Detecting IoT Attacks using Multi-Layer Data Through Machine Learning

Internet of Things (IoT) devices is being used in countless network applications. However, due to their insecure nature, the wide adoption of these devices has also increased the possibility of cyber-attacks. There is a need for a robust security mechanism to detect and safeguard against numerous threats. Machine Learning (ML) techniques have been used to detect attacks on different networking layers but training only the network, transport, or link-layer data has proven to be inadequate. Thus, opening paths for attackers to take control and penetrate the networks. Leveraging from this inadequacy, we have employed Machine Learning technology to detect attacks on IoT devices using the application, transport, and network layer data. In particular, we have focused on feature extraction of Application layer data to identify nefariousness in packets. Furthermore, for packet classification, we are also extracting features from the network layer and transport layer. Our simulation results have promised accuracy of 88% and 92% using different ML algorithms. We have also identified possible future work that can be used to validate the solution.