Guarding the LLM Galaxy: Security, Privacy, and Guardrails in the...
summary
A rapid survey of real failures and concrete defenses: a Samsung engineer pastes proprietary code into GPT and an API key theft leads to LLM jacking that racks up $46,000 in a day. Technical attacks include prompt injection, training‑data poisoning, model extraction and context‑window leaks. Practical mitigations shown are input/output filters, DLP and NER masking, rate limits and short‑lived JWTs, mutual TLS, RBAC, vector DB deletion plus suppression lists, and guardrail tools like NVIDIA’s NeMo Guardrails, LangChain utilities, rebuff, gizgard and GAR.
AI-written summary · about
chapters (5)
- 0:00LLM security risks
- 10:45Threat modeling vulnerabilities
- 26:16Guardrail architectures
- 39:51Secure LLM practices
- 44:48Key takeaways
related talks
50mSecuring LLM-Powered Applications: Overcoming Security and Privacy Challenges
44mSecuring LLM-Powered Apps: Overcoming Security and Privacy Challenges
49mBreaching LLM-Powered Applications: Overcoming Security and Privacy Challenges
51mSecuring LLM-Powered Applications: Overcoming Security and Privacy Challenges
20mSecurity Threats in Modern LLM Applications
53mBreaching LLM-Powered Applications: Overcoming Security and Privacy Challenges
29mBeyond the Hype: Building Trustworthy and Reliable LLM Applications with Guardrails
37mSecure by Design: Strategies for LLM Adoption in Cloud-Native Environm...