Towards Lightweight Hybrid Deep Learning Approach to Malware Detection Enhancement For IoT Based Systems
Sakshi Mittal, Prateek Rajvanshi · 2025
With rapid growth in use of smart devices and internet of things (IoT) the risk of cyber attacks has also been raised. Light weight machine learning methods to identify and stop cyber-attacks has become a topic of interest for the research community. Such systems are constrained in resources so they require light weight models that can operate in low resource environment. This work presents the parameters that support in determining if the model can be considered as light weight model. This work also contributes a unique light weight deep learning model combining Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU) over federated architecture. The research presents promising results to counter the malware attacks with real-world publicly accessible IoT network dataset, the study assessed the performance of the model using Aposemat IoT-23 dataset. We achieved around 96% and validation accuracy reaching 0.94% with 5 epoch. The results of this work shall contribute to advance deep learning approaches in cybersecurity and provide insightful analysis for the construction of more flexible and robust malware detection systems.