PDF Publication Title:
Text from PDF Page: 166
The paper is organized as follows: Section 2 presents related works. Section 3 describes the work performed. Section 4 describes the results obtained and finally, Section 5 provides discussion and conclusions. 2. Related works As far as we know, there is no previous application of a variational generative model based on neural networks to generate data of similar probabilistic structure to intrusion detection data. Therefore, the work presented in this paper is original in essence. There are a number of works for the application of variational generative models to generate images [1, 2, 17] and text [3, 4], but there is none in the area of intrusion detection. There is no work, similar to the present one, generating both continuous and categorical features. In [18] is presented a model that handles a discrete distribution for the “latent” layer, but it is applied to images with continuous features. In [19] the authors provide a solution using a VAE in the intrusion detection field, but it is used exclusively to implement a classifier, not to generate synthetic data according with the intrusion class as in the present work There is a vast number of works applying classification algorithms to NIDS [20, 21]. This paper is not related specifically with any classification technique, but we will show (Section 4.2) that the synthetic data generated with our model improves results obtained with different classifiers. Therefore, it is interesting to provide a summary of results on classification using the NSL-KDD dataset, which will help to put into perspective the results presented in this work. It is important to mention that comparison of results in this field is difficult due to: (1) diversity of reported performance metrics; (2) the aggregation of classification labels in different sets (e.g. 23 labels can be grouped hierarchically in different subsets or categories: 23, 5 or 2 final labels) making it difficult to compare results for different subsets; (3) reporting results on unclear test datasets. This last point is important to mention, because for example, for the NSL-KDD dataset, 16.6% of samples in the test dataset correspond to labels not present at the training dataset. This is an important property of this dataset and creates an additional difficulty to the classifier. These difficulties are shown in detail in [16]. Classification results for NSL-KDD are provided in several works. In [22] is achieved an accuracy of 79.9% for test data, for the 5-labels intrusion scenario. In [23] they provide, for the 2-labels scenario a recall of 75.49% on test data. Authors in [20] explain the reasons to create the NSL-KDD data set, providing results for several algorithms, being 82.02% the best accuracy reported when using the full NSL_KDD dataset for training and testing, for the 2- labels scenario. There is large literature related to algorithms dealing with imbalanced datasets [6, 7, 8], and in particular presenting algorithms for synthetic over-sampling [9, 10, 11], adaptive over- sampling [14], over-sampling followed by under-sampling [12], ensemble sampling [15] and specific combinations of methods [13]. 3. Work description In this section we present the dataset used for this work, a description of the variational method employed and details on different variants of the method. 3.1. Selected dataset We have chosen the NSL-KDD [16] dataset as our reference dataset. NSL-KDD is an Doctoral Thesis: Novel applications of Machine Learning to NTAP - 164PDF Image | Novel applications of Machine Learning to Network Traffic Analysis
PDF Search Title:
Novel applications of Machine Learning to Network Traffic AnalysisOriginal File Name Searched:
456453_1175348.pdfDIY PDF Search: Google It | Yahoo | Bing
Cruise Ship Reviews | Luxury Resort | Jet | Yacht | and Travel Tech More Info
Cruising Review Topics and Articles More Info
Software based on Filemaker for the travel industry More Info
The Burgenstock Resort: Reviews on CruisingReview website... More Info
Resort Reviews: World Class resorts... More Info
The Riffelalp Resort: Reviews on CruisingReview website... More Info
CONTACT TEL: 608-238-6001 Email: greg@cruisingreview.com (Standard Web Page)