DEEP-INTRUSION DETECTION SYSTEM WITH ENHANCED UNSW-NB15 DATASET BASED ON DEEP LEARNING TECHNIQUES

Growth in the number of devices and data has raised serious security concerns, that have increased the importance of the development of advanced intrusion detection systems (IDS). Deep learning can handle big data and in various fields has shown a great performance. Consequently, security specialists are aiming to adopt deep learning in an intrusion detection system. Numerous studies have been done on this topic which have led to many different approaches. Most of these approaches use predefined features extracted by an expert in order to classify network traffic. In addition, UNSW-NB15 dataset was developed in different separated files and labelled based on binary classification, in this research, we aim to merge the whole dataset to be in one file so it can test models once, instead of test models separately for each file. then used attacks families in the dataset as new label so that it will develop multi-classification labelled dataset. We investigated the performance of deep learning with the enhanced dataset, within two classification categories (Binary and Multi-Class). We compared our proposed deep learning model results with related works. We have used accuracy and loss to evaluate the efficiency of deep learning and machine learning models in the enhanced dataset. Our proposed Deep learning models Performed yielded accuracy of 99.59% in multi-class classification and 99.26% in binary classification.