Efficient Malware Classification with Spiking Neural Networks: A Case Study on N-BaIoT Dataset
Muhammad Umair, Wooi-Haw Tan, Yee‐Loo Foo · 2023
In recent years, there has been growing interest in the application of spiking neural networks (SNNs) for classification tasks. Compared to traditional neural networks, Spiking Neural Networks (SNNs) are a class of neural networks that model the dynamics of biological neurons, where information is represented in the form of spikes or action potential. This study explores the effectiveness of spiking neural networks (SNNs) in classifying the N-BaIoT dataset. SNNs model the dynamics of biological neurons and represent information through spikes or action potentials, enabling them to process temporal information and exhibit event-driven behaviour. This makes them a promising alternative for resource-constrained environments and low-power neuromorphic hardware implementations. The dataset was balanced and split into training and testing sets, and a two-layered SNN model with two LIF neurons was developed. The model achieved 71% accuracy on the test dataset, highlighting the importance of pre-processing steps for reliable results. While the accuracy may not be as high as some deep neural network models, the SNN’s event-triggered feature and sparse representation of information through spike patterns make it a promising alternative for classification tasks.