Anomaly Detection in IOT Site Using CatBoost
Rahul Kushwah, Ritu Garg · 2023
IoT environment is growing rapidly due to availability of large number of connected components like sensors and actuators. Due to heterogeneity and constrained nature of IoT devices, they are vulnerable to security issues, including attacks from unauthorized sources and anomalous behavior within the system. Anomaly detection helps to mitigate these security risks in the IoT infrastructure. Thus, the investigation in IoT network for detecting anomalies is a fast-emerging subject. There are various techniques developed by the researchers for the detecting anomalies and work on improving accuracy, precision etc, yet rarely consider the categorical data. In this paper, we presented a CatBoosting approach which enhances the anomaly detection for IoT devices and handle the categorical data. The proposed approach effectively monitors and distinguishes between normal and abnormal activities in the data by using the optimized techniques like gradient boosting and decision tree. To evaluate its effectiveness, the DS2OS dataset is considered and achieve the accuracy of 99.4% and precision 98.7%.