Security-Oriented Text Classification Using Optimized Naive Bayes in C++: A Lightweight Approach for Big Data Analytics

Abstract

With the increasing volume of unstructured text in cybersecurity environments, accurate and efficient classification of malicious content has become a major challenge. This study addresses the problem by designing and implementing a lightweight Naive Bayes classifier using C++ to detect and classify malicious versus benign software-related text messages. The proposed solution focuses on resource-constrained environments, ensuring secure processing and minimal memory overhead. A synthetic dataset containing 60 labeled messages (30 malicious, 30 benign) was used for testing. The classifier achieved 90% overall accuracy, 86.7% sensitivity, and 93.3% specificity. Performance was validated through confusion matrices, precision-recall analysis, and per-class evaluation. The findings confirm that Naive Bayes is a viable lightweight method for real-time text classification in security-sensitive big data applications. Future enhancements may include NLP integration, hybrid classifiers, and privacy-preserving machine learning techniques.