Loading...
Please wait, while we are loading the content...
Similar Documents
Deep Learning in Intrusion Detection System: An Overview
| Content Provider | Semantic Scholar |
|---|---|
| Author | Aminanto, Erza Kim, Kwangjo |
| Copyright Year | 2016 |
| Abstract | Identifying unknown attacks is one of the big challenges in network Intrusion Detection Systems (IDSs) research. In the past decades, researchers adopted various machine learning approaches to classify and distinguish anomaly traffic from benign traffic without prior knowledge on the attack signature. Extensive academic research on machine learning made a significant breakthrough in mimicking human brain recently. The state-of-the-art on machine learning breakthrough comes from deep learning which has been predicted to cause a powerful improvement in artificial intelligence field. Numerous complex applications have been accomplished by deep learning. One of the distinguished applications is AlphaGo from Google that uses Convolutional Neural Network. AlphaGo beat the Korean world champion in the “Go” game recently by showing superman-like capabilities in remote machine learning. The advancements on this learning algorithms may improve IDS ability to reach high detection rate and low false alarm rate. However, the deep learning implementations in intrusion detection applications may have some limitations. In this paper, we survey previous IDSs that embrace deep learning approaches. Deep learning methods such as deep belief network, restricted Boltzman machine, deep Boltzman machine, deep neural network, auto encoder, etc., are commonly used in IDSs. We examine such deep learning methods with their advantages and disadvantages in order to get better understanding on how to apply deep learning. We realize that there is a confusion of how to adopt deep learning in IDS application properly. Our claim is that deep learning is useful in IDS, especially for feature extraction. In order to support our claim, we provide future challenges and directions to employ deep learning in IDS accordingly. Finally, deep learning methods can enhance future research on unknown attack detection. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://caislab.kaist.ac.kr/publication/paper_files/2016/IRCET16_AM.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
