Machine Learning Empowered: Support Vector Machine-Based Selection of Encryption Techniques for Digital Image Security Levels

Recent advancements in multimedia systems have increased the demand for strong digital image security mechanisms. Traditional cryptosystem evaluations, dependent on manual statistical analysis, are computationally exhaustive and not scalable. This study proposes a machine learning-based framework, such as a Support Vector Machine (SVM) classifier, for the categorization of image encryption levels into three discrete classes: Strong, Acceptable, and Weak. The model is trained using statistical descriptors such as Peak Signal to Noise Ratio (PSNR), entropy, Mean Square Error (MSE), energy, correlation, homogeneity, and contrast extracted from encrypted image datasets. Feature normalization techniques, StandardScaler, have been used to ensure balanced input contributions. The proposed system, i.e., SVM with a Radial Basis Function (RBF) kernel, outperforms other kernels. The performance of the proposed model shows an average classification accuracy of 98%, precision up to 100%, and an F1-score of 97%. A web-based interface developed using Django integrates the model, enabling real-time analysis and visualization, making the proposed system a scalable solution for cryptographic strengt