@article{dziedzic2026performance,
  title = {Performance Evaluation of AI-Driven IDS: From Offline Forensic Accuracy to Real-Time Implementation Challenges},
  author = {Maurycy DZIEDZIC and Piotr KOZIKOWSKI and Maciej SOBIERAJ},
  year = 2026,
  url = {https://ibimapublishing.com/articles/CIBIMA/2026/467519/},
  journal = {Communications of the IBIMA},
  volume = 2026,
  pages = 10,
  doi = 10.5171/2026.467519,
  abstract = {This study presents a comparative performance analysis between a modern Artificial Intelligence-based Intrusion Detection System (AI-IDS) and the traditional Snort 3 platform. While traditional signature-based systems like Snort are highly efficient for real-time traffic processing, they often struggle to detect previously unknown "zero-day" threats. To address this, a proprietary AI-IDS utilizing the RandomForestClassifier algorithm was developed and tested within a virtualized environment against UDP flood attacks. The proposed AI model achieved 99% accuracy in detecting UDP flood attacks, demonstrating superior adaptability and predictive capabilities. However, testing revealed that the AI system's current Python-based implementation is better suited for offline forensic analysis due to real-time performance bottlenecks. The findings suggest that a hybrid architecture, combining the efficiency of signature-based methods with the precision of machine learning, provides the optimal defense against evolving cyber threats.},
  keywords = {intrusion detection system, artificial intelligence, machine learning, Snort, RandomForestClassifier, cybersecurity. },
  note = Article ID: 467519
}