DeepYardDeepYard
C

Cognitive Firewall

Hybrid edge-cloud security defense against prompt injection attacks on browser AI agents

Open SourceFree

About

Research-backed security architecture that protects browser-based AI agents from Indirect Prompt Injection attacks through a three-stage split-compute defense system. Combines a local visual Sentinel for real-time threat detection with a cloud-based Deep Planner for advanced analysis, optimizing the tradeoff between latency, privacy, and security. Designed for production deployments of autonomous agents that interact with untrusted web content.

Details

Type
Integrations
Language

Tags

autonomousobservabilityopen-sourcepythonsecurityresearch