Exploit catalog results

Showing 3 PoCs on this page

GitHub

tristanqtn/CVE-2024-37054

Repository PoCStars: 1Created 2026-05-19
ExploitCVE-2024-370542 files

18.5 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

The artifact is a Python exploit for CVE-2024-37054, a deserialization vulnerability in MLflow. It automates discovery of MLflow models, uploads a malicious cloudpickle payload to overwrite the model's python_model.pkl artifact, and triggers remote code execution via a reverse shell.

Backdoor review

No backdoor observed in reviewed code

The artifact is a proof-of-concept exploit for CVE-2024-37054. It demonstrates a legitimate deserialization attack against MLflow by uploading a malicious pickle payload. All behavior is consistent with the stated exploit purpose: it validates MLflow access, optionally registers/logs into a target application, trains a model, resolves run/experiment IDs, builds a reverse-shell pickle, uploads it to the MLflow artifact store, and triggers model loading. No concealed, deceptive, or unrelated harmful behavior was observed.

ClassificationExploit
Model confidence98%
AuthenticationRequired
LanguagesPython
Target softwareMLflow
Attack typesDeserialization of Untrusted DataRemote Code Execution
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The artifact contains a complete, automated Python script (exploit.py) that actively exploits CVE-2024-37054 by uploading a malicious pickle payload and triggering its deserialization to achieve remote code execution. The README describes it as an exploit and provides usage examples for obtaining a reverse shell. The code includes payload construction (build_revshell_pickle), artifact poisoning (upload_pickle), and trigger mechanisms (trigger_load), which are hallmarks of an exploit, not a scanner or writeup.

exploit.py:1-4exploit.py:202-207exploit.py:211-228exploit.py:232-243README.md:1README.md:19-29

Requirements

  • Network access to an MLflow tracking server (versions 0.9.0 through 2.14.1) with valid credentials (HTTP Basic Auth).README.md:4exploit.py:39-40
  • A registered MLflow model or the ability to trigger model training via the target application.exploit.py:148-164exploit.py:324-328
  • An application endpoint that calls mlflow.pyfunc.load_model() on the poisoned model, or an alternative trigger mechanism.README.md:40exploit.py:232-243

Observed behavior

  • Validates connectivity and credentials to the MLflow server by querying the registered-models/search API endpoint.exploit.py:48-75
  • Enumerates registered MLflow models and their versions when run in 'list' mode.exploit.py:303-313
  • Authenticates to a target web application (via login or throwaway account registration) to obtain a session cookie.exploit.py:79-144
  • Triggers model training on the target application by uploading a CSV file, causing a new model to be registered in MLflow.exploit.py:148-164
  • Resolves the experiment_id and run_id for a target model by querying MLflow REST APIs.exploit.py:168-198
  • Constructs a malicious cloudpickle payload that executes a bash reverse shell command via os.system upon deserialization.exploit.py:202-207
  • Uploads the malicious pickle to the MLflow artifact store, overwriting the model's python_model.pkl file.exploit.py:211-228
  • Triggers the deserialization by sending a request to the application endpoint that loads the poisoned model, resulting in remote code execution.exploit.py:232-243exploit.py:362-366
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Reverse Shell Payload
Payload withheldThe exploit builds a reverse shell command using attacker-supplied lhost and lport, which is the documented payload for the CVE.exploit.py:202-207
Target Application Interaction
Payload withheldThe script attempts to log into a target web application and falls back to registering a throwaway account to obtain a session cookie, which is then used to trigger model loading. This is part of the documented exploit workflow.exploit.py:79-144
Review boundaries

What the analysis did not establish

  • The evidence includes only two text files (README.md and exploit.py) from the repository snapshot. No other files (e.g., configuration, dependencies, or additional scripts) were provided, but the selected files are complete and sufficient for classification.
  • The exploit targets a specific lab environment (smarthire.htb) by default, which may limit its general applicability without modification.
  • Review is limited to the two text files provided; no binary or external dependencies were inspected.
  • The artifact's behavior depends on network targets and user-supplied parameters; only the static code was analyzed.
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

tristanqtn/CVE-2025-54123

Repository PoCStars: 0Created 2026-03-29
ExploitCVE-2025-541232 files

12.5 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

Python script that authenticates to a Hoverfly instance and exploits CVE-2025-54123 to achieve authenticated remote code execution via command injection in the middleware API endpoint.

Backdoor review

No backdoor observed in reviewed code

The supplied evidence consists of a README.md describing the CVE-2025-54123 vulnerability and an exploit.py script that implements a proof-of-concept for the described authenticated remote code execution. The exploit.py script performs only the expected exploit behavior: authenticating to a Hoverfly instance, sending a crafted payload to the vulnerable middleware endpoint, and executing user-supplied commands or establishing a reverse shell. No concealed executable behavior, persistence mechanisms, credential exfiltration to unrelated parties, or other backdoor functionality was observed.

ClassificationExploit
Model confidence100%
AuthenticationRequired
Languagespython
Target softwareHoverfly
Attack typescommand injectionremote code execution
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The artifact is a fully functional Python script that actively exploits CVE-2025-54123 by sending crafted requests to execute arbitrary commands on a vulnerable Hoverfly instance. It includes authentication, payload delivery, command execution, and output retrieval, which constitutes an exploit.

exploit.py:1-308README.md:4

Requirements

  • Valid Hoverfly credentials (username/password) or a session tokenexploit.py:38-50exploit.py:250-262
  • Target Hoverfly instance running a vulnerable version (<= 1.11.3)README.md:4

Observed behavior

  • Authenticates to the target Hoverfly instance using provided credentials or session tokenexploit.py:38-50exploit.py:53-64
  • Sends a PUT request to /api/v2/hoverfly/middleware with a JSON payload containing a user-controlled binary and scriptexploit.py:67-77exploit.py:80-94
  • Executes arbitrary shell commands on the target and returns output extracted from the error responseexploit.py:107-121
  • Provides an interactive shell and reverse shell capabilityexploit.py:129-144exploit.py:147-164
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Exploit Behavior
Payload withheldThis is the normal exploit behavior for CVE-2025-54123 as described in the README and CVE record.exploit.py:67-77exploit.py:80-94
Reverse Shell Functionality
Payload withheldThis is a standard feature of RCE exploits and is explicitly documented in the script's help text and the CVE description.exploit.py:147-164exploit.py:34
Review boundaries

What the analysis did not establish

  • Only the two text files (README.md and exploit.py) were provided; no other files from the repository were included in the evidence packet.
  • The analysis is limited to static review of the source code; dynamic behavior was not observed.
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

tristanqtn/CVE-2025-66034

Repository PoCStars: 1Created 2026-03-20
ExploitCVE-2025-660342 files

17.7 KiB

Analysisdeepseek-v4-pro:cloud ·

Technical assessment

A Python exploit for CVE-2025-66034 that plants a PHP webshell via an arbitrary file write in fontTools varLib, then provides command execution, interactive shell, and reverse shell capabilities.

Backdoor review

No backdoor observed in reviewed code

The repository contains a documented proof-of-concept exploit for CVE-2025-66034. The exploit.py script and README.md describe a legitimate attack chain against a vulnerable fontTools endpoint: crafting a malicious .designspace file to achieve arbitrary file write and embedding a PHP webshell payload. The script's behavior is fully consistent with the disclosed vulnerability and contains no concealed backdoor, unrelated payload, or deceptive behavior targeting the person running the PoC.

ClassificationExploit
Model confidence100%
AuthenticationNot required
LanguagesPython
Target softwarefonttools
Attack typesarbitrary file writewebshell uploadcommand injection
Evidence & reasoningClassification basis · observed behavior · safety review
Technical evidence

Classification basis and observed behavior

Classification basis

The artifact contains a complete, functional Python script (exploit.py) that actively exploits CVE-2025-66034 by uploading a malicious .designspace file to write a PHP webshell to an arbitrary path, then executing commands on the target. This is exploit code, not a scanner or writeup.

exploit.py:1-293

Requirements

  • A target web application that processes user-supplied .designspace files using a vulnerable version of fontTools varLib.README.md:18
  • The attacker must know the URL of the font processor endpoint, the URL where the webshell will be reachable, and the filesystem path to write the shell.README.md:77-79

Observed behavior

  • Crafts a malicious .designspace file containing a PHP payload in a CDATA-wrapped labelname element and an arbitrary output path in the variable-font filename attribute.exploit.py:52-91
  • POSTs the malicious .designspace and a minimal valid TTF master to the target font processor endpoint to trigger the file write.exploit.py:95-114
  • Sends commands to the planted PHP webshell via HTTP GET requests and extracts command output from the binary font response.exploit.py:117-139
  • Provides an interactive pseudo-shell loop and a reverse shell trigger using base64-encoded bash payloads.exploit.py:142-168
Safety-review evidence

Behaviors behind the backdoor verdict

Observables

Exploit Behavior
Payload withheldThe exploit.py script constructs a .designspace file with a PHP payload in a <labelname> CDATA block and an arbitrary output path in the <variable-font> filename attribute. It then POSTs this designspace and a minimal TTF to a user-supplied process URL, and provides functions to interact with the resulting webshell.exploit.py:52-91exploit.py:95-114exploit.py:117-139
Documentation
Payload withheldThe README explains the two weaknesses (unsanitized output filename and XML content passed through to output), the attack steps, and how to use the exploit script. This matches the exploit.py implementation.README.md:1-131
Review boundaries

What the analysis did not establish

  • Only the two text files (README.md and exploit.py) were reviewed; no binary files were present or analyzed.
  • The review does not assess the safety of executing the script against a real target, only whether the script itself contains a backdoor against its user.
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.