ai-security-toolkit 开源项目推荐（19 个）

microsoft /PyRIT

微软开源 AI 红队测试工具包。通过自动扫描、自适应多轮攻击和 Web UI 测试生成式 AI 系统。

cisco-ai-defense /model-provenance-kit

Toolkit to assess and determine model provenance

delphisecurity /xaidr

Runtime security for AI agents. In-process, zero dependencies, Apache 2.0.

shokribardiya /the-recursive-threshold

Oblivion / the-recursive-threshold. When AI starts helping build the AI that comes next: recursive self-improvement, autonomous agents, cyber risk, nuclear stability, and the control gap. Bardiya Shokri

ciberconscientes /nido

Reglas de seguridad para proyectos vibe-codeados, en castellano. Basado en OWASP.

spoonix-dev /nexusguard-ai

Ultra-fast AI reverse proxy with semantic cache, PII masking, budget controls, and smart failover — featuring a stunning TUI dashboard

Aveerayy /agent-guard

The open-source firewall for AI agents. Control what your agents do — policy enforcement, MCP security scanning, zero-trust identity, sandboxing, and reliability. 10/10 OWASP. One pip install.

kriskimmerle /ragaudit

RAG Knowledge Base Poisoning Detector - Zero-dependency static analyzer that scans documents BEFORE embedding. Detects prompt injection, hidden instructions, and content manipulation. Addresses OWASP Agentic Top 10 ASI06.

dibakshya01 /grey-panda

The calm guardian for AI, agent, and MCP code — a zero-dependency, OWASP-anchored AI security toolkit.

kriskimmerle /contextguard

Agent Context & Memory Poisoning Detector - Zero-dependency analyzer for agent memory stores and conversation logs. Detects OWASP Agentic Top 10 ASI06 poisoning patterns.

abooker30126 /Awesome-AI-Security

📁 #AISecurity

nand9lohot /llm-adversarial-attacks-textattack

A reproducible adversarial ML lab that demonstrates TextFooler, BERTAttack, and DeepWordBug attacks against transformer-based sentiment models, with Docker automation and adversarial security reporting.

Reider85 /OryaSecurity

LLM Security Scanner — a protective inline proxy between your AI app and the LLM provider Inspects incoming prompts for PII / secrets / prompt-injection, outgoing generations for leaks, returns an allow / block decision and a full report for SecOps.

kriskimmerle /llmoutput

LLM Output Sanitization Checker - static analyzer for unsafe LLM output handling

kriskimmerle /slopsquatcheck

AI Hallucination Squatting Detector for Python Dependencies - detects packages LLMs commonly 'invent'

kriskimmerle /vibeguard

Vibe-Coded Python Security Linter - catches AI-typical security mistakes

ElNaboulsi /ai-security-risk-addon-qatar-banks

Free AI security risk add-on for banks in Qatar: plugs into your existing risk register and adds the QCB and NCSA requirements for AI.

kriskimmerle /agentflow

Agent Orchestration Static Security Analyzer - Zero-dependency static analyzer for LangChain, CrewAI, AutoGen, and LangGraph agent code. Detects OWASP Agentic Top 10 security antipatterns.

snrgx /AI-Library

Infusing AI into process engineering: visual, bilingual (EN/AR) skills, plugins and MCP tools for founders and creators, with a friendly UI/UX so anyone can use them.