GitHub - bikini/exploitarium: A single archive of public exploit PoCs and vulnerability research writeups. At the time I post these, none have been reported. Feel free to report them yourself and take credit for the CVE if handed out lulz. Please do not abuse these. I do this so to allure people into the field, and I've always found this is the most efficient way.
Key Points
- 1The "Exploitarium" repository consolidates a series of proof-of-concept vulnerability research writeups, focusing on various software exploits and cybersecurity findings.
- 2The author emphasizes the use of automated fuzzing workflows driven by AI models paired with human oversight to identify and develop these security vulnerabilities.
- 3This collection is intended solely for good-faith, open-disclosure research purposes, and the author explicitly forbids the malicious use of any provided materials.
The provided repository, titled "Exploitarium," serves as a consolidated archive of public proof-of-concept (PoC) exploits and vulnerability research findings curated by the author. It functions as a central repository for various security vulnerabilities, transitioning from formerly standalone repositories into a single, organized structure. The archive spans a wide array of software targets, including critical components such as ffmpeg, libssh2, nmap, 7zip, imagemagick, and ghidra, among others.
Core Methodology and Fuzzing Workflow
The author utilizes a systematic approach to vulnerability research, heavily leveraging automated fuzzing. The core methodology is characterized by the following technical components:- AI-Driven Harness Generation: The author employs large language models—specifically identifying "GPT-5.5-3-Codex-Spark"—to automate the creation of fuzzing harnesses. The efficacy of this approach relies on the author's stated ability to provide high-quality, efficient harnesses to the AI, which reduces the complexity of the fuzzing logic required from the model.
- Human-in-the-Loop Oversight: Despite the high level of automation, the author emphasizes that the PoCs are not "vibe-coded." The author manually types the exploit code, applying academic rigor and human judgment to ensure the PoCs function correctly within the target environments. AI is used as a tool for assistance (e.g., in unfamiliar languages like Rust) rather than a replacement for expert analysis.
- Data Integrity and Consolidation: To ensure the validity of the archive, the author performed a rigorous verification process during the consolidation phase. This involved comparing the Git tree data of the former standalone repositories with the current folders. The criteria for success required:
- Matching relative file paths.
- Consistency in Git object types and tree modes (preserving executable bits).
- Identity of Git blob IDs, ensuring that for any given file, the SHA-1 hash of the file contents :
This verification covered 96 tracked entries across 12 repositories, confirming zero discrepancies between the original and consolidated versions.
Technical Scope
The repository includes a diverse collection of vulnerability types, including but not limited to:- Memory Safety Issues: Such as Use-After-Free (UAF) vulnerabilities in
c-ares. - Remote Code Execution (RCE): Found in
floci-apigatewayandghidra. - Escape and Impersonation: Observed in
docker-cpcontainer escapes andanydesk-printerimpersonation. - Logic and Security Bypasses: Such as
flowise-mcpenvironment case-bypass andnmapIPv6 extension length wrapping.
The project emphasizes open-disclosure and ethical research practices. The author asserts that the quality of findings is maintained through academic background and years of dedicated research, dismissing the notion that reliance on advanced AI models alone is sufficient for identifying meaningful vulnerabilities. The repository serves both as a historical record and a platform for ongoing security disclosures, with a commitment to providing high-quality, actionable PoCs for educational and research purposes.