Robust, Effective and Resource Efficient Deep Neural Network for Intrusion Detection in IoT Networks

Internet of Things (IoT) devices are becoming increasingly popular and an integral part of our everyday lives, making them a lucrative target for attackers. These devices require suitable security mechanisms that enable robust and effective detection of attacks. Deep Neural Networks (DNNs) offer a promise, but they require large amounts of computational resources to provide better detection, and their detection capabilities can be exploited by adversarial attacks. Therefore, this paper proposes a method to train Fully Connected Neural Network (FCNN) for IoT security monitoring in a robust, effective and resource-efficient way. The resulting model is assessed against various benchmark datasets created using commercial IoT devices, such as doorbells, security cameras, and thermostats. Experimental results demonstrate the model's ability to maintain state-of-the-art accuracy and F1-score while reducing training memory and time consumption by 99.99 and 99.80 percentage points than its benchmark counterpart.