Securing Multi-Agent Networks Against Shadow AI: An Explainable and Quantum-Resilient Framework

Document Type

Conference Proceeding

Source of Publication

International Conference on Artificial Intelligence Computer Data Sciences and Applications Acdsa 2026

Publication Date

2-5-2026

Abstract

The increasing use of Shadow AI-unauthorized or unmonitored artificial intelligence components, introduces practical risks to data integrity, transparency, and accountability in multi-agent systems. In response to these concerns, this study develops a security-oriented and explainable framework that combines AI Defense, Hyperledger-based auditing, Data Processing Units (DPUs), Wi-Fi 7 communication, and postquantum cryptography (PQC) to support more reliable interaction among distributed agents. The framework was implemented and evaluated through controlled simulations in Cisco Packet Tracer, GNS3, and Python, allowing reproducible assessment of both detection and communication performance. Across these experiments, the system achieved 94.6% detection accuracy, an explanation fidelity of 0.92, and a measurable reduction in communication latency relative to a baseline configuration. These results, obtained under consistent simulation conditions, suggest that the combined use of explainable detection, decentralized auditing, and PQC-based encryption can contribute to more transparent and efficient operation of multi-agent networks within practical computational limits.

ISBN

[9798331571917]

Publisher

IEEE

Disciplines

Computer Sciences

Keywords

Blockchain Auditing, Data Processing Units, Explainable AI, Multi-Agent Systems, PostQuantum Cryptography, Shadow AI, Trust, Wi-Fi 7

Scopus ID

105037590035

Indexed in Scopus

yes

Open Access

no

Link to Full Text

Share

COinS