Securing LLM-Powered Applications: Overcoming Security and Privacy Challenges
summary
Shows concrete exploits and fixes: a live demo where a malicious chat entry becomes part of RAG or chat memory and causes an LLM to cancel a booking, a Burp Suite interception that would allow path traversal file overwrite, and a tool call that executes SQL like "DROP TABLE products" after an LLM issues a function call. Recommends prompt injection defenses, limiting tool calling permissions to user scope, logging and human review, and offers model-specific notes (GPT‑3.5 vs GPT‑4, Llama 3.1) and on‑premises vs cloud tradeoffs.
AI-written summary · about
chapters (14)
- 0:00LLM security challenges
- 3:11LLMs use and risks
- 6:43Core components of LLM apps
- 8:09RAG vulnerabilities
- 13:48Path traversal and file uploads
- 19:57Chat memory and SQL injection
- 26:40Prompt injection demo
- 27:23Web shop chatbot scenario
- 31:17LLM permissions and tool calling
- 35:13Local LLM unintended actions
- 40:20MCP service risks
- 43:13On-prem LLM solutions
- 45:38Securing LLM applications
- 49:08Choosing an LLM
related talks
51mSecuring LLM-Powered Applications: Overcoming Security and Privacy Challenges
49mBreaching LLM-Powered Applications: Overcoming Security and Privacy Challenges
53mBreaching LLM-Powered Applications: Overcoming Security and Privacy Challenges
44mSecuring LLM-Powered Apps: Overcoming Security and Privacy Challenges
45mSecuring LLM Applications: Overcoming Security & Privacy Challenges
53mSecuring LLM-Powered Applications: Overcoming Security and Privacy Challenges
46mBreaching LLM-Powered Apps: Overcoming Security and Privacy Challenges
44mI Abuse Your LLM Live with 10 Security Vulnerabilities