Deep Learning-Based Intrusion Detection System for Cloud Infrastructure Security

Cloud infrastructure security has become a main issue that needs to be addressed since organizations are moving their sensitive operations to the cloud and thus are exposing their systems to different kinds of cyber threats like DDoS attacks, brute-force attempts, port scans and insider misuse. Traditional rule-based intrusion detection systems do not adapt to changing attack patterns most of the times; thus, advanced solutions should be developed. One of the key contributions of this research work is a Deep Learning-Based Intrusion Detection System (DL-IDS) utilizing UNSW-NB15 dataset which is designed for providing realistic traffic patterns including both normal and malicious activities. First step in the methodology comprises of applying preprocessing techniques such as one-hot encoding, normalization and feature selection, then followed by feature extraction through an autoencoder which helps in dimensionality reduction and noise elimination. Next, fully connected deep neural network (DNN) is used used for classification, optimized with Adam algorithm also early stopping to secure strong training. The experimental performance shows fantastic results measured in several metrics with 99.12% accuracy, 98.87% precision, 99.54% recall and 99.21% F1-score. Such outcomes point out the model’s ability to operate in middle with respect to sensitivity and specificity which is quite significant to detective reliably the malicious traffic and at same time keep the false alarms very low. High recall figure means that it is very effective in reporting true positives while the precision figure is the corroborating factor of the low rate of clerical mistakes made on the benign traffic. The F1-score is yet another proof of the system’s balanced performance, thus, it can be deployed in real-time in cloud environments. In sum, the DL-IDS framework proposed offers a dynamic, scalable, and effective intrusion detection method which overcomes the drawbacks of traditional systems and presents a significant improvement in the area of cloud infrastructure security.