#1Security papers, research projects, technical references, and reproducible material for defensive or authorized analysis.
Kitploit recommended

A lifecycle benchmark for black-box LLM extraction attacks, defenses, and adaptive attacks.

Research repository for CVE-2026-81000 (TUNderflow), a Linux kernel TUN/TAP receive headroom integer underflow enabling local privilege escalation,…

research on finding the bug and fix of CVE-2026-84616 and CVE-2026-84607

LLM-powered agent that autonomously fixes GitHub issues, finds cybersecurity vulnerabilities, and solves CTF challenges using configurable tool-use…

Cybersecurity oriented awesome list

ExploitGym is a large-scale, realistic benchmark built from real-world vulnerabilities designed to evaluate AI agents' ability to develop exploits.

Fine-tuning framework that applies first-order optimal safety calibration and periodic recalibration to LLMs, preserving safety-compatible updates…

Patch-diff research for published urllib3 issues: pinned releases, original offline harnesses, before/after behavior and benign controls.

Spicy malware 0day. Full kill-chain malware: exploit, pivot, c2, persistence. Rust converted to pseudo-code - if you're smart you can build it…

Embeds and detects keyed, style-based watermarks in LLM-generated Python, Java, and C++ code via CST rewriting, preserving functional correctness and…

Exploit chain research targeting CVE-2026-43499 on Samsung Galaxy S21

Kernel-level eBPF sandbox for securing LLM agent tool calls made through the Model Context Protocol (MCP)

An automated Themida / Code Virtualizer / WinLicense (v3.0-v4) unpacker and dump fixer tool

Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models (NeurIPS'26)

Research implementation of Hop-Decayed Influence (HDI) and the 3S attack framework, exposing structural auxiliary indexing vulnerabilities in…

LLM safety: analyze how fine-tuning shifts refusal behavior and evaluate layer-freezing and low-rank defenses.

Linux process identity cloaking tool that spoofs comm, argv, cmdline, environ, exe path, and VMAs via an 11-phase prctl pipeline to impersonate…

Research implementation for mitigating adaptive prompt injections via on-policy distillation, with training recipes and evaluators for SEP, PISmith,…