Designing Interpretable Ai Models With Lightweight Parallelism For Real-Time Malware Detection And Prevention
Document Type
Conference Proceeding
Publication Date
1-1-2026
School
Computing Sciences and Computer Engineering
Abstract
Network intrusion detection and prevention systems (NIDS/NIPS) have become less effective due to the growing complexity and sophistication of cyber threats. Traditionally, these systems use static signatures and rule-based logic to detect and subsequently prevent threats and malicious activities. However, new malware techniques, including obfuscation and evasion methods as well as adaptive behavior, critically weaken standard detection systems and create major threats to government and defense infrastructures as well as commercial networks. This research introduces a scalable, interpretable, and low-latency malware detection system that fuses parallel agentic AI, retrieval-augmented intelligence, and transparent decision pathways, offering substantial progress in operational cybersecurity environments. By using task-level parallelism the system distributes feature extraction analysis along with behavioral profiling and threat attribution tasks among multiple processes to achieve scalable performance even under heavy load conditions.
Publication Title
Communications in Computer and Information Science
Volume
2720 CCIS
First Page
393
Last Page
405
Recommended Citation
McCullough, Z.,
Martinez, J.
(2026). Designing Interpretable Ai Models With Lightweight Parallelism For Real-Time Malware Detection And Prevention. Communications in Computer and Information Science, 2720 CCIS, 393-405.
Available at: https://aquila.usm.edu/fac_pubs/22097
COinS