Published Paper Details:

Cryptography is a fundamental means for ensuring digital information confidentiality, integrity, and authenticity. Not every encryption algorithm is that powerful and effective in design, key size, and implementation. This paper introduces a machine learning system for automatically classifying the security grades of cryptographic algorithms from encrypted image data by feature extraction. The system classifies encryption outcomes into three grades: Weak, Acceptable, and Strong, in terms of statistical and structural image features.The method begins with a set of grayscale images and color images. An image is encrypted by using one of five algorithms: AES, DES, ChaCha20, Caesar Cipher, or XOR. In order to accommodate color images, features are computed independently on RGB channels and averaged to generate homogeneous descriptors. Grayscale images are handled natively. 14 features are obtained from each encrypted image, including entropy, contrast, correlation, energy, homogeneity, PSNR, MSE, skewness, kurtosis, SSIM, edge mean, edge standard deviation, edge entropy, and edge density. These features describe the statistical complexity and structural transformations caused by encryption.For dataset labeling, feature standardization and quantile-based binning are used collectively to label every sample with a security level. This is to enable an even dataset to prevent model bias. The labeled data are utilized to train a Support Vector Machine (SVM) classification with hyperparameters optimized via grid search and cross-validation. Comparison is also made using Random Forest and XGBoost models based on performance. Feature selection tools such as SelectKBest and Recursive Feature Elimination (RFE) are also utilized to identify top contributors to classification. Experimental results show the classification accuracy of 59% with high precision and recall for highly encrypted samples. F1-scores and confusion matrices confirm model reliability. Correlation, PSNR, and entropy features are found to be strong predictors of encryption strength.This paper presents a practical and scalable solution to the evaluation of encryption effectiveness using image analysis with security audit, encryption verification, and digital forensics implications. Future work is the addition of real-time analysis support and the extension to other media formats like audio and video.