Building Secure and Responsible LLM Applications Training Course

LLM Architecture and Attack Surface Overview

• Methods for building, deploying, and accessing LLMs via APIs
• Essential components within LLM application stacks (e.g., prompts, agents, memory, APIs)
• Identification and analysis of security issues in real-world scenarios

Prompt Injection and Jailbreak Attacks

• Definition of prompt injection and its associated dangers
• Scenarios involving direct and indirect prompt injection
• Techniques used for jailbreaking to bypass safety filters
• Strategies for detection and mitigation

Data Leakage and Privacy Risks

• Unintentional data exposure through system responses
• Leaks of Personally Identifiable Information (PII) and misuse of model memory
• Designing privacy-preserving prompts and retrieval-augmented generation (RAG) approaches

LLM Output Filtering and Protection

• Utilizing Guardrails AI for content filtering and validation
• Establishing output schemas and constraints
• Monitoring and logging unsafe outputs

Human-in-the-Loop and Workflow Strategies

• Determining optimal points for introducing human oversight
• Managing approval queues, scoring thresholds, and fallback mechanisms
• Calibrating trust and the role of explainability

Secure LLM Application Design Patterns

• Implementing least privilege and sandboxing for API calls and agents
• Applying rate limiting, throttling, and abuse detection
• Ensuring robust chaining with LangChain and prompt isolation

Compliance, Logging, and Governance

• Ensuring the auditability of LLM outputs
• Maintaining traceability and version control for prompts
• Aligning with internal security policies and regulatory requirements

Summary and Next Steps