DEEP LEARNING FOR SECURE MOBILE EDGE COMPUTING IN CYBER PHYSICAL TRANSPORTATION SYSTEM

Abstract

This study presents a novel deep learning–driven abnormality detection framework integrated with Mobile Edge Computing (MEC) for Cyber-Physical Transportation Systems. Conventional surveillance architectures rely heavily on centralized cloud infrastructures, leading to increased latency, bandwidth overhead, and potential privacy vulnerabilities. To address these limitations, the proposed mechanism performs real-time image analytics directly at edge nodes positioned close to data sources such as roadside cameras and in-vehicle sensors. The system employs a deep neural network trained to model normal traffic patterns and identify deviations indicative of anomalies, including accidents, irregular vehicle movements, and security threats. By combining spatial feature extraction with adaptive anomaly scoring, the framework enhances detection accuracy while minimizing false positives. Edge-based deployment ensures rapid inference, enabling immediate response to critical events without dependence on remote servers. the architecture reduces network congestion by limiting unnecessary data transmission and strengthens data security through localized processing. Experimental evaluation demonstrates improved detection speed, higher accuracy, and lower latency compared to traditional cloud-based approaches. The proposed solution offers a scalable and efficient paradigm for intelligent transportation monitoring, contributing to safer and smarter urban mobility systems.