SECNET ARCHITECTURE FOR AI-DRIVEN ACCESS CONTROL AND REAL-TIME THREAT DETECTION IN BLOCKCHAIN SYSTEMS
TL;DR
SecNet is introduced, a comprehensive and layered hybrid architecture that integrates smart-contract-based authorization mechanisms, AI-powered anomaly and threat detection, encrypted off-chain storage, federated learning, federated learning, and a token-based incentive layer.
Abstract
Storage systems must be verifiably secure, and access control must be fine grained and transparent, and must provide audit trails for modern cyber-physical and cloud ecosystems. While blockchain ensures immutability and traceability of data, it does not have the ability to make real-time authorization decisions, identify abnormal behavior, or facilitate privacy-preserving collaborative data analytics across organizations. This paper introduces SecNet, a comprehensive and layered hybrid architecture that integrates smart-contract-based authorization mechanisms, AI-powered anomaly and threat detection, encrypted off-chain storage, federated learning, and a token-based incentive layer. SecNet is based on a systematic review of twenty-one recent works from the areas of blockchain access control, federated learning, secure multi-party computation, zero-knowledge proofs, reputation systems, and healthcare data governance, the results of which form four evaluation dimensions: Access Authorization Index (AAI), Security Index (SI), Vulnerability Index (VI), and Latency Index (LI). A synthetic-data testbed is developed with 30 independent simulation runs for each system, which is used for comparison with SecNet and with a Traditional-Blockchain baseline and with a Cloud-Centralized baseline. SecNet improves the Mean Algorithmic Execution Latency by 71.9% over Traditional Blockchain and 66.7% over Cloud-Centralized, the Vulnerability Index by 79.0% over Traditional Blockchain, and the Access Authorization Index and Security Index by 68.3% and 50.2% over Traditional Blockchain baseline, respectively. Overall, the improvements are statistically significant (p < 0.001), according to Welch's ttests, and the individual contributions of AI-based access authorization and smart-contract enforcement can be determined through an ablation study. The results are seen as a repeatable, scalable and quantitatively-based model for trustworthy data sharing in distributed cyber-security ecosystems in the context of SecNet.