parHSOM: A Novel Parallel Hierarchical Self-Organizing Map Implementation
Researchers have introduced parHSOM, a novel parallel architecture for Hierarchical Self-Organizing Maps (HSOMs), aimed at enhancing Intrusion Detection Systems (IDS) in cybersecurity. While HSOMs are valued for creating trustworthy and explainable AI-based IDS, their traditional sequential training process is slow when handling large datasets. This study investigates the impact of parallel computation on HSOM training efficiency. The parHSOM model was rigorously tested across two different testbeds, four output grid sizes, and five distinct cybersecurity datasets. Experimental results demonstrate that parHSOM consistently achieves faster training times compared to the standard Sequential HSOM algorithm, without any significant loss in detection performance or accuracy. This development addresses a critical bottleneck in processing large-scale security data, offering a more efficient solution for real-time threat detection. Furthermore, the work establishes a foundational platform for future research into parallel HSOM implementations, potentially accelerating the adoption of explainable AI in cybersecurity infrastructure. The findings highlight the potential of parallel computing to optimize machine learning models used in critical security applications.
Wire timeline
parHSOM: A Novel Parallel Hierarchical Self-Organizing Map Implementation
Researchers have introduced parHSOM, a novel parallel architecture for Hierarchical Self-Organizing Maps (HSOMs), aimed at enhancing Intrusion Detection Systems (IDS) in cybersecurity. While HSOMs are valued for creating trustworthy and explainable AI-based IDS, their traditional sequential training process is slow when handling large datasets. This study investigates the impact of parallel computation on HSOM training efficiency. The parHSOM model was rigorously tested across two different testbeds, four output grid sizes, and five distinct cybersecurity datasets. Experimental results demonstrate that parHSOM consistently achieves faster training times compared to the standard Sequential HSOM algorithm, without any significant loss in detection performance or accuracy. This development addresses a critical bottleneck in processing large-scale security data, offering a more efficient solution for real-time threat detection. Furthermore, the work establishes a foundational platform for future research into parallel HSOM implementations, potentially accelerating the adoption of explainable AI in cybersecurity infrastructure. The findings highlight the potential of parallel computing to optimize machine learning models used in critical security applications.
cs.AI updates on arXiv.org