CoNN-IDS: Intrusion detection system based on collaborative neural networks and agile training. Issue 122 (November 2022)

• Record Type: Journal Article
• Authors: Lee, Jung-San; Chen, Ying-Chin; Chew, Chit-Jie; Chen, Chih-Lung; Huynh, Thu-Nguyet; Kuo, Chung-Wei
• Is Part Of: Computers & security. Issue 122(2022)
• Publication Date: 2022-11
• Subjects: Intrusion detection system -- Unknown attack -- Collaborative neural networks -- Agile training -- Group decision; Computer security -- Periodicals; Electronic data processing departments -- Security measures -- Periodicals; 005.805
• DOI: 10.1016/j.cose.2022.102908
• Languages: English
• ISSNs: 0167-4048

Abstract

Abstract: The intrusion detection system (IDS) is the first and the most significant shield in Internet security. Nevertheless, with regard to confronting unknown attacks, such as zero-day attacks, it is difficult for IDS to detect and respond. To address this issue, we have designed a brand-new IDS framework to detect vague attacks and learn new types of attack more effectively based on the collaborative neural network (CoNN-IDS). The framework consists of an unknown attack detection (UAD) module and a malicious behavior parsing (MBP) module, where the autoencoder and the deep neural network serve as the UAD module and the MBP module, respectively. Furthermore, the aggregation strategy based on the DBSCAN algorithm and group decision system is employed to retrain the IDS for long term use. In particular, an agile training mechanism is introduced to enhance the efficiency of our model by retraining low-impact neuron. The experiments have been simulated with two individual datasets to reveal that the new method has a higher accuracy in unknown attack detection and shorter time spent on the retraining phase, which collectively evidence its advancement compared with the general IDS system.