Deep Learning-Based Threat Intelligence Framework for Proactive Cyberattack Detection and Mitigation

The use of cybersecurity threat detection systems that were previously effective is becoming a thing of the past as a result of the spread and complexity of cyberattacks. This study suggests the CNN-LSTM architecture and the CIC-IDS2017 dataset as the foundation of the hybrid deep learning threat detection model. The extensive data pretreatment that is practiced by the offered technique to address the problem of severe imbalance between classes is covered by data cleaning, feature extraction, normalization, label encoding, and class balancing based on the Synthetic Minority Over-sampling Technique (SMOTE). After the 80: 20 split of the processed data into training and testing samples, the performance of the model is evaluated on standard indicators, such as as accuracy (ACC) and precision (PRE) and recall (REC), F1-score (F1) and ROC-AUC. With a 99.6% ACC rate, 97.5% PRE, 97.0% REC, 98.0% F1, and 0.99% ROC-AUC, the suggested CNN-LSTM model outperforms the state-of-the-art. The hybrid architecture has been effectively compared to other models due to its flexibility to large as well as real-time network intrusion detection systems.