Streamlined network intrusion detection: Feature selection optimization for higher accuracy and efficiency
Kunda Suresh Babu, V. M. S. R. Srinivas, Yamani Chandana, G. Satish, Rajesh R. Naik, Dodda Venkata Reddy · 2025
The primary task of network intrusion detection systems (NIDS) is to defend communication networks from any attack attempts. However, the problem arises in analyzing a large amount of such data since it often contains irrelevant and noisy data that increases computation time and decreases the detection accuracy. In order to overcome such difficulties, a new approach to feature selection optimization was created that targets key features that have a high influence on the value of the target variable. By using this approach, the irrelevant features were cut down by 35%, a tuned decision tree classifier achieved a detection accuracy of 99.91%, and the computation time has been reduced to more than 65% compared to the CICIDS-2017 dataset with all features. This method makes sure that only the necessary information gets processed, which in turn increases the performance and effectiveness of NIDS, which is important in quick response and targeted threat evaluation in the ever-evolving cyber world.