The increasing risk of cyber attacks necessitates the use of advanced and effective techniques for quick detection and categorization. Through the integration of preprocessing approaches and feature extraction techniques, the present study aims to improve cyber hacker breach categorization in terms of accuracy, precision, and recall. To be more precise, Z-Score standardization is used as a stage of preprocessing to normalize the data, and robust feature extraction is achieved by Independent Component Analysis (ICA) after that. For thorough breach categorization, the retrieved characteristics are subsequently combined using Multilayer Perceptron (MLP), Decision Trees (DT), and Naive Bayes classifiers. The research task assumes that Z-Score normalization, ICA, and MLP working together will produce better outcomes than any of them working alone. Experiments conducted on a wide range of cyber-hacking situations reveal that the suggested method works well, with improved memory, accuracy, and precision. The results highlight the value of integrating both preprocessing and feature extraction methods and highlight the opportunities for real-world security-related applications enabling active breach prevention and detection. From the outcomes attained, the proposed MLP yields accuracy rate of 91.50%, Precision of 0.90, Recall of 0.89. The tool used is Jupyter Notebook and the language used is python.