Improving Spam Intrusion Detection with the Machine Learning-Enhanced Chaotic Horse Ride Optimization Algorithm

Abstract

Automated spam detection, utilizing feature selection (FS) and machine learning (ML), categorizes and identifies unsolicited messages, like spam emails. The goal is to accurately differentiate and filter out spam, enhancing overall email security. This research presents the Chaotic Horse Ride Optimization Algorithm with Machine Learning-Driven Spam Detection and Classification (CHROA-MLSDC). CHROA-MLSDC efficiently classifies spam and non-spam through preprocessing and CHROA-based feature selection. It incorporates the Variation Auto Encoder (VAE) model and the Bat Algorithm (BA) for potential performance improvements. Simulations on Ling spam, Enron, Spam Assassin, and CSDM C2010 datasets demonstrate significant enhancements in precision, recall, accuracy, and execution speed compared to existing systems. CHROA-MLSDC achieves notable accuracy: 99.35% on Ling spam, 98.80% on Enron spam, 99.92% on Spam Assassin spam, and 98.02% on CSDM C2010 spam. Recall rates range from 97.82% to 98.05%. CHROA-MLSDC consistently outperforms similar methods, exhibiting accuracy from 96.91% to 97.55%. Execution time analysis reveals CHROA-MLSDC's consistently faster performance across all datasets. In summary, CHROA-MLSDC excels in spam detection, surpassing other methods across various evaluation metrics.

Keywords

• feature selection
• horse ride optimization
• machine learning
• metaheuristics
• spam detection