Lightweight Deep Learning-Based Malware Detection for Internet of Things Devices

Abstract

The rapid expansion of Internet of Things (IoT) devices has introduced numerous security vulnerabilities due to limited computational resources and heterogeneous network environments. Conventional malware detection methods often fail to identify sophisticated attacks while consuming significant computational power. This paper proposes a lightweight deep learning framework for malware detection specifically designed for resource-constrained IoT devices. The proposed model combines one-dimensional Convolutional Neural Networks (1D-CNNs) with attention mechanisms to extract discriminative behavioral features from network traffic and executable files. Model compression techniques, including pruning and quantization, reduce computational overhead without sacrificing detection accuracy. Experimental evaluation using benchmark malware datasets demonstrates superior performance in terms of accuracy, precision, recall, inference time, and memory utilization. The proposed framework provides an efficient cybersecurity solution suitable for deployment in large-scale IoT ecosystems.