Comparative Deep Learning Approach for Intrusion Detection
Nabin Kumar Sah, Mahi Kolli, Kadam Prajwal Dharmaraj, H. N. Vishwas · 2024
In the modern world, the security of computer networks from harmful activities is very important since cyber threats keep changing from time to time. However, Intrusion Detection Systems (IDS) have not been able to match this pace using traditional methods. This research is on checking how well deep learning models which are Convolutional Neural Networks (CNNs), Gated Recurrent Units (GRUs), Long Short-Term Memory (LSTM) networks, and Deep Neural Networks (DNNs) help in enhancing Intrusion Detection capabilities. The NSL-KDD dataset available on Kaggle provides various attack types and normal activities through diverse network traffic data. Rigorous data preprocessing has been done during the study through feature engineering, encoding categorical variables, and normalizing numerical attributes. After that, custom deep learning architectures are created and evaluated with accuracy being one of the standard metrics used for testing their performance among others like recall, precision, or F1-score so as to improve intrusion detection efficiency.