1. An ensemble approach-based intrusion detection system utilizing ISHO-HBA and SE-ResNet152.
- Author
-
Saikam, Jalaiah and Ch, Koteswararao
- Subjects
- *
OPTIMIZATION algorithms , *BLACK market , *COMPUTER networks , *DECISION trees , *INTRUSION detection systems (Computer security) , *NETWORK computers - Abstract
Over the past ten years, there has been a significant increase in computer network intrusions, partly due to a thriving black market for cybercrime and the availability of advanced tools for committing such breaches. The most effective method for stopping unwanted intrusions and identifying abnormal network behavioral patterns is an intrusion detection system (IDS). In IDS, transfer learning techniques are frequently employed. An ML-based IDS experiences problems with data imbalance and a greater false detection ratio due to a small training dataset. These ID systems can quickly and automatically recognize harmful threats. The network requires a complex security solution because dangerous threats constantly develop and appear. As a result, developing an efficient and intelligent ID system is a substantial scientific challenge. We suggested an effective ensemble strategy that improved the spotted hyena optimization algorithm (ISHO) and the honey badger algorithm (HBA) to address the data imbalance and overfitting problem. The dataset is balanced by increasing the number of data samples and the detection precision. The Squeeze-and-Excitation (SE)-Deep Residual Network 152 (SE-ResNet152) approach is utilized to remove the less critical features. Every iterative phase includes using a list of decision trees, which monitor the performance of the categorizer and prevent overfitting issues. We use the datasets UNSW-NB15, CSE-CIC IDS 2018, and CICIDS2019 to simulate and assess the model. Compared to other approaches, the proposed approach performs well on three datasets and obtains above 99% accuracy, precision, recall, and F-measure. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF