LSTM-Markov based efficient anomaly detection algorithm for IoT environment

Abstract

The seamless integration of wireless and IoT device in normal day to day life and for smart homes has enabled more security and privacy needs. The attack or unauthorized access to these devices may result in issues in decision making inside the smart environment. Attacks and anomalies in open IoT system could provide false alarms and cause delay in processing the information’s. The problem statement considered is, Anomaly detection systems, on the other hand, can be targets of attacks, h/w s/w failures, and thus fall short of their objectives. Because these are power-hungry devices carrying highly sensitive data, an effective attack and anomaly alert system is critical in an IoT-based environment. The present algorithms require high training and additional memory to identify the anomalies in the network, which is not practically feasible to simple edge-based computing devices. The anomalies and false information inside the network are handled with effective anomaly detection algorithm designed in this paper. An efficient anomaly detection method in real-time sensor is identified through markov and LSTM based network and the outliers in the data is clearly removed through the proposed approach. The proposed approach is tested with the real-time DHT sensor monitoring room temperature and room humidity. The proposed methodology provides 96.03% effective anomalies detection with 92.48% high training accuracy. The methodology showcase improved with 6.54% of effective anomaly rejection and 5.13% of training accuracy when compared with KNN algorithm.