FURI | Spring 2026

Access Controlled Website Interaction for Agentic AI with Delegated Critical Tasks

Security icon, disabled. A blue padlock, locked.

Autonomous artificial intelligence (AI) agents increasingly perform critical online tasks on behalf of users, yet current websites lack mechanisms for securely delegating such authority. This research designs and implements a website interaction model with fine-grained access control, extending the open-source Secure Swarm Toolkit (SST) to support delegated access through cryptographic session tokens. The evaluation demonstrates reliable authentication, strict enforcement of user-defined access scopes, and the prevention of unauthorized access with a moderate average latency of 135,050.467 ms. Future work will explore formal verification and broader deployment to strengthen security guarantees for large-scale agentic AI ecosystems.

Student researcher

Sunyoung Kim

Computer science

Hometown: Seoul, Seoul, South Korea

Graduation date: Spring 2026