Image Lock: Protecting Your Pixels

The increasing digitization of data has amplified concerns regarding the security of sensitive visual information. This project investigates the classification of images as sensitive or non-sensitive using various machine learning algorithms and introduces an innovative “locking” mechanism to secure sensitive images. Utilizing a dataset of 1,000 samples, each containing nine pixel intensity values (ranging from 0 to 255) and corresponding sensitivity labels, the system employs Logistic Regression, SVM, XGBoost, Random Forest, and Decision Tree algorithms for classification. Logistic Regression and SVM demonstrate the highest accuracy, effectively distinguishing sensitive images from non-sensitive ones. To ensure security, a novel locking mechanism is implemented, which modifies pixel intensity values using a key-based system, rendering sensitive images indecipherable while allowing recovery with the correct key. Visualizations of pixel distributions and model performance validate the efficacy of the proposed methods. This project highlights the practical integration of machine learning for image sensitivity classification and the application of pixel-level encryption for securing sensitive data, offering a scalable solution for protecting visual information in real-world scenarios.