Utilizing Machine Learning Models to Determine the Security Level of Different Cryptosystems

Abstract

Due to recent developments in multimedia technology, digital data security has emerged as an important concern. To improve upon the state of the art in terms of security, researchers often recommend modifications to procedures that have previously been implemented. However, many suggested encryption algorithms have proved unsafe over the previous several decades, putting sensitive data at risk. It is crucial to use the most suitable encryption strategy to defend against such attacks; nevertheless, the type of data being protected might affect the method that is most appropriate for each given situation. However, systematically evaluating various cryptosystems to choose the optimal one may consume significant computational effort. To rapidly and reliably choose the appropriate algorithm, an SVM (support vector machine) is proposed as a security-level identification tool for photo encryption algorithms. In this research, a dataset was compiled with the help of common encryption security criteria, including Security, Peak signal to noise ratio, Homogeneity, Correlation, Contrast, Energy, Entropy, and Mean Square Error. These values are used as extracted characteristics from various cypher pictures. There are three tiers of security for dataset labels: strong, acceptable, and weak. For evaluating the effectiveness of the proposed model, the calculated accuracy and results demonstrate the value of this SVM system.