panda-re/lava

LAVA: Large-scale Automated Vulnerability Addition

What it solves

LAVA addresses the shortage of ground-truth corpora for evaluating and improving bug-finding tools. It allows developers to create large-scale, synthetic but realistic vulnerability corpora on demand by automatically injecting bugs into software.

How it works

LAVA automatically injects vulnerabilities into target software programs. Each injected bug is paired with a specific input that triggers it, ensuring that the bug is reproducible and known. This creates a set of "ground truth" data where the bug and its triggering input are known, which can then be used to test if a bug-finding tool can actually locate the same vulnerability.

Who it’s for

This tool is designed for developers of bug-finding tools and security researchers who need high-quality, labeled datasets of vulnerable software to rigorously evaluate their analysis tools.

Highlights

  • Automated Bug Injection: Automatically creates vulnerabilities in existing software.
  • Ground Truth Generation: Every bug comes with a corresponding triggering input.
  • Realistic Synthetic Bugs: Bugs are embedded deep within programs to simulate real-world scenarios.
  • Extensible: Supports adding new target projects via configuration files in the target_configs directory.

Related

  • Project
    specula-org/Specula

    Specula uses AI coding agents to automatically generate TLA+ formal specifications from source code to find and reproduce deep bugs in concurrent and distributed systems.

  • Project
    cyberkaida/reverse-engineering-assistant

    ReVa is a Ghidra MCP server that enables AI models to interact with reverse engineering tools, allowing LLMs to perform deep binary analysis and automation while minimizing hallucinations.

  • Project
    Vasco0x4/AIDA

    AIDA is an open‑source autonomous pentesting agent that lets any LLM (Claude, Gemini, OpenAI Codex, etc.) drive a Docker‑based security testing environment. The model plans the assessment, calls built‑in tools (nmap, sqlmap, custom Python, raw HTTP), logs every step in a persistent notebook, auto‑scores findings with CVSS 4.0, and can export PDF reports. It ships with a lightweight `aida-pentest` container (or can attach Exegol for more tools) and includes a web dashboard, authentication, and notification integrations. The README cites real CVEs discovered by the system. Licensed under AGPL‑v3.

  • Project
    visa/visa-vulnerability-agentic-harness

    An autonomous pipeline for vulnerability discovery, remediation, and validation in source code using LLMs to reduce the time from detection to validated fix.

  • Project
    starVLA/starVLA

    An open-source research platform for developing Vision-Language-Action (VLA) models, providing a modular codebase for integrating different backbones, action heads, and training recipes for generalist robots.