Cloud Data Anomaly Detection using Machine Learning

In cybersecurity, anomaly detection has one of the most important roles in discovering threats that are a menace to an organization's information security. This paper describes the implementation of an anomaly detection system employing Isolation Forests of the Azure Function kind. Utilizing the possibility of the cloud service offered by Azure, the solution provides an effective anomaly detection. Raw data is normalized, features are reconstructed to address cases of missing data and the model then applied is the Isolation Forest with the purpose of differentiating normal and anomalous behavior. Standard performance measures, like accuracy, are used to measure the performance of the model. Furthermore, we give a clearer understanding of the model through the confusion matrix and the correlation heat map, respectively. That is why our approach shows good results in anomaly detection with a low percentage of false positives, which makes it appropriate for use in cloud systems. The integration with Azure Functions further confirms the simplicity, efficiency, and accessibility of serverless computing in more complex cybersecurity solutions.