Exploit catalog results

Showing 6 PoCs on this page

GitHub

0xCyberstan/CVE-2026-42533-POC

Repository PoCStars: 0Created 2026-08-05
ExploitCVE-2026-425339 files

73.9 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

The artifact is a full exploit for CVE-2026-42533, a heap buffer overflow in NGINX. It includes a Python script (poc.py) that chains an information leak to defeat ASLR and a heap overflow to achieve pre-auth remote code execution (RCE) by overwriting a pool cleanup handler to call system(). The repository also contains supporting scripts for calibration, leak validation, and triggering the vulnerability across multiple code paths.

Backdoor review

No backdoor observed in reviewed code

The repository contains a proof-of-concept exploit for CVE-2026-42533, a publicly disclosed heap buffer overflow in NGINX. All provided scripts and documentation describe legitimate exploit behavior targeting the stated vulnerability: triggering the overflow, leaking heap/libc pointers, and achieving remote code execution via a forged pool cleanup chain. No evidence of concealed backdoor behavior, credential theft, unrelated remote access, persistence mechanisms, or deceptive payloads was observed. The code is consistent with a security research PoC intended for defenders and reproduction.

ClassificationExploit
Model confidence100%
AuthenticationNot required
LanguagesPythonNginx configuration
Target softwareNGINX
Attack typesHeap-based Buffer OverflowInformation LeakRemote Code Execution
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The primary artifact, poc.py, contains code that exercises the CVE-2026-42533 vulnerability to achieve remote code execution. It implements a full exploit chain including an information leak, heap grooming, a heap overflow, and a control-flow hijack to call system(). The README explicitly describes the --rce mode as 'Full pre-auth RCE' and the code confirms this by checking for the creation of a file written by the executed command.

README.md:56-57exploits/poc.py:205-296

Requirements

  • A vulnerable version of NGINX (0.9.6 through 1.30.3, 1.31.2) must be running with a configuration that uses a map directive with regex matching and a string expression referencing the map's regex capture variables before the map output variable.README.md:3-7README.md:21-24
  • The attacker must be able to send HTTP requests to the vulnerable NGINX server.README.md:100-101

Observed behavior

  • The exploit sends a crafted GET request to trigger an information leak, causing the server to return a response body containing uninitialized heap memory, which includes a libc pointer and a heap pointer.exploits/poc.py:33-52docs/EXPLOITATION.md:29-67
  • The exploit uses the leaked pointers to calculate the base addresses of libc and the heap, defeating ASLR.exploits/poc.py:197-201docs/EXPLOITATION.md:58-67
  • The exploit grooms the heap by opening multiple connections and closing one to create a hole, then sends a crafted POST request that triggers a heap buffer overflow, overwriting a neighboring connection's pool metadata.exploits/poc.py:249-269docs/EXPLOITATION.md:103-118
  • The overflow rewrites the pool's cleanup function pointer to point to a fake cleanup structure in the POST body, which contains the address of system() and a command string.exploits/poc.py:74-109docs/EXPLOITATION.md:120-161
  • Closing the corrupted connection causes NGINX to tear down the pool and execute the attacker's command via system(), achieving remote code execution.exploits/poc.py:272-284docs/EXPLOITATION.md:169-180
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Exploit Behavior
Payload withheldThe PoC scripts (poc.py, leak_multi.py, named_capture_poc.py, calibrate.py) and documentation (README.md, EXPLOITATION.md) describe and implement a full exploit chain for the stated vulnerability, including heap grooming, pointer leak, and pool cleanup hijacking to call system(). This is normal exploit behavior against the target.exploits/poc.py:1-393docs/EXPLOITATION.md:1-280README.md:1-245
Command Execution
Payload withheldThe RCE payload writes the output of 'id' to /tmp/PWNED. This is a standard proof-of-execution marker used in exploit demonstrations and does not indicate a backdoor.exploits/poc.py:206exploits/poc.py:302
Target Identification
Payload withheldThe exploit is explicitly designed for the vulnerable NGINX version and configuration described in the CVE. It does not target unrelated software or perform actions outside the scope of the vulnerability.README.md:3-24configs/nginx_poc.conf:1-45
Review boundaries

What the analysis did not establish

  • The evidence includes 2 unclassified binary files (total 15168 bytes) that were not analyzed, but their content is not required for classification as the provided text files contain the complete exploit logic.
  • The analysis is based solely on the provided text content; the code was not executed, and its reliability or safety is not verified.
  • Two files (total 15168 bytes) were classified as unclassified/binary and were not analyzed; their content is unknown but they are flagged as metadata-only.
  • The review is limited to the supplied text evidence; no dynamic analysis was performed.
Model interpretation

This review is limited to the supplied PoC code and context. It does not assert that the code works or is safe to execute.

GitHub

0xCyberstan/CVE-2026-53360-POC

Repository PoCStars: 0Created 2026-07-04
ExploitCVE-2026-533604 files

39.8 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

A guest kernel module (trigger.c) that exploits CVE-2026-53360, a heap out-of-bounds vulnerability in the Linux kernel's KVM SEV-SNP Page State Change (PSC) handler. The module crafts malicious VMGEXIT requests to cause the host to read and write beyond an allocated buffer, leaking host heap layout and corrupting adjacent slab objects.

Backdoor review

No backdoor observed in reviewed code

The artifact is a proof-of-concept (PoC) for CVE-2026-53360, a Linux kernel vulnerability in KVM's SEV-SNP PSC handling. It consists of a README, a Makefile, and a guest kernel module (trigger.c). The code performs only the described vulnerability demonstration: it crafts malformed PSC requests to trigger out-of-bounds reads and writes on the host kernel heap. No backdoor, trojan, or deceptive behavior was observed. The module checks for an SEV-SNP environment, runs four stages of probing/writing/scanning/spraying, and unloads itself. There is no persistence, credential theft, unrelated remote access, or concealed payload.

ClassificationExploit
Model confidence98%
AuthenticationNot required
LanguagesC
Target softwareLinux kernelKVMSEV-SNP
Attack typesheap out-of-bounds readheap out-of-bounds writeinformation disclosure
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The artifact contains a complete, compilable kernel module (trigger.c) that actively exploits the vulnerability by sending crafted VMGEXIT requests to the host hypervisor, causing out-of-bounds reads and writes. The README explicitly describes it as a 'Proof of concept' that 'corrupts host kernel heap memory' and 'leaks host heap layout information', and the code implements the described exploitation stages.

README.md:3-7trigger.c:1-13trigger.c:344-427

Requirements

  • The attacker must control a guest VM running on an AMD SEV-SNP enabled host.README.md:30-31
  • The host kernel must be vulnerable (versions 6.10 through 7.1-rc5, before the fix).README.md:15-16
  • The attacker must be able to load a kernel module (trigger.ko) inside the guest VM.README.md:210-215

Observed behavior

  • The module crafts a PSC request with a small allocation size (24 bytes) and a large end_entry value, causing the host to read and write past the allocated buffer into adjacent kernel heap objects.trigger.c:112-139trigger.c:67-69
  • Stage 1 probes out-of-bounds entries to map the host heap (zero vs. non-zero memory).trigger.c:169-210
  • Stage 2 demonstrates a persistent out-of-bounds write by writing a value into an adjacent slab object and confirming it is skipped on a subsequent request.trigger.c:218-276
  • Stage 3 performs a single scan with end_entry=200 to measure the reach of the out-of-bounds access.trigger.c:282-302
  • Stage 4 fires 200 VMGEXITs with out-of-bounds entries to repeatedly trigger the vulnerability and generate KASAN reports on the host.trigger.c:310-342
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Vulnerability Demonstration
Payload withheldThe code matches the described PoC behavior: it allocates a 24-byte scratch buffer, sets end_entry beyond the buffer size, and issues VMGEXITs to cause host kernel slab out-of-bounds access.trigger.c:67-70trigger.c:112-139trigger.c:169-210trigger.c:218-276trigger.c:282-302trigger.c:310-342
Environment Check
Payload withheldThe module refuses to run outside an SEV-SNP guest, limiting its effect to the intended vulnerable host configuration.trigger.c:355-364
Self Termination
Payload withheldThe module does not persist in the kernel; it runs its stages and exits, leaving no resident code.trigger.c:425-426
Review boundaries

What the analysis did not establish

  • The evidence includes a complete README and trigger.c source code, but one file (LICENSE) is omitted from the text content, though its metadata is present. This does not affect classification.
  • The analysis is based solely on the provided source code and documentation; the code was not executed, and its effectiveness is not verified.
  • One file (LICENSE) was classified as binary and not analyzed for text content; it is a standard GPL-2.0 license file per the README and is unlikely to contain executable code.
  • The review is limited to the supplied text evidence; no dynamic analysis or execution was performed.
Model interpretation

This review is limited to the supplied PoC code and context. It does not assert that the code works or is safe to execute.

GitHub

0xCyberstan/CVE-2026-42533-Config-Scanner

Repository PoCStars: 29Created 2026-07-04
ScannerCVE-2026-425332 files

26.5 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

A Python-based static analysis scanner that parses nginx configuration files to detect patterns vulnerable to CVE-2026-42533, a heap buffer overflow in the complex_value two-pass capture-clobbering bug class. It does not exploit the vulnerability; it only reads config files and reports findings.

Backdoor review

No backdoor observed in reviewed code

The repository contains a static analysis scanner for the CVE-2026-42533 nginx vulnerability. The Python script and README describe and implement a read-only config parser that flags vulnerable patterns. No backdoor, deceptive payload, or concealed harmful behavior is present in the reviewed text.

ClassificationScanner
Model confidence100%
AuthenticationNot required
LanguagesPython
Target softwareNGINX PlusNGINX Open Source
Attack typesheap buffer overflowinformation disclosure
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The artifact is a read-only static analysis tool that parses nginx configuration files to detect vulnerable patterns. It does not contain any code to send crafted HTTP requests, trigger the buffer overflow, or execute a payload. The README explicitly states it 'does not exploit anything and does not touch the running server' and the code only performs file I/O and analysis.

README.md:3-7nginx_capture_clobber_scan.py:2-9

Requirements

  • Read access to nginx configuration files on the local filesystem.nginx_capture_clobber_scan.py:110-111
  • Python 3.6+ with standard library only.README.md:9

Observed behavior

  • Parses nginx configuration files, including following 'include' directives, to identify locations where a regex capture variable and a regex map variable share a two-pass buffer in an exploitable order.nginx_capture_clobber_scan.py:307-375README.md:35-52
  • Outputs a report (text or JSON) listing vulnerable configurations, cleared (safe but fragile) configurations, and any unreadable include files. Exits with code 1 if vulnerabilities are found, 0 otherwise.nginx_capture_clobber_scan.py:443-515README.md:80-92
  • Does not send any network requests, does not interact with a running nginx server, and does not attempt to trigger the vulnerability.README.md:3-7nginx_capture_clobber_scan.py:2-9
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Url
Payload withheldExternal reference to a vulnerability write-up and exploit chain, mentioned in README and script docstring. The URL itself is not executed by the artifact.README.md:12README.md:147
Review boundaries

What the analysis did not establish

  • The evidence packet reports complete_artifact_coverage as false, indicating that not all files in the repository were included. Only two text files (README.md and nginx_capture_clobber_scan.py) are provided, and the analysis scope notes TEXT_EXCEEDS_ONE_CALL_BUDGET. Other files, if any exist in the repository, were not inspected.
  • Only the two text files (README.md and nginx_capture_clobber_scan.py) were provided and reviewed. No binary files were present. The review does not verify the correctness or safety of the external URL payload withheld.
Model interpretation

This review is limited to the supplied PoC code and context. It does not assert that the code works or is safe to execute.

GitHub

0xCyberstan/CVE-2026-31694-POC

Repository PoCStars: 16Created 2026-06-30
ExploitCVE-2026-3169411 files

33.5 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

A local privilege escalation exploit for CVE-2026-31694, a FUSE readdir cache out-of-bounds write in the Linux kernel. The exploit.c file implements a FUSE server that returns an oversized dirent, causing a 24-byte heap overflow. It uses page grooming to target the page cache of /etc/passwd, overwriting the root password field to achieve passwordless root access.

Backdoor review

No backdoor observed in reviewed code

The repository contains a proof-of-concept exploit for CVE-2026-31694, a Linux kernel FUSE readdir cache out-of-bounds write. The provided source code (exploit.c, Makefile, README.md) implements the described vulnerability exploitation and local privilege escalation to modify /etc/passwd. No concealed backdoor, unrelated remote access, persistence mechanism, or deceptive payload targeting the person running the PoC was observed. The code's behavior is consistent with the documented exploit functionality.

ClassificationExploit
Model confidence100%
AuthenticationRequired
Languagescmakefileshell
Target softwarelinux kernelfuse
Attack typeslocal privilege escalationheap overflowout-of-bounds write
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The artifact contains a complete, compilable C program (exploit.c) that actively exploits CVE-2026-31694 to achieve local privilege escalation by corrupting /etc/passwd. The README describes it as a 'marker PoC for the write primitive and a local privilege escalation' and provides instructions for running it to gain root. This is code intended to exercise a vulnerability, not merely detect it.

README.md:8-10exploit.c:1-19exploit.c:394-527

Requirements

  • Requires a vulnerable Linux kernel (6.15+ before fix) with FUSE support and a local unprivileged user account.README.md:12-13
  • Requires fusermount3 and the ability to mount FUSE filesystems.exploit.c:367-369

Observed behavior

  • Sets up a malicious FUSE server that returns a dirent with namelen=4095, causing a 24-byte overflow past a page boundary.exploit.c:193-201
  • Drains per-CPU page freelists and allocates a pool of pages to groom physical adjacency between a freed page and the /etc/passwd page cache.exploit.c:226-256
  • Triggers the overflow to overwrite the root entry in /etc/passwd with a passwordless line, then uses su to gain a root shell.exploit.c:302-353exploit.c:469-521
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Exploit Behavior
Payload withheldThis is the intended and documented behavior of the PoC, not a hidden backdoor.exploit.c:56-57exploit.c:302-353exploit.c:461-521
Payload
Payload withheldThis payload is directly related to the LPE goal of creating a passwordless root user.exploit.c:56-57
Post Exploitation
Payload withheldThis is part of the documented LPE flow to finalize the privilege escalation.exploit.c:497-506
Review boundaries

What the analysis did not establish

  • Only 3 of 11 files in the repository were provided as text; 8 files (including shell scripts in vm/ and a LICENSE file) were omitted due to budget constraints.
  • The evidence packet reports complete_artifact_coverage as false, so the full repository content is not available for analysis.
  • 8 files (likely shell scripts in the vm/ directory) were omitted from the text evidence. Their content was not reviewed, but the README describes them as VM setup and execution scripts, which is consistent with the PoC's purpose.
Model interpretation

This review is limited to the supplied PoC code and context. It does not assert that the code works or is safe to execute.

GitHub

0xCyberstan/CVE-2026-46215-POC

Repository PoCStars: 11Created 2026-06-18
ExploitCVE-2026-462153 files

26.7 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

A full local privilege escalation exploit for CVE-2026-46215, a use-after-free in the Linux kernel's DRM GEM change_handle ioctl. The exploit races GEM_CHANGE_HANDLE against GEM_CLOSE to create a dangling handle, reclaims the freed object with a pipe_buffer spray, leaks a kernel pointer to bypass KASLR, uses FLINK to set the PIPE_BUF_FLAG_CAN_MERGE flag, and overwrites /etc/passwd to create a passwordless root account.

Backdoor review

No backdoor observed in reviewed code

The repository contains a proof-of-concept exploit for CVE-2026-46215, a Linux kernel DRM GEM use-after-free vulnerability. The code performs a local privilege escalation by racing ioctls, reclaiming freed memory with pipe buffers, leaking a kernel pointer, and overwriting /etc/passwd via a DirtyPipe-style technique. All behavior is consistent with the stated exploit goal; no concealed backdoor, unrelated payload, or operator-directed harm was observed.

ClassificationExploit
Model confidence100%
AuthenticationNot required
Languagescshell
Target softwarelinux kerneldrm subsystem
Attack typesuse-after-freerace conditionlocal privilege escalationkaslr bypassdirty pipe
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The artifact contains a complete, functional exploit (poc.c) that exercises the CVE-2026-46215 vulnerability to achieve local privilege escalation by overwriting /etc/passwd. The README explicitly describes it as an exploit and provides a full exploitation chain.

README.md:1-8README.md:39-49poc.c:1-8

Requirements

  • Access to a DRM render node (/dev/dri/renderD* or /dev/dri/card*)README.md:5-7
  • A vulnerable Linux kernel (v6.18-rc1 through 6.18.31, 6.19 through 7.0.8, 7.1-rc1/rc2) built without CONFIG_KASANREADME.md:11-13README.md:68-69
  • A supported DRM driver (virtio_gpu or nouveau) that exposes object size/name at expected offsetsREADME.md:70-71

Observed behavior

  • Races DRM_IOCTL_GEM_CHANGE_HANDLE and DRM_IOCTL_GEM_CLOSE to create a dangling GEM handlepoc.c:158-186poc.c:346-422
  • Sprays pipe_buffers to reclaim the freed GEM object's slab slotpoc.c:302-332
  • Leaks a kernel pointer (pipe_buf_ops) via a driver info ioctl to bypass KASLRpoc.c:538-550
  • Uses DRM_IOCTL_GEM_FLINK to set PIPE_BUF_FLAG_CAN_MERGE on the overlapping pipe_bufferpoc.c:572-592
  • Writes to pipes to merge data into the page cache and overwrite /etc/passwd with a passwordless root entrypoc.c:595-604
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Exploit Behavior
Payload withheldThe entire codebase implements the described exploit chain against the targeted CVE. No actions beyond the stated LPE are performed.poc.c:1-655run_exploit.sh:1-146
Target File Modification
Payload withheldThe exploit's final payload writes a passwordless root entry to /etc/passwd, which is the documented goal of the PoC.poc.c:596-598poc.c:626-632
Network Activity
Payload withheldNo network sockets, connections, or data exfiltration are present in the code.poc.c:1-655
Persistence Mechanism
Payload withheldThe exploit modifies /etc/passwd for immediate privilege escalation but does not install any persistent backdoor, cron job, or service.poc.c:1-655
Review boundaries

What the analysis did not establish

  • Evidence is limited to the three text files provided (README.md, poc.c, run_exploit.sh). No binary files were inspected.
  • The exploit's reliability and safety are not verified; the analysis is based solely on the supplied source code and documentation.
  • Review is based solely on the supplied text files; no binary analysis was performed.
  • The artifact's behavior was not executed or dynamically verified.
  • The review does not assess the safety or reliability of the exploit code.
Model interpretation

This review is limited to the supplied PoC code and context. It does not assert that the code works or is safe to execute.

GitHub

0xCyberstan/CVE-2025-64459-Poc

Repository PoCStars: 2Created 2025-12-01
ExploitCVE-2025-6445913 files

8.9 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

A Dockerized Python exploit for CVE-2025-64459, a SQL injection in Django's Q object via dictionary unpacking of the _connector argument. The code constructs a malicious payload, executes a vulnerable filter, and prints the generated SQL and bypassed query results.

Backdoor review

No backdoor observed in reviewed code

The supplied evidence is a standard proof-of-concept for CVE-2025-64459. It demonstrates a SQL injection vulnerability in Django's Q object via dictionary unpacking. The code creates a local test environment, executes the documented exploit, and prints the resulting SQL and query results. No concealed executable behavior, persistence, credential theft, unrelated remote access, or other backdoor indicators are present.

ClassificationExploit
Model confidence95%
AuthenticationNot required
Languagespython
Target softwaredjango
Attack typessql injection
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The artifact contains executable code that constructs a malicious payload and triggers the vulnerable code path to produce a SQL injection, demonstrating actual exploitation rather than just detection or analysis.

CVE-2025-64459-PoC/app/webapp/management/commands/poc.py:9-10CVE-2025-64459-PoC/app/webapp/management/commands/poc.py:24-28

Requirements

  • Django application using Q(**user_input) or filter(**user_input) with unsanitized user-controlled dictionary keysCVE-2025-64459-PoC/app/webapp/management/commands/poc.py:9

Observed behavior

  • Creates test users in the databaseCVE-2025-64459-PoC/app/webapp/management/commands/poc.py:17-19
  • Constructs a malicious dictionary with _connector key set to ') OR 1=1 OR ('CVE-2025-64459-PoC/app/webapp/management/commands/poc.py:24-28
  • Passes the malicious dictionary to Q(**search_dict) and executes a filter queryCVE-2025-64459-PoC/app/webapp/management/commands/poc.py:9-10
  • Prints the generated SQL query showing injected OR 1=1 clauseCVE-2025-64459-PoC/app/webapp/management/commands/poc.py:37-40
  • Prints query results, demonstrating filter bypass and exposure of admin userCVE-2025-64459-PoC/app/webapp/management/commands/poc.py:44-50
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Vulnerability Demonstration
Payload withheldThis is the documented CVE-2025-64459 exploit behavior and is not a backdoor.CVE-2025-64459-PoC/app/webapp/management/commands/poc.py:24-28
Local Environment Setup
Payload withheldThese settings are consistent with a local PoC and do not indicate a backdoor.CVE-2025-64459-PoC/app/sqli/settings.py:3CVE-2025-64459-PoC/app/sqli/settings.py:27
Review boundaries

What the analysis did not establish

  • Two Python __init__.py files are listed as unavailable as text, but they are typically empty and unlikely to affect classification.
  • The evidence does not include the Dockerfile or requirements.txt, so the exact vulnerable Django version cannot be confirmed from the provided files.
  • Two empty __init__.py files were not included as text but are standard Python package markers and unlikely to contain behavior.
  • Seven additional files (e.g., models.py, apps.py, Dockerfile, requirements.txt) were omitted from the text evidence; their content is unknown but the included files show no backdoor behavior.
Model interpretation

This review is limited to the supplied PoC code and context. It does not assert that the code works or is safe to execute.