An Adaptive Detection Mechanism for IoT Devices Anomalies Using AI/ML Based on User Pattern

Internet of Things (IoT) has been showing tremendous growth across the multiple domains from smart home/cities, finance, medical, industry, agriculture and commercial management, but security challenges for IoT growth rate also in similar lines. IoT devices have unique IP address which helps to connect across the devices using different protocols (OCF, ZigBee, Zee wave, …) standards and data misuse on the connection layer raises flaws of devices design and communication channels. In this paper, we proposed novel AI-based method to identify anomaly in IoT system based on user usage patterns and announcement to respective stakeholder for actions. Usage patterns may be like user interactions and operation pattern with IoT devices or the IoT devices events pattern in an IoT system. The proposed framework has novel aspect to identify IoT system anomaly using linearly built multilayered AI model to learn context, sequence and point type of anomalies through efficient and accurate manner. The dataset is efficiently reshaped and moved across the layers to increase the coverage and accuracy. The proposed framework has been applied into Samsung IoT platform named as SmartThings, which helped to detects all types of anomalies more efficiently and accurately with covering more than 90% anomalies with accuracy of > 95%. The proposed framework had a potential to expand for any IoT domains to detect anomalies and provide next course of actions like security alert system, business actions, industry parts manufacturer, connected systems, etc.