Federated Learning Assisted IoT Malware Detection Using Static Analysis
Madumitha Venkatasubramanian, Arash Habibi Lashkari, Saqib Hakak · 2022
The Internet of Things (IoT) has millions of connected devices and has paved the way for a highly connected society. IoT devices have also increased the threats caused by malware and its variants. In addition, IoT devices are often heterogeneous, receive updates in infrequent intervals, and remain out of sight for prolonged periods, causing security and privacy challenges to the IoT systems. With technological advances, there are various emerging techniques to address this problem. The previous research works have focussed on utilizing centralized learning techniques such as Machine Learning and Deep Learning for IoT malware detection. But these techniques do not efficiently protect confidential user data as they share the user data with a centralized model and receive back the updated model from the centralized servers causing several privacy and security concerns. This paper proposes a Federated Learning-based approach that employs a random forest model for detecting IoT malware samples. Through the Federated Learning solution, we ensure that the local IoT device data stays locally and is not moved off the device. The results from our proposed model achieve an accuracy of 95% and efficiently classify the malware and benign samples. The overall comparative results between our proposed decentralized model and a centralized format demonstrate a significant improvement in the accuracy of malware detection while preserving the privacy and security of the user data.