Network Intrusion Detection using Deep Learning by Kwangjo Kim & Muhamad Erza Aminanto & Harry Chandra Tanuwidjaja

Network Intrusion Detection using Deep Learning by Kwangjo Kim & Muhamad Erza Aminanto & Harry Chandra Tanuwidjaja

Author:Kwangjo Kim & Muhamad Erza Aminanto & Harry Chandra Tanuwidjaja
Language: eng
Format: epub
ISBN: 9789811314445
Publisher: Springer Singapore


5.1 Generative

This sub-chapter groups IDSs that use deep learning for feature extraction only and use shallow methods for the classification task.

5.1.1 Deep Neural Network

Roy et al. [1] proposed an IDS by leveraging deep learning models and validated that a deep learning approach can improve IDS performance. DNN is selected comprising of multilayer feedforward NN with 400 hidden layers. Shallow models, rectifier and softmax activation functions, are used in the output layer. Two advantages of feedforward neural network are to provide a precise approximation for complex multivariate nonlinear function directly from input values and to give the robust modeling for large classes. Besides that, the authors claimed that DNN is better than DBN since the discriminating power by characterizing the posterior distributions of classes is suitable for the pattern classification [1].

For validation, KDD Cup’99 dataset was used. This dataset has 41 features that was given as the input to the network. The authors divided all the training data into 75% for training and 25% for validation. They also compared the performance of a shallow classifier, SVM. Based on their experimental result, DNN outperforms SVM by the accuracy of 99.994%, while SVM achieved 84.635% only. This result showed the effectiveness of DNN for IDS purposes.



Download



Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.