awesome-ai-security - A curated list of the latest AI-powered security t

项目说明

Awesome AI Security Tools

> A curated list of the latest AI-powered security tools — offensi, and research-focused.
> Updated regularly. Contributions welcome.

![Awesome]()

---

Contents

---

Prompt Injection & LLM Testing

| Tool | Description | Usage |
|------|-------------|-------|
| PIMap | Automated black-box prompt injection scanner — sqlmap for LLMs. 8 attack modules, 33 payloads, HTML reports | python3 pimap.py --url --param message --level 3 |
| Garak | LLM scanner — tests for prompt injection, jailbreaks, data leakage, hallucination | python3 -m garak --model openai --model_type gpt-3.5-turbo --probes all |
| PyRIT | Microsoft's Python Risk Identification Toolkit for LLMs. Red-teaming framework with attack strategies | from pyrit.orchestrator import PromptSendingOrchestrator |
| promptmap | Automatic prompt injection testing for ChatGPT plugins | python3 promptmap.py --plugin-url |
| LLMFuzzer | Fuzzing framework for LLM APIs — jailbreaks and prompt injection | python3 llmfuzzer.py --target openai --model gpt-4 |
| Vigil | LLM prompt injection and jailbreak detection — defensi | python3 vigil.py --text "your prompt here" |

---

AI-Assisted Penetration Testing

| Tool | Description | Usage |
|------|-------------|-------|
| PentestGPT | GPT-powered penetration testing assistant — guides through recon, exploitation, and reporting | python3 main.py then describe your target |
| HackingBuddyGPT | LLM-powered autonomous pri escalation agent — tested on Linux targets | python3 wintermute.py --target 192.168.1.1 --user lowpriv |
| AutoAttacker | Automated network penetration using LLMs as reasoning engine | Research tool — see repo for setup |
| Burp AI Extension | Built-in AI assistant in Burp Suite Pro — explains findings, generates payloads | A in Burp Suite Pro UI |
| NightShade | AI-augmented payload generator for web application testing | python3 nightshade.py --target |

---

Research & Triage

| Tool | Description | Usage |
|------|-------------|-------|
| COIT | C Impact Translator — maps any CVE to OT/ICS operational context, MITRE ATT&CK for ICS, Purdue levels | python3 coit.py C --backend ollama |
| | AI-powered analysis and C assistant | python3 .py --c |
| ai-exploits | Collection of real-world AI/ML exploits — MLflow, Jupyter, Ray | Reference + PoC collection |
| Nuclei AI | Nuclei templates now include AI-generated detection logic | nuclei -t ai-templates/ -u |

---

Malware Analysis

| Tool | Description | Usage |
|------|-------------|-------|
| Firmware Threat Model | Automated STRIDE threat modelling from firmware binaries — extracts filesystem, analyses strings, maps to ATT&CK | python3 firmware_threat_model.py firmware.bin --backend ollama |
| Malware Re | LLM-based static malware analysis — explains obfuscated scripts in plain English | python3 re.py --file suspicious.ps1 |
| gepetto | IDA Pro plugin — uses GPT to explain decompiled functions and rename | IDA Pro plugin — F2 to query |
| BinGPT | Binary analysis with GPT — asks natural language questions about disassembled code | python3 bingpt.py --binary target.exe --ask "what does this function do" |
| CAPA + LLM | CAPA detects capabilities in executables — pair with LLM for explanation | capa suspicious.exe --json \| python3 explain.py |

---

SAST / Code Security

| Tool | Description | Usage |
|------|-------------|-------|
| Semgrep AI | Semgrep Assistant — AI triage and fix suggestions for SAST findings | Semgrep Cloud Platform |
| CodeShield | Meta's LLM-based insecure code detector — identifies CWEs in Python, C, Ja | python3 codeshield.py --file app.py |
| Snyk Code AI | DeepCode AI — AI-powered SAST with fix suggestions | snyk code test |
| insider | SAST tool with AI-enhanced pattern detection for mobile and web | insider -tech ja -target ./src |

---

Threat Intelligence

| Tool | Description | Usage |
|------|-------------|-------|
| OT C | Maps all OT/ICS CVEs to MITRE ATT&CK for ICS — annual research dataset with HTML/PDF reports | python3 ot_c.py -- schneider --html report.html |
| OpenCTI | Open Cyber Threat Intelligence platform — AI-enriched IOC correlation and graph analysis | docker-compose up -d |
| MISP + AI | MISP threat sharing with AI enrichment modules | MISP modules — see docs |
| ThreatFox AI | IOC repository with ML-based classification | REST API — curl |

---

OT / ICS Security

| Tool | Description | Usage |
|------|-------------|-------|
| COIT | C Impact Translator — C don't tell OT engineers what matters. COIT does | python3 coit.py C |
| Firmware Threat Model | Upload any firmware binary, get a STRIDE threat model with IEC 62443 mapping | python3 firmware_threat_model.py plc_firmware.bin |
| OT C | Systematic 2024 OT C — ATT&CK for ICS mapping, rankings, research reports | python3 ot_c.py --json dataset.json --pdf report.pdf |
| GrassMarlin | NSA passive ICS/SCADA network mapping tool — AI-assisted device classification | Standalone app — see releases |

---

Network Security

| Tool | Description | Usage |
|------|-------------|-------|
| NetworkGPT | Natural language network analysis — ask questions about PCAP files | python3 networkgpt.py --pcap capture.pcap --ask "show me all DNS queries" |
| Zeek + ML | Zeek network monitor with ML-based anomaly detection plugins | zeek -r capture.pcap |
| PacketWhisperer | LLM-based PCAP analysis and threat hunting | python3 packetwhisperer.py --pcap traffic.pcap |

---

Cloud Security

| Tool | Description | Usage |
|------|-------------|-------|
| CloudGPT | AI-powered cloud security analysis — reviews IAM policies, S3 buckets, security groups | python3 cloudgpt.py --provider aws --resource iam-policy.json |
| Cloudsplaining AI | AWS IAM analysis with AI-generated risk explanations | cloudsplaining scan --input policy.json |
| Prowler AI | Cloud security assessment with AI-summarised findings | prowler aws --ai-summary |

---

Defensive / Blue Team

| Tool | Description | Usage |
|------|-------------|-------|
| Vigil | Real-time LLM prompt injection detection for production apps | pip install then wrap your LLM calls |
| LLM Guard | Input/output sanitisation library for LLM apps — detects injection, PII, toxicity | from llm_guard import scan_prompt |
| Rebuff | Prompt injection detector with canary tokens and heuristics | from rebuff import Rebuff |
| Microsoft Presidio | PII detection and anonymisation for LLM outputs | python3 -m presidio_analyzer analyze --text "text here" |
| Wazuh + AI | SIEM/XDR with AI-dri detection and alert correlation | docker-compose up -d |

---

Research & Datasets

| Resource | Description | Access |
|----------|-------------|--------|
| OWASP LLM Top 10 | The definitive LLM security risk classification | owasp.org |
| MITRE ATLAS | Ad Threat Landscape for AI Systems — ATT&CK for ML | atlas.mitre.org |
| MITRE ATT&CK for ICS | ICS/OT specific ad techniques | attack.mitre.org |
| AI Incident Database | Real-world AI failure and incident reports | incidentdatabase.ai |
| HarmBench | Standardised e framework for LLM red-teaming | Research benchmark |
| PromptBench | Microsoft's ad robustness benchmark for LLMs | Research benchmark |

---

Contributing

Pull requests welcome. Please follow this format for new entries:

Categories to consider: is it offensive or defensive? Is it LLM-specific or AI-assisted? Where does it fit?

---

From the iosec.in Toolkit

Tools built and maintained by iosec.in:

| Tool | Repo | Focus |
|------|------|-------|
| PIMap | iosec-shekhar/pimap | Prompt injection testing |
| COIT | iosec-shekhar/coit | OT C translation |
| OT C | iosec-shekhar/ot-c | OT threat landscape research |

---

*Maintained by iosec.in · MIT License*

项目信息

开发语言：未知

Star 数：2

技术要点

该项目属于防护思路方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕A curated list of the latest AI-powered security tools — offensive, defensive, and research-focused. Updated regularly. Contributions welcome.展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。