Lightweight Deep Learning Models for Edge-Based Cyber Threat Detection

Abstract

The rapid proliferation of Internet of Things (IoT) devices and distributed networks has shifted the cybersecurity landscape toward the edge, where real-time threat detection is essential. However, deploying deep learning models on edge devices presents challenges due to limited computational power, memory constraints, and energy efficiency requirements. This paper explores the development and implementation of lightweight deep learning models for edge-based cyber threat detection, emphasizing architectural optimizations, compression techniques, and adaptive intelligence. By leveraging model pruning, quantization, and knowledge distillation, researchers have enabled high-performance models that operate effectively in resource-constrained environments. Furthermore, integrating these optimized models into edge networks facilitates faster detection of malicious activities, reduces data transmission to centralized servers, and enhances privacy by processing sensitive information locally. The paper also examines challenges such as adversarial robustness, model update synchronization, and data heterogeneity across edge nodes. Ultimately, lightweight deep learning represents a pivotal advancement in creating scalable, privacy-preserving, and energy-efficient cyber defense systems for next-generation edge infrastructures.