Client/Malicious Actor
The client or malicious actor who interacts with the LLM system, potentially attempting to exploit vulnerabilities.
View vulnerabilities →
LLM System Components
Explore the different components of an LLM system and their associated security vulnerabilities. Each component represents a potential attack surface that must be secured in production deployments.
System Components List
Ingress
The entry point for user inputs to the LLM system, handling queries before processing.
View vulnerabilities →
LLM Service
The core language model service that processes inputs and generates responses.
View vulnerabilities →
Vector DB
Database storing vector embeddings used by the LLM for retrieval-augmented generation.
View vulnerabilities →
Training Pipeline
The pipeline responsible for training and fine-tuning the language model.
View vulnerabilities →
Security Layer
The security mechanisms that protect the LLM system from various threats.
View vulnerabilities →