Enhanced Network Intrusion Detection using ResNet50 and LSTM: A Deep Learning-based Cyber Defence Approach

Abstract

Concern about cybersecurity has grown among businesses of all sizes and in all industries as the frequency of cyberattacks has increased. Recently, deep learning (DL) and artificial intelligence (AI) have arisen as potent enabling technologies for bolstering cyber defense systems. These methods outperform conventional rule-based systems in several respects, including their ability to detect threats in real-time, analyze user behavior, and make automated decisions. Ensuring robust adoption and practical deployment of AI-driven cybersecurity requires significant worldwide collaboration between industry, academia, and government. This research delves into the latest deep learning techniques employed by NIDS, specifically ResNet50 and LSTM architectures, which are known for their effectiveness in detecting intrusions. Testing the accuracy, consistency, and practical applicability of ML models is required before fully automating cyber detection systems, even if they have already been used to supplement or even replace first-level security analysts. Analyzing ML and DL methods for spam, malware, and intrusion detection is the main focus of this study. Results from experiments show that the ResNet50-based model is best for high-accuracy intrusion detection because of its high recall, precision, and F1-score; on the other hand, the LSTM model improves temporal threat awareness by effectively capturing sequential patterns in network traffic.