Machine Learning for Intrusion Detection for Massive IoT Networks

Wide area networks such as fog and internet of things often encounter network level security. There would exist a continued trade-off between the error rate (authentication metric), system overhead, computational complexity and latency of the system. Hence an extremely meticulous system design with appropriate choice of stochastic parameters and authentication scheme should be adopted. In this proposed work, an acceleration learning based LSTM network has been proposed to detect attacks in IoT networks. It can be observed from the obtained results that the proposed system attains better performance compared to previously existing system. The performance enhancement can be attributed to additional features computed and the LSTM with acceleration used to train and further detect errors.