Honey Shield: A Deceptive Security Model to Study and Prevent Cyber Attacks

Authors: Abinaya M, Harshada R, Karthika U, Priyavarshini D

Abstract: Traditional network security systems that rely on static rules and signature-based detection are increasingly inef- fective against dynamic, automated, and zero-day attacks. This paper presents Honey Shield, an AI-driven deceptive security model implemented as a microservice-based Intelligent Network Gateway that intercepts connections, uses a cloud-backed dy- namic blocklist, and leverages the Google Gemini API for real- time payload analysis. Honey Shield’s Gateway captures initial payloads and source metadata, the HoneypotService orchestrates a two-stage analysis (DynamoDB blocklist lookup followed by Gemini AI analysis), and a closed-loop feedback mechanism updates the blocklist to enable self-learning defenses. We present system architecture, module descriptions, dataset and testing approaches, and validation results from a proof-of-concept de- ployment. The approach demonstrates an adaptive, low-latency mechanism for identifying and mitigating application-layer at- tacks while minimizing manual intervention.