Authors: Assistant Professor P. Cathrine Ranjana, Assistant Professor Dhanusha Mol K P
Abstract: With the continuous increase in complexity of cyber threats like zero-day attacks and APTs, the effectiveness of conventional signature-based intrusion detection approaches becomes less relevant. This paper presents a novel approach of AI-based cyber threat intelligence (CTI) system by incorporating deep learning and real-time threat intelligence correlation capabilities for detecting such advanced attacks. This research uses CNN-LSTM model to detect cyber threats, which achieves an accuracy of 94.2% with CICIoTDataset2023. To enrich the system with threat intelligence, RAG technique along with large language models (LLMs) is used in the proposed framework for recognizing zero-day cyber attacks. The proposed CTI solution detects zero-day cyber attacks accurately with 98.5% accuracy. It also offers substantial
