Federated Learning for SMS Spam Detection: A Privacy-Focused Approach
Satvik Vats, Suryakant Shastri, Shiva Mehta · 2024
Herein proposes a new classification technique of SMS spam within a federated learning system, based on which the process occurs on the user’s device. The objective is to realize the perfect balance of the two; on the one hand, it should be proactive with spam detection, and on the other hand, it should have the highest respect for privacy protection. Our research demonstrates that this method has significantly boosted classification accuracy without privacy issues by equating an appropriate evaluation technique for extensive testing and assessment. The statistical analysis shows that our federated learning algorithm significantly reduces loss; according to the statistical analysis results, this value increases from 0.267 to 0.389 from the first to the tenth epoch. Also, the system reveals that this mechanics remains excellent and keeps showing values of more than ${9 0 \%}$, and the third epoch shows nothing but a high accuracy value of ${9 8. 3 \%}$. In effect, we analyze the cornerstone of our knowledge based on the method of the critical priorities potentially tightened privacy protection for federated learning. The federated learning solution offers a transparent approach that enables users to access essential services directly on their devices, and then the trained machine learning models are merged securely. This study underscores the competency to differentiate spam from anything useful in electronic communication and considers it a significant issue as regards privacy protection. The outcome gives a sense of hope in the idea of federated learning to have specific SMS spam. It creates a good gap between effectiveness and privacy protection, which is especially important considering that mobile communication covers a vast territory today.