<?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 Fahri Korkmaz on Medium]]></title>
        <description><![CDATA[Stories by Fahri Korkmaz on Medium]]></description>
        <link>https://medium.com/@r4bb1t?source=rss-10ba5c21b52------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*MZTvTwZDBp8FD1pUOYBjAQ.jpeg</url>
            <title>Stories by Fahri Korkmaz on Medium</title>
            <link>https://medium.com/@r4bb1t?source=rss-10ba5c21b52------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 00:19:44 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@r4bb1t/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[Azure Discovery and Recon Cheatsheet]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-snippet">Azure Tenant Discovery</p><p class="medium-feed-link"><a href="https://infosecwriteups.com/azure-discovery-and-recon-cheatsheet-d2ffb6b115aa?source=rss-10ba5c21b52------2">Continue reading on InfoSec Write-ups »</a></p></div>]]></description>
            <link>https://infosecwriteups.com/azure-discovery-and-recon-cheatsheet-d2ffb6b115aa?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/d2ffb6b115aa</guid>
            <category><![CDATA[security]]></category>
            <category><![CDATA[penetration-testing]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[red-team]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Fri, 18 Apr 2025 08:29:53 GMT</pubDate>
            <atom:updated>2025-04-18T08:29:53.025Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[The relevance of Azure Pentesting]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://systemweakness.com/the-relevance-of-azure-pentesting-1c229a377ca2?source=rss-10ba5c21b52------2"><img src="https://cdn-images-1.medium.com/max/1920/1*6KmIG7NjAtK356GLUu6CSA.jpeg" width="1920"></a></p><p class="medium-feed-snippet">Businesses in today&#x2019;s digital landscape are increasingly shifting their workloads to the cloud. One of the cloud providers used by these&#x2026;</p><p class="medium-feed-link"><a href="https://systemweakness.com/the-relevance-of-azure-pentesting-1c229a377ca2?source=rss-10ba5c21b52------2">Continue reading on System Weakness »</a></p></div>]]></description>
            <link>https://systemweakness.com/the-relevance-of-azure-pentesting-1c229a377ca2?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/1c229a377ca2</guid>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[red-team]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[penetration-testing]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Tue, 15 Apr 2025 18:22:26 GMT</pubDate>
            <atom:updated>2025-04-16T13:59:22.354Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[How a Chinese APT used Visual Studio Code for Command & Control]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://infosecwriteups.com/how-a-chinese-apt-used-visual-studio-code-for-command-control-d7ff66ba5644?source=rss-10ba5c21b52------2"><img src="https://cdn-images-1.medium.com/max/1094/1*qMpGfDEXAHeSTsQz1LEXlw.png" width="1094"></a></p><p class="medium-feed-snippet">As described in Unit42&#x2019;s analysis, Chinese APTs have used Visual Studio Code&#x2019;s tunnel feature to establish reverse shells.</p><p class="medium-feed-link"><a href="https://infosecwriteups.com/how-a-chinese-apt-used-visual-studio-code-for-command-control-d7ff66ba5644?source=rss-10ba5c21b52------2">Continue reading on InfoSec Write-ups »</a></p></div>]]></description>
            <link>https://infosecwriteups.com/how-a-chinese-apt-used-visual-studio-code-for-command-control-d7ff66ba5644?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/d7ff66ba5644</guid>
            <category><![CDATA[hacks]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[penetration-testing]]></category>
            <category><![CDATA[red-team]]></category>
            <category><![CDATA[blue-team]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Mon, 16 Sep 2024 19:53:21 GMT</pubDate>
            <atom:updated>2024-10-06T17:09:45.531Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Web Application Security]]></title>
            <link>https://systemweakness.com/web-application-security-0b729ada86db?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/0b729ada86db</guid>
            <category><![CDATA[udemy]]></category>
            <category><![CDATA[web-app-security]]></category>
            <category><![CDATA[penetration-testing]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Fri, 06 Sep 2024 17:26:52 GMT</pubDate>
            <atom:updated>2024-11-08T17:07:42.227Z</atom:updated>
            <content:encoded><![CDATA[<p>In 2024, web application security remains a cornerstone of modern cybersecurity, protecting digital platforms from a growing spectrum of online threats. With the increased reliance on web-based services for everything from financial transactions to social networking, the security of these applications is critical. Data breaches, identity theft, and hacking attempts continue to rise, making robust security measures essential for businesses, governments, and individuals alike. Attack vectors like SQL injection, cross-site scripting (XSS), and session hijacking pose constant threats, as cybercriminals refine their tactics to exploit vulnerabilities in web applications.</p><p>For cybersecurity professionals, mastering web application security is crucial because it equips them with the skills to defend against these attacks and fortify applications against exploitation. With organizations facing high stakes — financial loss, reputation damage, and legal consequences — students who understand how to secure web applications will be highly sought after in the job market. As applications move increasingly toward the cloud and as mobile usage grows, learning web security ensures that cyber security professionals are prepared for the current and future demands of the industry.</p><p>In essence, mastering web application security in 2024 not only offers a strategic advantage in the cybersecurity field but also provides professionals with the knowledge needed to protect critical online infrastructures against a backdrop of rising cyber threats.</p><p>In this article, I want to go over three common web application vulnerabilities, that might have a critical impact on the confidentiality, integrity, and availability of a web app’s data. However, there are much more web vulnerabilities. If you want to learn more about web application security, check out my new free course on Udemy “<a href="https://www.udemy.com/course/common-web-vulnerabilities-and-how-to-exploit-them/">Common Web Vulnerabilities and How to exploit them</a>”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s9W3TuOT3daJ3zNZFoIKIw.jpeg" /></figure><h3>Insecure Direct Object Reference (IDOR)</h3><p><strong>Insecure Direct Object Reference (IDOR) vulnerabilities</strong> are a type of access control flaw that occurs when an application provides direct access to objects (such as database records, files, or resources) based on user-supplied input, without properly verifying that the user is authorized to access them. This vulnerability allows attackers to manipulate input values (like object IDs, filenames, or other identifiers) to access data or perform actions that they should not have permission to access.</p><p>For example, if a web application URL includes a user ID like https://example.com/user/123, an attacker might change the ID to 124 and access another user&#39;s account if proper authorization checks are not in place. IDOR attacks can lead to exposure of sensitive data, unauthorized actions, and data tampering.</p><p>The impact of an Insecure Direct Object Reference (IDOR) vulnerability can be severe, depending on what type of data or resource is exposed and how an attacker can exploit the weakness. One significant impact is the unauthorized access to sensitive data. For example, if an attacker manipulates a URL or API parameter to access a different user’s data, such as changing an ID in a URL from /user/123 to /user/124, they might gain access to personal information like names, addresses, or even financial or medical records. This kind of data exposure can lead to privacy violations, identity theft, and compliance failures with regulations like GDPR or HIPAA.</p><p>IDOR vulnerabilities also pose the risk of privilege escalation. If an attacker can access higher-privileged accounts, such as admin interfaces, through ID manipulation, they may be able to take over accounts, compromise the entire system, or gain control of sensitive organizational functions. This type of exploitation can lead to full system compromise, where an attacker could carry out actions as an administrator, including user management, service disruption, or theft of intellectual property.</p><p>IDOR vulnerabilities are often found in poorly implemented access control mechanisms, especially in APIs or web applications with insufficient checks. They highlight the importance of robust authorization mechanisms and proper security validation of user input.</p><h3>Cross-Site Scripting (XSS)</h3><p><strong>Cross-Site Scripting (XSS)</strong> is a web application vulnerability that allows attackers to inject malicious scripts into web pages viewed by other users. The attack occurs when a website or web application does not properly sanitize user input, enabling attackers to inject client-side scripts — usually JavaScript — into the content that other users will see and interact with. When another user visits the compromised page, the malicious script is executed in their browser, allowing the attacker to steal sensitive data, manipulate the page content, or perform actions on behalf of the victim.</p><p>The consequences of an XSS attack can be wide-ranging and severe. One of the most common impacts is <strong>session hijacking</strong>, where an attacker can steal a user’s session cookie and use it to impersonate the user on the website, gaining access to their account. This can lead to unauthorized actions, theft of sensitive data, or fraudulent transactions. Another serious impact is <strong>data theft</strong>, where the attacker uses malicious scripts to access sensitive information stored in the browser, such as login credentials or personal data.</p><p>Preventing XSS vulnerabilities requires developers to properly <strong>sanitize</strong> and <strong>validate</strong> all user input before rendering it in a webpage. This means ensuring that any data entered by users — whether through forms, URL parameters, or other sources — is properly escaped or encoded to prevent it from being executed as code. Web applications should also implement strong <strong>Content Security Policies (CSPs)</strong>, which help prevent browsers from executing unauthorized scripts, even if they are injected. Using security frameworks and libraries that handle input validation and output encoding can significantly reduce the risk of XSS.</p><p>In summary, cross-site scripting is a dangerous and common vulnerability that allows attackers to execute malicious scripts in the context of a user’s browser. XSS can lead to data theft, account compromise, and manipulation of web content. Preventing XSS requires careful input handling, validation, and secure coding practices.</p><h3>SQL Injection (SQLi)</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*QS4TXq_ETVcXrWwNM3qgnQ.jpeg" /></figure><p><strong>SQL Injection (SQLi)</strong> is a type of security vulnerability that occurs when an attacker manipulates a web application’s SQL queries by injecting malicious SQL code into input fields. SQL Injection exploits the improper handling of user input in database queries, allowing the attacker to gain unauthorized access to the underlying database, manipulate its contents, or even bypass authentication mechanisms. This vulnerability arises when user-provided input is directly included in SQL queries without proper sanitization or validation.</p><p>The consequences of an SQL Injection attack can be severe. It can lead to <strong>Data Breaches</strong>, where attackers can extract sensitive information from the database, such as usernames, passwords, credit card numbers, or personally identifiable information (PII).</p><p>Furthermore, <strong>data manipulation</strong> can also occur. In some cases, attackers can modify or delete database records, causing data loss or corruption. They could also create new records, potentially inserting malware or fake users into the system.</p><p>In summary, SQL Injection is a dangerous and common web vulnerability that allows attackers to manipulate database queries, leading to unauthorized data access, modification, or system compromise. Developers can prevent SQLi by adopting secure coding practices such as parameterized queries, input validation, and minimal database privileges.</p><h3>Wrap-Up</h3><p>These are just a few of many web vulnerabilities. To fully understand web application vulnerabilities a cyber security professional needs knowledge of web app technologies. As penetration testing is a very hands-on skill, professionals also need a way to practice their skills.</p><p>If you want to learn the fundamentals of web app technologies and get to know more web application vulnerabilities by doing hands-on labs, you can check out my new free course “<a href="https://www.udemy.com/course/common-web-vulnerabilities-and-how-to-exploit-them/">Common Web Vulnerabilities and How to exploit them</a>”.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0b729ada86db" width="1" height="1" alt=""><hr><p><a href="https://systemweakness.com/web-application-security-0b729ada86db">Web Application Security</a> was originally published in <a href="https://systemweakness.com/">System Weakness</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Bsides Munich 2023]]></title>
            <link>https://r4bb1t.medium.com/bsides-munich-2023-5a3f5c2285bb?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/5a3f5c2285bb</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[conference]]></category>
            <category><![CDATA[munich]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Wed, 25 Oct 2023 08:27:22 GMT</pubDate>
            <atom:updated>2023-10-25T08:27:22.269Z</atom:updated>
            <content:encoded><![CDATA[<p>I recently had the pleasure of attending the <a href="https://2023.bsidesmunich.org/">Bsides Munich</a> conference this year (2023). It is an independently organized security event in Munich, Germany. <a href="http://www.securitybsides.com/">Security Bsides</a> events are community-driven conferences, which are hosted all over the world. So there might be one near to your place of living!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RrsHBobzZhGF9xOzbhc1dQ.jpeg" /><figcaption>Got some gifts :p</figcaption></figure><p>The event started on 14/10/2023 with workshops and ended on 15/10/2023 with the conference containing the talks. The workshops were hosted at <em>Munich University of Applied Sciences</em> and the conference took place at <em>The Westin Grand Munich</em>. Unfortunately, I was not able to attend the workshops, but I know that <a href="https://www.linkedin.com/in/claudia-ully/">Claudia Ully</a> from <a href="https://www.nviso.eu/">NVISO</a> did a very interesting workshop on Mobile Penetration Testing, which was called “<a href="https://2023.bsidesmunich.org/workshops/003-01_MWSLBH_the-hitchhacker-s-guide-to-the-mobile-galaxy/">The Hitchhacker’s Guide to the Mobile Galaxy</a>”. I really would have loved to participate in that workshop, because I’d also like to expand my Penetration Testing knowledge on Mobile Apps. However, I am pretty sure there will be an opportunity to attend Claudia’s workshop anytime soon.</p><p>The conference started with a single keynote by Dr. Mario Heiderich and continued with at least two talks in parallel. The conference ended with a closing keynote by Ana Oprea. You can get the full agenda and the slides of the talks <a href="https://2023.bsidesmunich.org/agenda/">here</a>. Furthermore, during the conference, there was also a hardware village, but unfortunately, I did not have the time to attend it :| Also, there were free snacks in the morning and lunch in the afternoon. The event organizers are listed on their <a href="https://2023.bsidesmunich.org/organizers/">web page</a>. There were multiple organizations sponsoring the event, including <a href="https://www.nviso.eu/">NVISO</a>, <a href="https://www.medel.com/">MED-EL</a>, etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*unu1UDsJr5BvljfFM0f5ZA.jpeg" /><figcaption>Some snacks during the conference</figcaption></figure><h3>Talks</h3><p>There were a lot of amazing talks, I basically was listening to talks all day long. There were also other great talks, which unfortunately I could not attend — sometimes you just wish to have more than 24 hours per day :| However, <a href="https://www.linkedin.com/in/lisihocke/">Lisi</a> attended other talks, which I did not, and made a <a href="https://www.lisihocke.com/2023/10/askappsec-bsides-munich-2023.html">blog post</a> about it, including <a href="https://mastodon.social/@lisihocke/111246266722202567">her notes</a>, which are really beautiful! She also attended the Workshop Day and shared her experience. You should definitely check her out.</p><p>In this section of the article, I’d like to share notes, which I have taken during the talks. I’ll also share what I liked and disliked about the talks.</p><h4>The Seven Sins. And Virtues. Of IT Security. And how they affect our world.</h4><p>The day started with none other than the cyber-high-priest himself, Dr. Mario Heiderich from <a href="http://cure53.de/">Cure53</a>. His talk was themed based on religion, which he also apologized for. However, I think that was not really necessary, as the theming made it unique, and nice to listen to and understand.</p><p>He mapped the seven sins, Lust, Gluttony, Greed, Sloth, Wrath, Envy, and pride, as well as the seven virtues, chastity, temperance, charity, diligence, patience, kindness, and humility to cyber security.</p><p><strong>The sins</strong></p><p><em>Lust </em>in cyber security for him means that someone has the desire to access and control data, e.g. by unethically exploiting systems. Examples of the sin of <em>gluttony </em>would be the hoarding of a lot of data or over-consumption of resources. Also not spending money where it is needed results in that sin. In the case of <em>greed</em>, someone tries to achieve more profit at the expense of other things. This happens a lot in cyber, for example when managers cut security budgets. But another interesting habit, that results in this sin, is when just classic checkbox pentesting is done — just to say “We have done a pentest”, which shows that actual security is not a priority.</p><p>If you’re lazy or lack due diligence in maintaining systems, then you committed another sin: <em>Sloth</em>. Every time uncontrolled anger is faced, this means the sin of <em>Wrath </em>is committed, such behavior is observed in Hackbacks and other destructive actions with IT. <em>Envy </em>or <em>jealousy </em>is seen in Cyber in the case of industrial espionage, or just thinking that you are not an interesting target. The sin of <em>pride </em>gets committed, if you have a big ego and overconfidence in your security posture.</p><p><strong>The virtues</strong></p><p>You achieve the virtue of <em>Chastity</em> if you use tech and data in an ethical and responsible manner. <em>Temperance </em>can be achieved in a balanced approach to technology adoption. You achieve <em>Charity </em>by creating and sharing secure code. If you want to achieve <em>Diligence </em>then you should stay up-to-date with the latest security threats and defenses.</p><p>I think the next virtue is very important: <em>Patience. </em>Security of organizations is everyone’s job, not just the job of security professionals. Security is always depending on the weakest link — that’s often the human factor. I’ve worked in Blue Teaming, and it is always the same: Only because someone clicked on a phishing link, security folks (SOC Analysts, IT Admins, Incident Responders, etc.) assume the victim is dumb. But it is not like that. Social Engineering victims become victims because they are <strong>human</strong>! So always be patient with users and expect them to make mistakes. I would even say, that if an organization gets ransomwared because the adversary gained an initial foothold via Phishing, then that’s mainly the security folks&#39; fault. Because there are much more points of failure, that got triggered before the objective of the ransomware operators has been achieved. And most of the time until objectives are achieved, there are many possibilities to catch and stop the adversary. It is always important to pursue a defense-in-depth approach. My philosophy is: I will blame myself first, before blaming others.</p><p>If you achieve <em>patience</em> you will have it easy to achieve <em>kindness</em>, by treating your users with respect. And also take your ego and kick it out of the door, so you can achieve <em>humility</em>.</p><p>By taking small steps towards the virtues and away from the sins, you will be able to achieve greater security and be a better security professional. But also keep in mind, that it is only possible to build security together.</p><h4>(In)direct Syscalls: A journey from high to low</h4><p>This talk was done by Daniel Feichter (Founder <a href="https://redops.at/">RedOps</a>) and you can find him on Social Media via the handle @VirtualAllocEx. BTW: RedOps has an awesome <a href="https://redops.at/en/knowledge-base">knowledge base</a>, where you can find more content by Daniel. Furthermore, the contents of the talk can be found in <a href="https://github.com/VirtualAlllocEx/DEFCON-31-Syscalls-Workshop/tree/main">Daniel’s GitHub Repository</a>. I really liked that talk because it was very technical and extremely relevant for any Red Teamer out there.</p><p>A normal program, e.g. your browser, executes in user space. However, sometimes these programs need access to lower-level functionality, for example, to save files on the disk. User Space programs can’t directly access low-level functionality. They have to do it via a syscall to switch to the kernel space. Each syscall has its unique ID, which is retrieved through NTDLL and is a process to get temporary privileged access. Syscalls are very interesting for Red Teamers, especially to execute shellcode with higher privileges or to dump credentials from lsass.exe.</p><p>Because syscalls can be used for malicious purposes they are heavily monitored through EDRs by hooking them. The following hooking techniques are used: SSDT Hooking, Inline API hooking, and Import Address Table Hooking. Often inline hooks are used in NTDLL, however other DLLs, such as Win32.dll, get hooked by EDR. Still, it is not possible for EDRs to hook everything, because it will result in performance issues, that’s why EDR hooking mainly targets API calls, such as NtAllocMemory. You can identify hooks with a debugger because hooks contain unconditional jump instructions.</p><p>The talk introduced 4 types of loaders that can bypass security measures depending on the use of hooking. In the following, we will look into them.</p><p><strong>Win32-API Loader</strong></p><p>This loader mainly uses the Windows API. First of all, you declare your shellcode, which will be executed later in a new thread. After that, the calling process will allocate memory with VirtualAlloc. The shellcode gets copied over to the newly allocated memory. Finally, a new thread gets created, which executes the shellcode. When the execution is finished the main thread will finish.</p><p><strong>Native-API Loader</strong></p><p>This loader doesn’t use the high-level API calls from the Windows API, but instead uses the Native APIs, which is considered medium level. It is done by directly accessing the Native APIs in ntdll.dll.</p><p>First of all, because these headers can’t be directly accessed, it is necessary to manually define the structure of the functions and manually load them. Afterwards, they can be used to execute the shellcode, by using NtWriteVirtualMemory, to allocate memory for the shellcode and NtCreateThreadEx to execute the shellcode.</p><p><strong>Direct Syscall Loader</strong></p><p>This low-level method directly uses syscalls without accessing NTDLL. The syscall stub is in the .text section of the assembly and the pointer to functions have to be defined manually. From there the malware will use NtAllocateVirtualMemory, NtWriteVirtualMemory, and NtCreateThreadEx to finally execute the payload. Even though you will bypass user-mode hooks in ntdll.dll, EDRs will most likely still catch you, because they use Event Tracing for Windows. This sequence uses very odd return addresses, which can be fixed via indirect syscalls.</p><p><strong>Indirect Syscall Loader</strong></p><p>This method will execute the syscalls in memory of ntdll.dll. In order to do that, first the malware has to retrieve a handle to ntdll.dll. After that, it gets the start address of the Native APIs inside ntdll.dll. When adding the appropriate offset to the Native API pointer, it is possible to get the memory address of the syscall instruction. Which can be used to execute the syscall and run the payload.</p><p>With indirect syscalls, it is possible to bypass user-mode hooks and checks based on Windows Event Tracing because the call stack is much more legitimate. But they can still be detected, because the entire call stack has to be spoofed, not such the syscalls. This can be done with Obfuscation/Encryption of functions, and import tables, as well as through thread call stack spoofing, which was recently introduced in a <a href="https://youtu.be/_2lH90C2nOM?si=Ko-AdAUr9uKaWqFz">Defcon talk by Alessandro Magnosi</a>.</p><h4>Cracking the chaos ransomware family</h4><p>This talk was done by Alexander Andersson from <a href="https://de.truesec.com/">Truesec</a>. You can watch the full talk on <a href="https://youtu.be/z_6ldJE7klE?si=Cv6AF7kadWxZx9fc">YouTube</a>.</p><p>On the Darknet Forum “XSS”, this malware was first released with a GitHub link containing the source code. The authors claimed it to be ransomware, however, the encryption was implemented in Version 2, and in previous versions, it was just a Wiperware. This ransomware has been deployed in a steel plant in Ukraine but is also used by people trying to get Roblox money in exchange for the decryption keys. Alexander said in his talk, that cyber attacks that use this malware, range from kids playing around to cyber war. However, maybe threat actors use Roblox because they can <a href="https://www.forbes.com/sites/cyrusfarivar/2023/03/30/hundreds-of-roblox-users-may-be-engaged-in-money-laundering-court-filing-says/">easily launder money with the game</a>?</p><p>The problem with the ransomware family is that they use the normal random function from C# to generate a seed for the encryption algorithm. However, this is not secure if your security relies on true randomness. Furthermore, the ransomware uses weak cryptographic modes, which makes them vulnerable to plaintext attacks. Based on these findings Truesec was able to create decryptors, which can be downloaded from their <a href="https://github.com/Truesec/TSDecryptors">GitHub Repository</a>. However keep in mind, that some variants of the malware have a persistence mechanism, so you should remove that before running the decryptor because otherwise your files will get encrypted again.</p><p>This is a really inspiring story. Truesec had exposure to this malware because they investigated an incident. During that incident, they were able to create a decryptor in 1.5 weeks, which is really amazing! Imagine being the victim and a cyber security company that comes to help you is able to fully recover your files by cracking the ransomware — that’s really cool!</p><h4>Let me do it for you — Automating OSINT and Recon</h4><p>This talk was done by Paul Zenker, you can find him on Twitter through the handle @Secbyaccident.</p><p>The talk focused on tools you can use during OSINT which help you to automate the time-consuming task of collecting intelligence. The OSINT cycle starts with publicly accessible information, such as from the Internet. That data has to be collected, then cleaned, and finally stored. Afterward, you can play with the data during the data presentation phase and extract useful intelligence.</p><p><strong>Python Libraries</strong></p><p>Useful Python libraries are requests , shodan , and regex . If you want to automate Google dorking, you should try out DuckDuckGo, as it has a Python library duckduckgo_search and is basically like Google dorking but on steroids. Furthermore, you can create bots as Python has libraries to communicate with the Discord or Slack API. This is also pretty useful for Bug Bounty hunting, for example, to start a DNS enumeration on a program. If you prefer to build your own API, then Paul recommended using Flask. However, I would recommend you use FastAPI , which makes building Rest APIs easy and creates swagger documentation automatically. In case you have to parse data from HTML, then use Beautifulsoup, or Selenium, if it is a Single Page application, that gets built via JavaScript.</p><p><strong>Data Storage</strong></p><p>For storing data, you can use SQL databases such as MongoDB or NoSQL ones such as MongoDB. However, some data is best represented in graph format, for example, social networks. In that case, it is recommended to use a graph database, such as Neo4J. For time series data, InfluxDB can be used. If you want to build dashboards on top of your data, then use Grafana. If you just need a fast key-value store, then use Redis.</p><p><strong>Deployment</strong></p><p>Your programs can be deployed in the cloud, for example by utilizing Docker. In order to manage your docker instances, you can use Portainer. Also, a lot of database providers have a free tier. You might also use Cloud services, such as AWS or Azure, but keep in mind, that it might push you towards poverty if you configure it wrong :)</p><p><strong>Artificial Intelligence</strong></p><p>Paul also mentioned artificial intelligence. ChatGPT can be used to assist you during coding. There are also other free language models at <a href="https://huggingface.co/">Huggingface</a>. I’d also would like to add, that ChatGPT is perfectly suited for building regular expressions. Also, I’d like to add that GPT-4 has the ability to extract certain information from unstructured data, such as text, and transform it into machine-readable format. It can also map certain things to more standardized definitions. For example attack techniques to the MITRE ATT&amp;CK framework.</p><p><strong>OSINT Tools</strong></p><p>Paul also mentioned cool OSINT Tools: Spiderfoot, ProjectDiscovery, n8n, Apache Airflow, BabyAGI, LangChain, Maltego, Netlas.io, search.t0.rocks, start.me, TomNomNom, Cipher387, and the GitHub Repository “<a href="https://github.com/awesome-selfhosted/awesome-selfhosted">Awesome selfhosted</a>”.</p><p>I think Paul&#39;s talk was very valuable in regard, to getting to know certain tools. But it would be also nice if he had talked about use cases that chain these tools to reach a certain objective. I’ve already given him the feedback, and he will include this next time.</p><h4>Bio-Lock: The future and ethics around DNA Cryptography</h4><p>This talk was done by a rockstar speaker: Tayla Micael Sellschop. You can reach out to her via <a href="https://www.linkedin.com/in/tayla-micael-sellschop-77059b148/">LinkedIn</a> or Email (<a href="/service/mailto:tayla@telspace.co.za">tayla@telspace.co.za</a>). She has an awesome and entertaining way of presenting. It was hard to believe to me, when she said, that this was her first talk. And I also think this was the <strong>most unique talk on Bsides</strong>. It is a very interesting and young area of research (the first paper is from 2017), which seems promising for the future. In the future, when quantum computers are available, it won’t be possible to use our current cryptographic methods, which rely on elliptic-curve cryptography. Hence new cryptography methodologies have to be researched, that are secure against the computing power of quantum computers. That’s also why I really enjoyed the talk.</p><p>DNA is a long chain of molecules consisting of Adenine, Thymidine, Guanine, and Cydosine. DNA is millions of years old and has been used to store information about us humans. It can be used to encrypt data through a DNA sequence and can hold a lot of data, which makes it also suitable for storage: One gram of DNA can hold up to 108 TB of data.</p><p>The researchers have implemented the encryption library with Java. The Bio Java implementation uses several more steps of a substitution cipher. DNA can be used both for symmetric and asymmetric encryption. However, it also has some flaws. Research on DNA encryption is lacking and is incredibly expensive because specialized skills (Biology, Computer Science, Cryptography) and laboratories are required.</p><p>Furthermore, DNA has the problem that it can be changed through certain diseases, such as cancer, hepatitis, or HIV. Furthermore, identical twins have the same DNA, which makes it less practical for cryptography and storage.</p><p>Also, there are some ethical issues. What if our DNA gets leaked in a breach? What if companies sell our DNA for financial gain and advertising? On the other hand, our DNA is everywhere…</p><h4>Rooting the Cradlepoint IBR600 and other Stories</h4><p>The research from this talk is from <a href="https://github.com/vegantransistor">Sebastien</a> (Email: <a href="/service/mailto:vegantransistor@mailbox.org">vegantransistor@mailbox.org</a>) and <a href="https://github.com/stulle123">Dawin</a> (<a href="https://twitter.com/dschmidt0815">Twitter</a>), while Sebastien has performed the talk at Bsides Munich. During the talk, they describe how they perform research on the <strong>Cradlepoint IBR600</strong> router.</p><p>The Cradlepoint IBR600 router has an in-build WiFi, LTE, and Modem, which allows for LAN &amp; WAN connection. Furthermore, it has a web server running as well as connected to a Cloud Service (Netcloud) for device management purposes, hence having a wide attack surface.</p><p>Previous related work was able to find hardcoded passwords, a privilege escalation vulnerability because of a backdoor account, as well as passwords that were encrypted by using a hardcoded key.</p><p>If you want to hack IoT devices, you have to look out for a UART interface. UART stands for Universal Asynchronous Receiver/Transmitter. It’s a hardware communication protocol used for serial communication between devices. Most of the time developers use UART for debugging purposes. In case of hardware hacking it is also very useful to get access to firmware. Furthermore, the device uses uboot as a boot loader, which is typical for embedded devices. Also, no secure boot was configured, so the researchers could modify environment variables, which helped them in further research. What was very unique for that device, is that they used a Python script for their middleware. The researchers have continued their research on the Python code.</p><p>The code was in bytecode, however, it can be easily decompiled to actual Python code. That script contained a function to enable silent mode during the boot process. Furthermore, there was also a custom, limited shell implementation with Python, called cpshell. The goal of the researchers was to patch the firmware to enable a root shell. They loaded an openWRT live image into the device in order to flash the manipulated firmware and get a root shell after reboot.</p><p>The next vulnerability that exploded was an insecure way of updating the firmware. Normally the firmware is encrypted, but the decryption key was on the root file system. So the researchers could recover the key. Furthermore, signature verification was skipped for versions below 7.0.0. Hence secure boot was completely broken because the version number can be specified by an attacker to load a malicious firmware on the device.</p><p>Next, they researched the cloud communication. The communication was encrypted with TLS. However as the researchers already had root access to the device, they could add a malicious Root CA to the trusted store, which enabled them to sniff the network communications with <a href="https://mitmproxy.org/">mitmproxy</a>.</p><p>In the communication, they could spot a Base64 encoded string, which contained a pickled stream. Pickle is a serialization library for Python, which is not secure and can be exploited for RCE, as long as you control the data stream — which the researchers did.</p><p>Furthermore, they could also find a registration vulnerability, which could allow a malicious actor to disconnect any device. The Python code had a function called insecure_activation() . This function was used to activate devices based on MAC addresses. An attacker could just use the MAC address of another device to connect to Netcloud, which disconnects the legitimate device.</p><p>You can find the full research on <a href="https://github.com/vegantransistor/Rooting-the-Cradlepoint-IBR600">GitHub</a>.</p><h4>Christmas Hancitor Campaign</h4><p>Do you know what would be nice? Detecting a cyber-attack before it even happens! In this talk, <a href="https://www.linkedin.com/in/artem-artemov-82796bb0/">Artem Artemov</a> from <a href="https://www.group-ib.com/">Group-IB</a> described an incident where they could prevent further damage to an organization by locating the Command and Control server just in time.</p><p>It started with a post on a DarkNet Forum, where a criminal tried to sell a SOCKS5 Backconnect Module. The analysts identified the systems via <a href="https://www.shodan.io/">Shodan</a> search. They could gain access to the admin panel of SystemBC, where they were able to identify victims. One of the victims was a Belgium company, hence they alerted the Belgium CERT, which started the incident response process.</p><p>During incident response, they were able to identify, that the threat actor is currently in the stage of lateral movement. The threat actor has already mapped the network and dumbed credentials, including admin credentials. They also tried to get insides into the network traffic with WinPcap. The treat actor started their actions on objective during the holiday season. First, they tried to remove the backups — which failed. Because of that, the threat actor sold their access on a DarkNet forum.</p><p>The second threat actor used Cobalt Strike for command and control. Because the C2 server was already known, it got blocked and the analysts could attribute the second threat actor to CUBA/Mlock ransomware operators.</p><p>This is a story where threat intelligence was really shining and helped shift the incident response process from investigating an attack to preventing an attack. Very inspiring story! If you want to read more about it, Group-IB has also published a <a href="https://www.group-ib.com/blog/christmas-miracle-incident-response/">blog post</a> on it.</p><h4>Honeypot Boo Boo: Better Breach Detection with Deception Inception</h4><p>This was a very interesting approach to detecting an adversary. The talk was done by Security Philosopher, Justin Varner. You can reach out to him via Email (<a href="/service/mailto:justin@radzen.io">justin@radzen.io</a>) and get further information via <a href="https://radzen.io/">this website</a>.</p><p>The goal is to set up honeytraps where an attacker might use them, which will trigger an alert and help detect the adversary with less alert fatigue — as someone who has worked in a SOC, I can for sure say that’s a real thing and I am grateful for any approach that helps to address that problem. In traditional Security Monitoring, you will monitor every endpoint and the network for malicious activity. This creates a lot of alerts and stress for SOC Analysts, why this role has a high rate of burnout and fluctuation. So it is necessary to make a low volume of high-fidelity alerts, where honeytraps are perfectly suitable. This helps to reduce the mean time to detect.</p><p>An example of a honeytoken would be an API token for example to AWS. This token won’t be used by anyone. If an attacker stumbles across the token, they will probably use it. When used an alert will be triggered. Other forms of tokens could be for Slack, GitHub, etc.</p><p>Another example is honeypots, for example, one that emulates a domain controller. Another one is by using domain names as honeytraps, so as to get alerted when they get queried. Web redirect tokens will redirect the attacker to a legitimate website, but will grab metadata, such as the user agent, source IP, etc. in the process before redirecting. Which might help attribute an adversary.</p><p>A very interesting way of using honeytokens is via a Wireguard VPN token, which will be placed on your phone. When an attacker uses that VPN, then you know that your phone got pwned. You can create such a token through <a href="https://canarytokens.org/generate">https://canarytokens.org/generate</a> for free.</p><p>Another interesting approach to active defense is by playing mind games. You can for example create a lot of database servers, that sound interesting but contain useless information, which frustrates the attacker. You can use LimaCharlie, which is a security automation platform, to send canary alerts and investigate them with it.</p><p>From a Red Teamers perspective, it is hard to detect such traps. You could look out for obvious patterns, or credentials/tokens that are known for certain honeytrap vendors.</p><h4>Security by Design</h4><p>In an interconnected world, it is very important to enable security by default. But how can it be achieved? <a href="https://www.linkedin.com/in/anaoprea/">Ana Oprea’s</a> talk goes into the details of Security by Design.</p><p>First of all security risks should be managed accordingly. Start with proper risk assessment. Be sure, that you might be a target and begin implementing the basics of security. Never underestimate your adversary and they aren’t always afraid of being caught. Furthermore implement security design strategies with the least privilege, for example by using Zero Trust.</p><p>If you want to design for least privilege, start with auditing and detecting flaws. You should review all access logs and justifications to ensure that they are appropriate for your organizational needs. But keep in mind that auditors might miss context and/or objectives.</p><p>Furthermore, shifts security and reliability to the left side. They are central elements in the architecture of systems.</p><p>Ana also mentioned some resources for further reading: “The Site Reliability Workbook”, “Site Reliability Engineering”, and “Building Secure &amp; Reliable Systems”.</p><h3>Conclusion</h3><p>I hope next time I will also be able to attend the workshops because there were really interesting ones. Nevertheless, I had a lot of fun during the talks. Most of them were done by very entertaining speakers, and there were very unique ones, as well as very technical ones, which I really appreciate! It was also an awesome event to network with like-minded folks. Even though it was a local event, it was still very international. People attended from the U.S., Turkey, Russia, Switzerland, Austria, and South Africa!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5a3f5c2285bb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My experience getting a Cybersecurity Job as a fresh graduate]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://r4bb1t.medium.com/my-experience-getting-a-cybersecurity-job-as-a-fresh-graduate-5f009f50ab6f?source=rss-10ba5c21b52------2"><img src="https://cdn-images-1.medium.com/max/640/1*yzaqCRigFnBbfTD0jdwwOg.jpeg" width="640"></a></p><p class="medium-feed-snippet">In this article I want to share my experience in getting a cybersecurity job after freshly graduating my master&#x2019;s in computer science.</p><p class="medium-feed-link"><a href="https://r4bb1t.medium.com/my-experience-getting-a-cybersecurity-job-as-a-fresh-graduate-5f009f50ab6f?source=rss-10ba5c21b52------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://r4bb1t.medium.com/my-experience-getting-a-cybersecurity-job-as-a-fresh-graduate-5f009f50ab6f?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/5f009f50ab6f</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[job-search]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[job-interview]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Fri, 06 Oct 2023 19:36:53 GMT</pubDate>
            <atom:updated>2024-11-20T22:07:24.073Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[What is the Cyber Kill Chain?]]></title>
            <link>https://r4bb1t.medium.com/what-is-the-cyber-kill-chain-83e7d4937f97?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/83e7d4937f97</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[cyber]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[it-security]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Thu, 31 Aug 2023 22:02:25 GMT</pubDate>
            <atom:updated>2023-08-31T22:02:25.505Z</atom:updated>
            <content:encoded><![CDATA[<p>The cyber kill chain is a model that describes the attack process on IT systems and identifies the stages that an attacker goes through to carry out a successful attack. It consists of the following phases:</p><ol><li>Reconnaissance: gathering information about the target to develop an attack plan.</li><li>Weaponization: The creation of malware or other attack tools.</li><li>Delivery: transferring the malware to the target, e.g. via email attachments or social engineering.</li><li>Exploitation: using a vulnerability in the system to gain access.</li><li>Installation: Install malware on the target system.</li><li>Command and control (C2): Transmission of commands and data to and from controlled systems.</li><li>Actions on objectives: execution of the attacker’s desired actions on the targets, e.g., data theft or system takeover.</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=83e7d4937f97" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What is Cloud Security?]]></title>
            <link>https://r4bb1t.medium.com/what-is-cloud-security-202444b2af86?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/202444b2af86</guid>
            <category><![CDATA[security]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[cloud]]></category>
            <category><![CDATA[it-security]]></category>
            <category><![CDATA[blue-team]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Sat, 01 Jul 2023 08:02:20 GMT</pubDate>
            <atom:updated>2023-07-01T08:02:20.521Z</atom:updated>
            <content:encoded><![CDATA[<p>First of all, what is the cloud anyway? The cloud is a model for delivering and accessing IT resources and applications over the Internet. Instead of being delivered locally on a single computer or local network, data and applications are delivered to a centralized environment called the cloud. Users can access IT resources and applications over the Internet without installing or storing them on their own computers.</p><p>Cloud security thus refers to the measures taken to ensure the security of data, applications, and infrastructures in cloud computing environments. This includes measures such as monitoring networks and data, controlling access, encrypting sensitive data, conducting reviews and audits, and responding to security incidents. The goals of cloud security are to prevent data loss and leakage, protect against attacks, and maintain the integrity, availability, and confidentiality of data.</p><p>To work in cloud security, the following skills are typically required:</p><ol><li>Deep understanding of networks and security technologies: Understanding how networks work and the various security technologies is essential for cloud security.</li><li>Knowledge of cloud technology: knowledge of the various cloud platforms and their architecture, as well as their security requirements and measures, is essential.</li><li>Experience in dealing with compliance and regulations: It is important to have knowledge in the various compliance and regulatory requirements for cloud services and data.</li><li>Knowledge of data privacy and data security: it is important to have knowledge of the various data privacy and data security regulations in order to be successful in cloud security.</li><li>Ability to analyze and assess risk: an important skill in cloud security is the ability to identify and assess risks and take appropriate action.</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=202444b2af86" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Cyber Security Trends 2023]]></title>
            <link>https://r4bb1t.medium.com/cyber-security-trends-2023-c30cf82a3089?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/c30cf82a3089</guid>
            <category><![CDATA[it-security]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[cyber]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Wed, 15 Mar 2023 09:02:11 GMT</pubDate>
            <atom:updated>2023-03-15T09:02:11.160Z</atom:updated>
            <content:encoded><![CDATA[<p>Cybersecurity experts are in demand like never before. With each passing year, more and more destructive hacks and vulnerabilities are coming out. To keep businesses safe in 2023, it’s important to stay on top of IT security trends. In this article, I go over 9 trends that will keep us busy this year.</p><h3>Web 3 Security</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*bv_DSNakeoOgb-WF.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/sicherheit-schutz-antivirus-265130/">Pixabay</a> by <a href="https://pixabay.com/de/users/pixelcreatures-127599/">Werner Moser</a></figcaption></figure><p>Web 3 is a vision for the future of the Internet, where data and content are decentralized. It aims to represent a new era of the Internet where users have more control over their data and online identity and enjoy greater privacy. It uses technologies such as blockchain and smart contracts to provide a more secure and trustworthy online experience.</p><p>Web 3 thus refers to the next generation of the World Wide Web, which is oriented toward decentralization, data protection, and responsibility. Unlike the current World Wide Web, where much of the data is collected and managed by large corporations, Web 3 will have a decentralized structure where data is managed and secured by users themselves. In addition, the interaction between users and applications will take place via blockchain technology, which should ensure greater transparency and data security.</p><p>Attacks on the blockchain of companies have already been observed, in which hackers were able to capture several million in cryptocurrency. Maintaining the security of Web 3 applications requires new skill sets, and since Web 3 has a trend curve similar to that of the Internet, it is advantageous to acquire these skills.</p><h3>Zero Trust</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/889/0*L1Ll3XTh_93F1kBz.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/teamgeist-zusammenarbeit-2447163/">Pixabay</a> by <a href="https://pixabay.com/de/users/anemone123-2637160/">Anemone123</a></figcaption></figure><p>Zero Trust is a concept where any access to systems and data, regardless of location or device, is accompanied by strict authentication and monitoring. The idea behind Zero Trust is that no network node or device is automatically trusted and that all access to data and applications must always be verified before it is approved. This is to help prevent data theft, malware attacks, and other threats.</p><h3>Cloud Security</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*yXpRdbaJfDmR2uVD.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/wolken-himmel-hell-tageslicht-1282314/">Pixabay</a> by <a href="https://pixabay.com/de/users/pexels-2286921/">Pexels</a></figcaption></figure><p>More and more companies are taking the path to the cloud. This allows companies to develop software quickly on the one hand and save costs on the other. However, this move also comes with its own pitfalls. This is because cloud systems have their own security vulnerabilities. For example, sensitive data can be leaked through the insecure configuration of an AWS S3 bucket. This is another reason why it may be worthwhile to familiarize yourself with cloud security in 2023. Attack vectors on cloud systems are listed in the Cloud Matrix of the MITRE ATT&amp;CK Framework.</p><h3>Artificial intelligence</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*RB1ZGMEWkVb0RDcv.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/k%c3%bcnstliche-intelligenz-gehirn-hirn-3382521/">Pixabay</a> by <a href="https://pixabay.com/de/users/geralt-9301/">Gerd Altmann</a></figcaption></figure><p>At the latest with ChatGPT, everyone has realized the potential of artificial intelligence. AI will be integrated more and more into society and will be found more and more often in software. And that’s where the problem lies because AI systems can still be insecure.</p><p>This is because AI systems still require software and infrastructure so that they can be developed and executed. These systems should be protected. But there are also attack methods that explicitly target AI systems. For example, if someone develops an AI that is used to write code. Then that AI also needs data. If the developer of the AI collects data from GitHub for training, then an attacker could simply create noisy repositories where there is a backdoor in the code. Then the AI learns that the code to be generated must also contain this backdoor. This type of attack is also called “Poison Training Data”.</p><p>For this reason, I think it will become more important in IT security to adequately secure Machine Learning models. The MITRE ATLAS framework shows further concrete attack vectors on AI systems.</p><h3>Supply Chain Attacks</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*XVelFpoUAQeX4Y4R.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/containerschiff-nacht-hafen-schiff-7383545/">Pixabay</a> by <a href="https://pixabay.com/de/users/makabera-6773476/">Markus Kammermann</a></figcaption></figure><p>The supply chain is a chain of vendors required for the manufacturing process of a product or service. An attack on the supply chain refers to any attempt to impair or compromise any of these components to cause a negative impact on the final product or service.</p><p>For example, a supply chain attack can occur when a hacker manipulates or infects a product in the supply chain before it is delivered to the end customer. In this way, an attacker can introduce malicious software on a computer and thus spy out data or compromise the system.</p><h3>State-sponsored hacking</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*mKEZR9nzygQUFSrJ.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/internet-sicherheit-hacker-online-3410923/">Pixabay</a> by <a href="https://pixabay.com/de/users/tumisu-148124/">Tumisu</a></figcaption></figure><p>Especially with the war between Russia and Ukraine, we have seen that hacking is also increasingly used by state actors such as intelligence agencies. Critical infrastructure has also already been attacked in various countries by state actors, so-called Advanced Persistent Threats. The war is exacerbating the situation, making attacks by APTs on businesses and government agencies increasingly likely. APTs have also been seen behind supply chain attacks, such as the attack on Solarwinds.</p><p>For organizations to protect themselves effectively against APTs, they must improve their cyber resilience. Resilience refers to the ability of a system to cope with disruptions and quickly return to normal functions. In IT security, this can be understood as resilience to cyber attacks. It includes measures to minimize the risk of failure and damage and to enable rapid recovery in the event of an incident.</p><p>Because it’s hard to protect yourself from APTs. After all, they have access to all the resources a state has and pretty much infinite time. Which makes these types of attackers very dangerous. For this reason, organizations need to change their mindset from Prevent-The-Breach to Assume-The-Breach. To find out if the measures, especially in the area of resilience, are good enough, companies should implement Red Teaming or Purple Teaming.</p><h3>Internet of Things</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*nfLyYzb8_DUVWe6v.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/system-netz-netzwerk-verbindung-2660914/">Pixabay</a> by <a href="https://pixabay.com/de/users/geralt-9301/">Gerd Altmann</a></figcaption></figure><p>With Industry 4.0, companies have been relying on IoT for years. However, IoT is now also increasingly being used in private homes and in public. States are trying to transform their cities into so-called smart cities, i.e. cities, using IoT. Of course, this also brings a high risk, as attacks on cities now pose a high threat to the population. Especially due to the increase in state attacks, attacks on IoT networks have become more likely.</p><h3>OT Security</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*w4fwNsQzlCarhOE4.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/fabrik-kraftwerk-landschaft-wolken-4338627/">Pixabay</a> by <a href="https://pixabay.com/de/users/randgruppe-12327121/">Randgruppe</a></figcaption></figure><p>Operational Technology, or OT for short, refers to the technology used to monitor and control industrial processes, productions, and infrastructures. These technologies are often used in the context of the Internet of Things (IoT) and include sensors, actuators, control systems, and networks. OT is an important part of industrial automation and has a major impact on the efficiency, reliability, and safety of industrial processes.</p><p>Often, critical infrastructure is operated through OT. Since attacks by state hackers are becoming increasingly likely and these hackers also like to target OT. It is therefore important that companies and critical operators secure their OT infrastructure as well as possible.</p><p>Many IT security experts specialize in normal IT infrastructure, which means that experts for OT security are desperately needed. Therefore, it is a good strategy to educate yourself in the direction of OT security.</p><h3>Ransomware</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*4-H1ORMUmx1TRA9p.jpg" /><figcaption>Picture from <a href="https://pixabay.com/de/photos/vorh%c3%a4ngeschloss-gesperrt-gesichert-428549/">Pixabay</a> by <a href="https://pixabay.com/de/users/jarmoluk-143740/">Michal Jarmoluk</a></figcaption></figure><p>Ransomware is a type of malware that prevents access to files on a computer or network by encrypting them. The attacker then demands a ransom, called a ransom, to release the data. This can have a significant impact on business operations and data protection for affected companies or individuals.</p><p>Ransomware has caused many cybersecurity experts sleepless nights in recent years. This year, ransomware attacks will certainly not stop. Because many hacker groups earn big money with it.</p><p>For this reason, securing systems against ransomware will become an important topic in 2023. Fighting cybercrime through law enforcement investigations is also likely to increase this year.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c30cf82a3089" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Analysis and Modeling of U.S. Cyber Security Salaries]]></title>
            <link>https://systemweakness.com/analysis-and-modeling-of-u-s-cyber-security-salaries-b1de03afb2d2?source=rss-10ba5c21b52------2</link>
            <guid isPermaLink="false">https://medium.com/p/b1de03afb2d2</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[salary]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[pandas]]></category>
            <dc:creator><![CDATA[Fahri Korkmaz]]></dc:creator>
            <pubDate>Mon, 20 Feb 2023 12:10:58 GMT</pubDate>
            <atom:updated>2023-04-06T13:15:28.159Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*ZHMvfK3hYxQiSFBGLyyIUQ.jpeg" /><figcaption>Photo by <a href="https://www.pexels.com/de-de/@tima-miroshnichenko/">Tima Miroshnichenko</a> from <a href="https://www.pexels.com/">Pexels</a></figcaption></figure><p>It is believed that Cyber Security professionals have one of the highest salaries in the Information Technology industry. But is that true? And what is realistically expected salaries in the Cyber Security domain? What factors affect the salary of Cyber Security professionals? If you ever asked yourself these questions, then you came to the right place!</p><p>This article will analyze U.S. Cyber Security Salaries and build a machine learning model to predict expected salary based on experience level and company size.</p><p>The dataset can be found at <a href="https://infosec-jobs.com/salaries/download/">salaries.infosec-jobs.com</a> and is licensed under CC0. infosec-jobs.com tries to “make global salaries in […] Cybersecurity […] more open and transparent” [1].</p><h3>Exploratory Data Analysis</h3><p>For exploratory data analysis, the Python libraries Pandas, Matplotlib, and Seaborn were used.</p><h4>Company Location</h4><p>The dataset contains data from various countries. But most of the data, around 78%, is from the U.S. That is why this article focuses more on salary data from the United States. To get a list of top 10 countries, which are represented in this dataset and plot them in a bar chart, the following command can be used:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/31de18306d10aab01e81785ce2a99c4c/href">https://medium.com/media/31de18306d10aab01e81785ce2a99c4c/href</a></iframe><p>First of all, the command will group by company location and then count the values for each location. The values are then sorted with sort_values in ascending order. After that the head method is used to get the first 10 elements, which are the top 10 countries represented in the dataset. In order to get only one column the first command ends with iloc[:, 1] . Finally the plot method from pandas is used to plot a bar chart and set x-Axis, y-Axis description, and title. The command results in this bar chart, clearly show that most data points are from U.S.-based Cyber Security professionals. For the pie plot, I have set kind=&quot;pie&quot; and ylabel=&quot;&quot; .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/580/1*LTgTDo-zCAZexcx1SPY_yw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/389/1*H42NVoS2_DvjCebjvh0Y2A.png" /></figure><h4>Filtering data</h4><p>Next, the data will be filtered, so it only includes data from the U.S. The company location is called “US”. So the filtering can be done with the following command:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/804a058d84f656208f7c18cd288b4573/href">https://medium.com/media/804a058d84f656208f7c18cd288b4573/href</a></iframe><h4>Job Titles</h4><p>The dataset contains a column called “job_titles”. It describes the role of the professional, e.g. “Security Analyst”. There are 94 unique job titles in the dataset. But when we count every job title, with value_counts , we get the following output:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/476/1*qwCzMOm_rlFXqeFTnjdQfA.png" /></figure><p>It seems like there are multiple job titles, that represent the same role. For example “Security Analyst” and “Cyber Security Analyst”. So we have to clean the data if we want to work with the job titles column. In order to clean the data, I have created the following function. It takes a role and assigns it to a more generic job title.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/15546720c3972d530b477394bdc0124d/href">https://medium.com/media/15546720c3972d530b477394bdc0124d/href</a></iframe><p>After defining the clean-up function, it can be applied to the job_title column by picking that column with loc and calling the apply method:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/5a85bccef9dc567ce2002e5e52551d22/href">https://medium.com/media/5a85bccef9dc567ce2002e5e52551d22/href</a></iframe><p>Finally, there are only 15 unique job titles left.</p><h4>Employment Type</h4><p>The dataset contains a column called “employment_type”. This column has 4 possible values [2]:</p><ul><li>FT: Full Time</li><li>PT: Part Time</li><li>CT: Contract</li><li>FL: Freelance</li></ul><p>When values for each employment type are counted, then it is clear to see that most of the data is from full-time employees:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/452/1*IRNNFfLLYiw3I1ZxujB5vQ.png" /></figure><p>This means we can discard Part-Time, Contract, and Freelance employment from further analysis, as they do not possess any statistical relevancy.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/eab37bd3f435b8ee1b13e5e7df4fa1e9/href">https://medium.com/media/eab37bd3f435b8ee1b13e5e7df4fa1e9/href</a></iframe><h4>Experience Level</h4><p>The dataset contains another column called “experience_level”. That column has the following possible values [2]:</p><ul><li>EN: Entry-level / Junior</li><li>MI: Mid-level / Intermediate</li><li>SE: Senior-level / Expert</li><li>EX: Executive-level / Director</li></ul><p>After plotting experience level and salary in a bar plot, it is clear, that both variables correlate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*DyeX86QOrYW2KWvCLg6F1A.png" /></figure><h4>Company Size</h4><p>The dataset contains a column called company size, which describes the size of the company with the values S (less than 50 employees), M (50 to 250 employees), and L (more than 250 employees) [2]. When we plot the company size and the salary, we see that people at medium-sized companies will get higher pay.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*IHa_w4plqGXD6YouIdesyQ.png" /></figure><p>When making an observation like this it is also important to find the cause. Because it might be that in this dataset there are more Executive Level professionals working at medium-sized companies. That’s why I have created the following bar plot setting the hue to “experience_level”. I have used seaborn to create the plot:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/27ec0f59eec9dbead81aa418d624c724/href">https://medium.com/media/27ec0f59eec9dbead81aa418d624c724/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*HEOepUHPRtcs7h1f95mNsQ.png" /></figure><p>After seeing the second plot, it is safe to assume that Cyber professionals at medium-sized companies tend to have higher pay, as long as the professional is Junior or Mid experienced. As an Executive or Senior professional, the salary at Large sized companies is higher.</p><h4>Remote Ratio</h4><p>Another column in the dataset is “remote_ratio”. This column consists of 3 possible values and describes how much remote work the professional is allowed to do [2]:</p><ul><li>0: No remote work (less than 20%)</li><li>50: Partially remote</li><li>100: Fully remote (more than 80%)</li></ul><p>In the next steps, we can plot remote_ratio and the mean salary for each group. There is no clear evidence that remote work affects salary.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/5a516b235378a3c92a92d1354a64709b/href">https://medium.com/media/5a516b235378a3c92a92d1354a64709b/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*scgXFi_6k6CjCD4q9YPbkw.png" /></figure><p>The next step should also take the experience level into consideration:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/4411e6f50c088ad3fe6095f7048c3792/href">https://medium.com/media/4411e6f50c088ad3fe6095f7048c3792/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*eO0stDDclcCfjHJ9MDDZrw.png" /></figure><p>This plot shows, that higher experienced professionals will more likely work remotely.</p><h4>How have the salaries changed between 2020 and 2023?</h4><p>In order to answer the question, I first calculated the mean salary for each year and plotted it in a line chart.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/4f4e3e4655d52e8e20c32bcaeede267a/href">https://medium.com/media/4f4e3e4655d52e8e20c32bcaeede267a/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*aHJrO-QT3OTKbuP-KJJmxQ.png" /></figure><p>It is safe to assume, that Cyber Security Salaries are on the rise. However, to be sure that this effect is not because the dataset contains more data in 2022 / 2023 from senior experienced professionals, I have created a Barplot of the mean salary per year and experience level. From the plot we see that the observation from the previous plot is not because of the dataset:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/1e9e5ee81c8bf6b7ae9700fa1bda0d58/href">https://medium.com/media/1e9e5ee81c8bf6b7ae9700fa1bda0d58/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*8VCFw-LmBfDlwHvULqzRxQ.png" /></figure><h4>Which Cyber Security roles get paid the highest salary?</h4><p>To answer this question, I have grouped them by job title, calculated the mean of the salaries, and sorted it in descending order. Finally, the result is plotted with a bar chart.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/1def0bbf036f145cf4e4d8034e2a0b8d/href">https://medium.com/media/1def0bbf036f145cf4e4d8034e2a0b8d/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/598/1*7SU9WUAswshARMe4C6QHCQ.png" /></figure><p>The interesting part of this plot is, that a Manager or even the CISO is not in the top 3. But maybe it is because the dataset does not include total compensation but only the salary.</p><h3>Modeling</h3><p>As we have already seen, the two important features are experience level and company size. As we want to predict the salary, which is a regression problem, I have used linear regression as a model.</p><p>But before training the model, we have to transform the data.</p><h4>1. One-Hot Encoding</h4><p>Experience level and company size are both categorical data. This means we have to introduce dummy variables, which one-hot encode the data. In One-Hot encoding, we represent a certain value with a certain bit pattern. For example, we can represent a Junior experience level as 1000. This is what the get_dummies function from pandas does for a given feature [3].</p><p>First of all, let’s encode this data:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/515/1*85l_CrY1gy-fHvRnP3RPPg.png" /></figure><p>The following code will One-Hot encode the experience level and company size columns:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/963cb7b106e092015ec53b2e7cf7f60b/href">https://medium.com/media/963cb7b106e092015ec53b2e7cf7f60b/href</a></iframe><p>Finally, we should have the following data frame:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FjD56xvxdsJltza4brzmCw.png" /></figure><h4>2. Drop unnecessary columns</h4><p>As we don’t need remote_ratio, experience_level and company_size columns, we can drop them:</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/df215eb282b3d5b794bb364cabe5a0d9/href">https://medium.com/media/df215eb282b3d5b794bb364cabe5a0d9/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/729/1*dUGnI-A1aafFLwnzTLiuQg.png" /></figure><h4>3. Train Test Split</h4><p>In the next step, we will split data into test and training data. First of all extract the target variable (y) and the features (X). After that, we can use train_test_split from Scikit-Learn. The train test split will create a test dataset from 20% of the data.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/74a3ae0bb758d0cc6bed868c415f660b/href">https://medium.com/media/74a3ae0bb758d0cc6bed868c415f660b/href</a></iframe><h4>4. Create validation data</h4><p>In order to evaluate the models and only use the test data once, we have to create a validation dataset. That dataset will also contain 20% of the original dataset, making the test_size parameter set to 0.25, because 0.2 / 0.8 = 0.25</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/eaf894038b499dc463d4ee18523e1703/href">https://medium.com/media/eaf894038b499dc463d4ee18523e1703/href</a></iframe><h4>5. Scaling</h4><p>In the next step, the target feature (salary) will be scaled down to values ranging from -1 to +1. Machine Learning models can perform better on these values. For that purpose, I have used StandardScaler from Scikit-learn. The Scaler is fit to the training data.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/61dcabfb012545a45c68ecab09f9b978/href">https://medium.com/media/61dcabfb012545a45c68ecab09f9b978/href</a></iframe><h4>6. Training and evaluating the model</h4><p>Finally, we can use LinearRegression from Scikit-Learn. After training on the dataset, I determine the R² score and the Mean Squared error (MSE).</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/c85c7b5c811f4c1744b72f0802558dad/href">https://medium.com/media/c85c7b5c811f4c1744b72f0802558dad/href</a></iframe><p>This will output an R² score of 0.3263 and an MSE of 0.4549 .</p><h4>7. Coefficients</h4><p>We can print out the coefficients of the model with reg.coef_ . The output is:</p><pre>array([[-0.98157661,  1.24608955, -0.48704373,  0.2225308 ,  0.10616191,<br>         0.08739461, -0.19355652]])</pre><p>The first 4 coefficients are for the experience level. The last 3 are for the company size. From the coefficients we can clearly see, that experience makes the most impact on salary.</p><h4>8. Making predictions</h4><p>Now we can make predictions. For example, if we want to predict the expected salary of a Junior professional at a large size company, we have to first create the input vector, which is: [1, 0, 0, 0, 1, 0, 0] . In the next step, we will have to call the predict method of our model and finally rescale the output to the actual salary.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/c090c66d2236a75555c954c74e9cdc3b/href">https://medium.com/media/c090c66d2236a75555c954c74e9cdc3b/href</a></iframe><p>In my case, the model predicts, that a junior professional at a large company will make 79,424.79 USD.</p><h3>Conclusion</h3><p>In this article, we have discussed how Data Science can be applied to the Cyber Security domain. We first analyzed the dataset from infosec-jobs.com and were able to identify the following key takeaways for Cyber Security professionals in the U.S.:</p><ul><li>Experience level makes the most impact on salary</li><li>Junior and Intermediate professionals will have higher salaries in medium-sized companies while Senior and Executive-level will have higher salaries at large companies</li><li>There is no positive or negative impact of remote work on the salary, however, more senior professionals tend to work more often remotely</li><li>Cyber Security salaries are rising</li><li>It is possible to have a higher salary as a Security Consultant, Security Architect, and Security Researcher compared to a CISO</li></ul><p>You can get the complete code and notebooks from my <a href="https://github.com/Korkmatik/DataScience/tree/main/Projects/Cyber%20Security%20Salaries">GitHub profile</a>. If you want to download the model and scaler, you can do so from the <a href="https://github.com/Korkmatik/DataScience/releases/tag/v1">releases page</a>.</p><h4>References</h4><p>[1] infosec-jobs.com, <a href="https://infosec-jobs.com/salaries/form/">infosec-jobs.com Salaries | infosec-jobs.com</a> (2023), infosec-jobs.com</p><p>[2] infosec-jobs.com, <a href="https://infosec-jobs.com/salaries/download/">Download InfoSec/Cybersecurity salary data | infosec-jobs.com</a> (2023), infosec-jobs.com</p><p>[3] Aurélien Géron, Hands-On Machine Learning with Scikit-Learn, Keras &amp; TensorFlow (2022), O’Reilly Media, Inc.</p><h3>Resources</h3><ul><li><a href="https://github.com/Korkmatik/DataScience/tree/main/Projects/Cyber%20Security%20Salaries">DataScience/Projects/Cyber Security Salaries at main · Korkmatik/DataScience</a></li><li><a href="https://github.com/Korkmatik/DataScience/releases/tag/v1">Release Cyber Security Salaries · Korkmatik/DataScience</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b1de03afb2d2" width="1" height="1" alt=""><hr><p><a href="https://systemweakness.com/analysis-and-modeling-of-u-s-cyber-security-salaries-b1de03afb2d2">Analysis and Modeling of U.S. Cyber Security Salaries</a> was originally published in <a href="https://systemweakness.com/">System Weakness</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>