Empirical Evaluation of SNN for IoT Network Anomaly Detection

Internet of Things (IoT) devices flourish fast along with the rapid progress of wireless network technologies. Since IoT devices usually deal with sensitive information from nearby users, protecting them from malicious network activities is critical. Artificial neural network (ANN) based approaches are known to be effective in detecting network anomalies. However, it is hard for IoT devices to apply such approaches due to constrained resources. Spiking Neural Networks (SNN) is a new type of neural network that requires low power consumption and computational overhead, which is proper for IoT devices. In this paper, using several network intrusion datasets, we conduct extensive experiments to compare the performance of ANN and SNN for identifying network attacks. Our experiment results demonstrate that SNN yields comparable performance to ANN in terms of accuracy and outperforms ANN in detecting frequently appearing attacks while consuming less energy.