Table 1.
The IDS works in the literature.
Fig 1.
Classification taxonomy for intrusion detection systems.
Fig 2.
ML algorithms in cyber security.
Fig 3.
Basic methodology of an ML-based intrusion detection system.
Fig 4.
Basic methodology of an ML-based intrusion detection system.
Fig 5.
(a) One-hot encoding (b) Binarisation and discretisation process for continuous features.
Fig 6.
Flowchart of proposed architecture for intrusion detection.
Fig 7.
Images of intrusion data samples.
Fig 8.
(a) Residual block and (b) basic architecture of the ResNet50 model [59].
Fig 9.
Basic architecture of the GoogLeNet model [60].
Fig 10.
Basic architecture of the AlexNet model [61].
Table 2.
Hyperparameters of the suggested model.
Table 3.
Summary of public benchmark datasets.
Table 4.
Composition of 10% KDDCUP’99 and NSL-KDD datasets.
Table 5.
Distributions of samples in the UNSW-NB15 dataset used for training.
Table 6.
The evaluation metrics formulas.
Fig 11.
Quantitative results for the (a) UNSW-B15, (b) NSL-KDD, and (c) KDDCUP’99 datasets.
Fig 12.
ROC-AUC of the four models (a) AlexNet (b) GoogleNet (c)Resnet-50 (d) Proposed Model.
Fig 13.
Confusion matrices obtained by applying the proposed method to the NSL-KDD dataset, for five types of intrusion.
Table 7.
The comparison of results from the proposed architecture and state-of-the-art algorithms on the UNSW-B15, NSL-KDD and KDDCUP’99 dataset.