LLM Proxy — Building a Validation Layer for LLMs
summary
A presentation about building a validation/proxy layer for large language models (LLMs), covering practical cyber‑security techniques to validate and filter LLM inputs and outputs. It focuses on implementing a proxy that enforces security checks and policies for LLM interactions to reduce risks when integrating LLMs into applications.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 3:13Building first LLM Proxy
- 9:13Business needs and challenges
- 13:52LLM risks
- 18:49How LM Proxy works
- 22:39Guardrail types
- 24:02Communication contexts
- 30:00On-prem and cloud models
- 36:02Intermediary systems and orchestration
- 38:50Architectural challenges
- 41:57Components and scaling
- 44:45Summary
related talks
47mA Step-by-Step Guide to Securing Large Language Models (LLMs)
20mSecurity Threats in Modern LLM Applications
29mBeyond the Hype: Building Trustworthy and Reliable LLM Applications with Guardrails
49mA Practical Approach to Using LLMs: Peculiarities and Pitfalls
44mI Abuse Your LLM Live with 10 Security Vulnerabilities
50mBeyond the Prompt: Evaluating, Testing, and Securing LLM Applications
23mSecurity and auditing tools in Large Language Models (LLM)
39mBeyond the Prompt: Evaluating, Testing, and Securing LLM Applications