parHSOM: Parallel Hierarchical Self-Organizing Map for Faster Intrusion Detection
Researchers have developed parHSOM, a novel parallel implementation of Hierarchical Self-Organizing Maps (HSOMs) aimed at accelerating training times for intrusion detection systems (IDSs). Traditional HSOMs are trained sequentially, making them slow on large datasets. parHSOM leverages parallel computation to reduce training time without sacrificing accuracy. The architecture was tested on two testbeds, four output grid sizes, and five cybersecurity datasets. Results show consistent speed improvements over the sequential HSOM algorithm. This work addresses a key bottleneck in AI-based IDSs, which are critical for cybersecurity in the digital age. The paper is available on arXiv under ID 2605.08164.
Key facts
- parHSOM is a parallel HSOM architecture for intrusion detection.
- Traditional HSOMs are trained sequentially, causing slow training on large datasets.
- parHSOM was tested on two testbeds, four grid sizes, and five datasets.
- Performance metrics show parHSOM trains faster than sequential HSOM.
- The research focuses on the effect of parallel computation on HSOM training time.
- HSOMs are used for trustworthy, explainable AI-based IDSs.
- The digital age has transformed information processing, making cybersecurity crucial.
- The paper is published on arXiv with ID 2605.08164.
Entities
Institutions
- arXiv