https://doi.org/10.1140/epjp/s13360-025-07001-9
Regular Article
BayesIDS: a Bayesian-optimized learning-based intrusion detection system
1
Department of CSE, BMS Institute of Technology and Management, 560119, Bangalore Karnataka, India
2
Department of AIML, , BMS Institute of Technology and Management, 560119, Bangalore, Karnataka, India
3
Department of CSBS, BMS Institute of Technology and Management, 560119, Bangalore, Karnataka, India
4
Department of AIML, BMS Institute of Technology and Management, 560119, Bangalore, Karnataka, India
a
This email address is being protected from spambots. You need JavaScript enabled to view it.
Received:
29
April
2025
Accepted:
27
October
2025
Published online:
6
November
2025
Abstract
The power of intrusion detection systems (IDS): It serves as the backbone of modern network security, protecting against various threats, including malware, DoS attacks, and advanced persistent threats. Traditional IDS methods, such as signature-based and anomaly-based systems, face challenges in identifying zero-day attacks and changing, unpredictable threats. They also struggle to generalize across datasets and remain robust in the presence of adversarial attacks, where malicious changes are made to the input data to trick the model. As such, advanced detection frameworks that utilize deep learning capabilities are increasingly essential for boosting performance, generalization, and endurance, given these shortcomings. This paper presents BayesIDS, a hybrid deep learning-based intrusion detection framework that combines Convolutional Neural Networks, Bidirectional Long Short-Term Memory, and machine learning-based anomaly detection approaches. It attempts to model spatial and temporal patterns/types, which provides an improved accuracy against different classes of attacks, such as zero-day attacks and adversaries. Following that, a model was developed utilizing feature extraction and accurate anomaly detection, and it was shown to outperform previously used methodologies. On multiple datasets, including CIC-IDS2017 and UNSW-NB15, experimental results show superior performance of BayesIDS in terms of accuracy, recall, and F1 score compared to existing models. The model demonstrates strong generalization across datasets and resilience to adversarial perturbations. The experimental findings indicate that the presented framework is effective and practical, providing a promising basis for real-time, scalable, and adaptive network security systems.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Società Italiana di Fisica and Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
