Establishing Cloud Security using Modern Learning Approaches

During the present times, it is very challenging and crucial to develop an efficient as well as simple system for detecting attacks and ensuring security against cyber-threats in cloud systems. For the cloud network's security, different kinds of detection methodologies based on machine learning were developed in traditional work. However, that methodology faces difficulties like taking too much space and time, complexity in understanding, and complex system designs. Therefore, to provide the cloud system with superior security, the proposed work intends to formulate and develop a model for the detection of cyber threats. Optimization, extraction, feature pre-processing, normalization and prediction are the modules present in the proposed framework. Relevant features are extracted from Principle Component Analysis (PCA). For operations of testing and training, a model named baseline classifier model (LR, DT and RF) is used to select the relevant features. In the final stage, a BERT classifier is employed to predict if it entertains intrusive or normal data flow. Three benchmark datasets are available for the public and popular to implement the proposed system. Parameters for performance analysis like F1-score and detection accuracy (AC) are measured for the validation stage, and the same is compared with the proposed system proposed by us. Superior performance results like ACC value of 99% and F1 scores of 98% is achieved on average by the anticipated system when experiments are conducted in the given datasets. Other traditional security models were outperformed by the proposed method, and the outcomes prove the same. As a result, the research proposed contributes considerably in cloud systems security fortification in contemporary times by making the cloud system secure against threats simply and efficiently.