Network attacks have significantly increased as a result of Internet of Things (IoT) technologies and the extensive use of wireless networks. So, Intrusion Detection Systems (IDS) were required for safe and secure networks. In this research, an Unsupervised Machine Learning Called Improved Isolation Forest is proposed to identify abnormalities in network traffic communication systems. The NSL-KDD dataset is used to gather and store the data. In order to secure the IoT network, it processes the data using One Hot Encoding which is utilized for analyzing the anomalous activity. The suggested improved isolation forest was then used to train the anomaly detection model based on network data in order to identify abnormalities and potential threats. The anomalous score was analyzed for evaluating the performances. The results illustrated that the proposed Isolated Forest approach outperformed than existing models like SVM, Naïve Bayes, and Automated ML, with superior performance accuracy of 99.45%, precision of 99.56%, F1-measure of 99.80%, and recall of 98.12%.