Securing The Gateway And Mitigating Risks In LLM API Integration
summary
Includes concrete breach examples and working exploits: prompt injection that exposed AWS secret key in an orom mini 3B deployment and a prompt that achieves remote code execution via terminal prompt through a LangChain llm_math path. Documents prompt‑injection incidents on Bing and a plugin that toggled GitHub repos public. Describes a scanner with ~58 templates integrated into CI that flags leaks, uses a cheap local LLM plus NLP filters for intent and profanity, enforces sandboxed function calls instead of direct DB access, and logs and rate‑limits calls.
AI-written summary · about
chapters (13)
- 0:00LLM security challenges
- 2:19LLM integrations need attention
- 3:00Recent LLM breaches
- 7:21LLM vulnerabilities demo
- 11:01Orom 3B data leakage
- 12:34RCE in LLM integrations
- 14:01LLM security testing
- 15:25Why security fails for LLMs
- 17:33Securing LLM APIs playbook
- 22:26Q&A input sanitization
- 24:31Q&A LLM security tools
- 25:35Q&A RCE in LangChain
- 26:18Q&A instruction models
related talks
53mBreaching LLM-Powered Applications: Overcoming Security and Privacy Challenges
49mBreaching LLM-Powered Applications: Overcoming Security and Privacy Challenges
50mSecuring LLM-Powered Applications: Overcoming Security and Privacy Challenges
44mI Abuse Your LLM Live with 10 Security Vulnerabilities
36mCompromising LLMs: The Advent of AI Malware
44mSecuring LLM-Powered Apps: Overcoming Security and Privacy Challenges
20mUnderstanding and Mitigating Risks in LLM-Powered Web Apps
28mSecuring LLMs: Insights into OWASP top 10