Secure Anomaly Detection with Light GBM and Enhanced Privacy Preservation

Cloud infrastructure is essential for modern businesses and other services since it allows for scalable and reliable operations. This work entails storing and retrieving gigantic amounts of information that have the potential to revolutionize human life cycle, business, education, and related domains, so altering the world in profound ways. The term "intrusion detection" refers to the steps used to find instances of data breaches in cloud infrastructure that do not adhere to established security protocols. In order to find out how machine learning algorithms have worked in real-world situations, that is what this thesis is all about. More specifically, this effort aims to provide a secure anomaly detection system that can be implemented and has higher performance in terms of execution time and classification accuracy. As a means to increase the algorithms' verifiability, a hybrid approach was developed, combining the Light GBM algorithm for classification with RFE for feature selection. In a cloud setting, the Trusted Authority may give key pairs to both the user and the compute engine, enhancing the efficacy of the RSA method in protecting data privacy. In order to identify features and classify them, the computation engine uses the NSL-KDD dataset.