Deep Learning-Based Attack Detection in the Internet of Things

The exponential growth of the Internet of Things in various domains has escalated the rise of concern in this digital era. The concern is primarily due to the evolution of cyber-attacks leading to the emergence of numerous threats and anomalies. The bottlenecks in traditional security techniques have unclouded the vision of learning techniques for intrusion detection. The use of classical ML techniques for the identification and classification has been around for a long time, but it suffers from the issue of scalability and feature engineering, which limits its usage. In this paper, we have analyzed the use of deep learning for intrusion and anomaly detection. The deep learning algorithms used here are deep neural networks (DNN) and long short-term memory recurrent neural networks (LSTM). Denial of service, malicious control, wrong setup, data type probing, scan, spying, malicious operation are the attacks against which the algorithms are tested. An accuracy of 99.30% is achieved for deep neural networks and 97.50% for the LSTM model.