huhusmang/awesome-llms-for-vulnerability-detection
O índice mais abrangente e continuamente atualizado da comunidade sobre pesquisas em Large Language Models para detecção de vulnerabilidades de software — artigos sobre detecção em nível de função, nível de repositório, agentes e contratos inteligentes, além de conjuntos de dados, benchmarks e pesquisas.
Uma lista curada de artigos, projetos e habilidades de agentes sobre o uso de LLMs para detecção e descoberta de vulnerabilidades.
Mostrando apenas 2025 e posteriores. Para trabalhos anteriores, consulte Arquivo de Artigos (2024 e anteriores).
| Título | Local | Ano | Artigo | Github |
|---|
| VulnGym: Benchmarking Coding Agents for Repository-Level Vulnerability Detection | 2026 | link | link | |
| VulTriage: Triple-Path Context Augmentation for LLM-Based Vulnerability Detection | 2026 | link | link | |
| Synthesizing Multi-Agent Harnesses for Vulnerability Discovery | 2026 | link | link | |
| QRS: A Rule-Synthesizing Neuro-Symbolic Triad for Autonomous Vulnerability Discovery | 2026 | link | ||
| Seclens: Role-specific Evaluation of LLM's for security vulnerablity detection | 2026 | link | link | |
| Do Fine-Tuned LLMs Understand Vulnerabilities? An Investigation into the Semantic Trap | 2026 | link | ||
| Sifting the Noise: A Comparative Study of LLM Agents in Vulnerability False Positive Filtering | ISSTA | 2026 | link | |
| AgenticSCR: An Autonomous Agentic Secure Code Review for Immature Vulnerabilities Detection | 2026 | link | ||
| LLM-based Vulnerability Detection at Project Scale: An Empirical Study | 2026 | link | ||
| MulVul: Retrieval-augmented Multi-Agent Code Vulnerability Detection via Cross-Model Prompt Evolution | 2026 | link | ||
| VulnLLM-R: Specialized Reasoning LLM with Agent Scaffold for Vulnerability Detection | 2025 | link | ||
| VULPO: Context-Aware Vulnerability Detection via On-Policy LLM Optimization | 2025 | link | ||
| VulInstruct: Teaching LLMs Root-Cause Reasoning for Vulnerability Detection via Security Specifications | 2025 | link | ||
| From Large to Mammoth: A Comparative Evaluation of Large Language Models in Vulnerability Detection | NDSS | 2025 | link | |
| Benchmarking LLMs and LLM-based Agents in Practical Vulnerability Detection for Code Repositories | ACL | 2025 | link | link |
| A Systematic Literature Review on Detecting Software Vulnerabilities with Large Language Models | 2025 | link | link | |
| LLMxCPG: Context-Aware Vulnerability Detection Through Code Property Graph-Guided Large Language Models | Usenix | 2025 | link | link |
| CLeVeR: Multi-modal Contrastive Learning for Vulnerability Code Representation | ACL Findings | 2025 | link | link |
| Mono: Is Your "Clean" Vulnerability Dataset Really Solvable? Exposing and Trapping Undecidable Patches and Beyond | 2025 | link | link | |
| Learning to Focus: Context Extraction for Efficient Code Vulnerability Detection with Language Models | 2025 | link | ||
| SV-TrustEval-C: Evaluating Structure and Semantic Reasoning in Large Language Models for Source Code Vulnerability Analysis | SP | 2025 | link | link |
| SecVulEval: Benchmarking LLMs for Real-World C/C++ Vulnerability Detection | 2025 | link | link | |
| CVE-Bench: Benchmarking LLM-based Software Engineering Agent's Ability to Repair Real-World CVE Vulnerabilities | NAACL | 2025 | link | link |
| R2Vul: Learning to Reason about Software Vulnerabilities with Reinforcement Learning and Structured Reasoning Distillation | 2025 | link | link | |
| Neuro-symbolic Static Analysis with LLM-generated Vulnerability Patterns | 2025 | link | ||
| Context-Enhanced Vulnerability Detection Based on Large Language Model | 2025 | link | ||
| Everything You Wanted to Know About LLM-based Vulnerability Detection But Were Afraid to Ask | 2025 | link | link | |
| MOS: Towards Effective Smart Contract Vulnerability Detection through Mixture-of-Experts Tuning of Large Language Models | 2025 | link | ||
| Abundant Modalities Offer More Nutrients: Multi-Modal-Based Function-Level Vulnerability Detection | TOSEM | 2025 | link | link |
| Generative Large Language Model usage in Smart Contract Vulnerability Detection | 2025 | link | ||
| Closing the Gap: A User Study on the Real-world Usefulness of AI-powered Vulnerability Detection & Repair in the IDE | ICSE | 2025 | link | link |
| Vulnerability Detection with Code Language Models: How Far Are We? | ICSE | 2025 | link | link |
| Combining Fine-Tuning and LLM-based Agents for Intuitive Smart Contract Auditing with Justifications | ICSE | 2025 | link | |
| LAMD: Context-driven Android Malware Detection and Classification with LLMs | 2025 | link | ||
| LLMs in Software Security: A Survey of Vulnerability Detection Techniques and Insights | 2025 | link | link | |
| One-for-All Does Not Work! Enhancing Vulnerability Detection by Mixture-of-Experts (MoE) | 2025 | link |
| Nome | Descrição | Github |
|---|---|---|
| OpenAnt (Knostic) | Descoberta de vulnerabilidades em múltiplos estágios com verificação adversarial baseada em LLM | link |
| DeepAudit | Plataforma multiagente de red-team com validação de exploits em sandbox Docker | link |
| AutoCVE | Detecção e relatório automatizados de vulnerabilidades com arquitetura multiagente | link |
| Darkmoon | Plataforma autônoma de pentest com IA de código aberto (GPL-3.0) e host MCP; subagentes ofensivos por tecnologia, cobertura de Active Directory e Kubernetes, mais de 80 ferramentas orquestradas, trilha de evidências por descoberta | link |
| strix | Ferramenta autônoma de pentest com IA de código aberto instalável via pip | link |
| deepsec (Vercel) | Harness de segurança para varredura profunda de vulnerabilidades em bases de código com agentes de codificação | link |
| Nome | Descrição | Link |
|---|---|---|
| codex-security (OpenAI) | Varredura autônoma de vulnerabilidades em nível de repositório via agentes Codex | link |
| defending-code-reference-harness (Anthropic) | Habilidades de referência para modelagem de ameaças, varredura, triagem e correção com Claude Code | link |
| security-audit-skill (Cloudflare) | Habilidade de auditoria de segurança em seis fases com agentes de caça paralelos e validação adversarial | link |
Captura e atualização diária automatizada de artigos do Arxiv para palavras-chave especificadas por meio de workflows.
