Securing LLM-Powered Applications: Overcoming Security and Privacy Challenges
summary
Shows concrete attacks and defenses: a Spring Boot app using LangChain for J and a GPT-3.5 / Llama 3.1 stack lets an attacker upload a crafted file to overwrite RAG documents and change policy text so the chatbot cancels bookings; SQL injection corrupts chat memory to inject fake assistant messages; function-calling lets the model create a new user and book a Tesla; GPT‑4 resists some injections; use Burp Suite for exploitation and apply input guardrails and limited tool permissions plus human confirmation.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:50AI as a universal solution
- 4:16Prompt injection risks
- 7:26LLM system vulnerabilities
- 8:53RAG context manipulation
- 10:18Demo: Path traversal attack
- 18:32Chat memory exploitation
- 24:18Demo: Prompt injection
- 32:28LLM permissions abuse
- 35:41Tool calling problems
- 41:50Data privacy with RAG
- 44:59Avoiding free chatbot misuse
- 46:25General security rules
related talks
44mSecuring LLM-Powered Apps: Overcoming Security and Privacy Challenges
53mBreaching LLM-Powered Applications: Overcoming Security and Privacy Challenges
50mSecuring LLM-Powered Applications: Overcoming Security and Privacy Challenges
49mBreaching LLM-Powered Applications: Overcoming Security and Privacy Challenges
53mSecuring LLM-Powered Applications: Overcoming Security and Privacy Challenges
46mBreaching LLM-Powered Apps: Overcoming Security and Privacy Challenges
45mSecuring LLM Applications: Overcoming Security & Privacy Challenges
28mSecuring LLMs: Insights into OWASP top 10