A Framework for Effective Intrusion Detection and Mitigation at the Edge of Things Using Deep Learning Architectures

Edge computing lets data and tasks run close to where people use them. That usually speeds up response and can cut down on bandwidth.  Still, the Edge of Things raises privacy and security issues. These issues show up across the layers of an edge setup. As a result, it becomes harder to do intrusion detection in systems that are not centralized. Current intrusion detection systems (IDS) face issues with complexity and accuracy, rendering them unsuitable for resource-limited edge devices. To address this, an efficient lightweight deep learning-based IDS is proposed for EC to safeguard EoT devices. The framework extracts key network features using MNasNet and the elastic collision seeker optimization algorithm (ESOA), followed by SqueezeNet for intrusion detection and classification. A defensive module restores network connectivity post-attack, while the Auxiliary Classifier Generative Adversarial Network (ACGAN) approach effectively mitigates data imbalance. Validation with Edge-IIoT and IoT-23 datasets shows the system excels in identifying diverse attack types in the EC environment.