Zero-Shot Based Hybrid Neural Network System for Enhancing Zero-Day Attack Detection

Zero-Shot Based Hybrid Neural Network System for Enhancing Zero-Day Attack Detection

연구 분야: Strategies

학회: 2024 IEEE 21st India Council International Conference (INDICON)

Ensuring network security is increasingly vital as our digital interactions and network infrastructures expand. With the continuous evolution of cyber threats, particularly zero-day attacks that exploit unknown vulnerabilities, there is a pressing need for advanced intrusion detection systems (IDS). These zero-day vulnerabilities, undetected and unpatched, pose significant risks, especially in the burgeoning Internet of Things (IoT) landscape, which has seen a dramatic increase in such attacks. To address this challenge, we developed a hybrid deep learning model combining the strengths of CNNs and LSTMs, creating a robust framework for detecting zero-day attacks. Our model was trained on the CICIoT2023 dataset, a comprehensive dataset specifically designed for IoT security. The evaluation, conducted using a confusion matrix, demonstrated the model's efficacy in accurately identifying both known and unknown threats. The results indicate that the proposed hybrid deep learning approach significantly improves the detection of zero-day attacks, offering a promising solution for enhancing network security in an increasingly connected world. This work highlights the potential of advanced deep learning techniques in creating more effective IDS and enhancing the overall resilience of network infrastructures.