Enhancing Cybersecurity: Zero-Day Attack Detection in Network Traffic with Deep Learning Model
Sri Krishna Kireeti Nandiraju
*
University of Illinois at Springfield, United States.
Sandeep Kumar Chundru
University of Central Missouri, United States.
Mukund Sai Vikram Tyagadurgam
University of Illinois at Springfield, United States.
Venkataswamy Naidu Gangineni
University of Madras, Chennai, India.
Sriram Pabbineedi
University of Central Missouri, United States.
Ajay Babu Kakani
Wright State University, United States.
*Author to whom correspondence should be addressed.
Abstract
A big risk to networks comes from zero-day attacks since no patches are available until after the attack takes place. Such attacks can escape detection by traditional, signature-based IDSs. Therefore, better analysis methods are needed to catch threats that examining data has not revealed before. A zero-day hack is the most dangerous thing that can happen to network security. Zero-day attacks are hard to spot because they act in ways that haven't been seen before. A lot of people are interested in intrusion detection systems (IDS) because they can find these kinds of threats. Machine learning (ML) and deep learning (DL) are being used a lot more in intrusion monitoring systems. It has been shown that these methods can find zero-day risks. The study aims how to detect Zero-Day Attack in network traffic with deep learning model and to enhancing cybersecurity. This study shows a way to find cyberattacks using deep learning and a Convolutional Neural Network (CNN) trained on the UNSW-NB15 dataset. Adopted the UNSW-NB15 dataset to reflect real-world and comprehensive cyber-attack patterns. The collection has a lot of different types of real-world network attacks. Support Vector Machine and Random Forest are not as good as the CNN model that was mentioned. The CNN model achieved the highest accuracy at 93.8%, demonstrating its superior capability in capturing complex patterns in network traffic data. A score of 95.1% means it is mostly correct, 93.8% means it is mostly precise, 96.5% means it remembers things, and 94.6% means it is most likely correct. Experiments show that the model is good at learning, generalising, and being sturdy without becoming too perfect. This proves that it can find complex zero-day attacks. This study contributes to enhancing cybersecurity through an efficient and reliable deep learning framework for network traffic analysis. The power grid and other similar systems play a big role in keeping cyber-physical systems safe. When someone breaks the control code in a power grid system, it could cause a lot of damage. If the issue is found early, it will do less damage in the future. Future tasks involve collecting more diverse datasets, building combinations of existing approaches and validating the model directly on-site during real cyber-attacks.
Keywords: Cybersecurity, zero-day attack, network traffic analysis, threat detection, internet of things