PDF Publication Title:
Text from PDF Page: 063
6. METHODOLOGY In this section it is explained the approach taken for the different papers. All papers have similar points of interest: objectives, datasets, models and results. In the following sections these common points of interest are analysed in detail for the three papers. In Fig. 17 is provided the workflow connecting all points of interest for the papers. It is depicted in a model diagram, with arrows defining the relationship between the points. We can see in Fig. 17 that the objectives established at the start of the research define the datasets, and likewise the datasets finally available can constrain the achievable objectives. Once the objectives and datasets are established, both are crucial to select the possible applicable models. The different models that are finally chosen may have specific requirements in the format of the dataset used. Different combinations of models and datasets will normally generate different results that will ultimately confirm (or not) the objectives set at the beginning. The influence of the objectives to define the datasets is clear in all papers since all of them make use of a dataset specifically suited for the mission. The constraint imposed by the datasets on the objectives is also clear, for example in [3] where we had to limit the intrusions that could be detected to the ones provided by the NSL-KDD dataset. In all cases, we have tried models appropriate to the objectives and available datasets. For example, in [1] where the objective is to predict the future activity of IoT devices with a dataset of time-series of past activity, we applied time-series algorithms and re-formatted the dataset to allow the use of other algorithms initially not suitable for time-series data. In [2], we have made possible an innovative application of deep learning algorithms to the detection of the type-of-service by formatting the data appropriately and using the specific models more suited to the detection problem (CNNs for representation learning and LSTMs for time-series information processing). In [3], since intrusion detection is a stochastic problem, we wanted to try models specifically suited to learn the probability distribution of intrusions and their co- occurrence with the available predictors (features), which is the reason to try a generative model based on a conditional VAE to an intrusion detection problem. In [4], where the goal is to perform QoE estimation using information extracted from network flow packets, in order to generate the data that the QoE prediction models will use, it was necessary to establish an experimental setup that would allow identifying the QoE of the video transmissions while recording the associated network flow packets. The resulting data are multivariate time series, in time-steps of 1-second, which contain the network packets transmitted in each time-step plus the presence or absence of seven video transmission errors in that time-step. Finally, in [5] where the objective is to generate synthetic data that could be used as new training data for an intrusion detection classifier, we selected the NSL-KDD dataset which has a large number of continuous features and three categorical features with high cardinality, this imposes a higher difficulty to the problem of data generation what makes this dataset appropriate for the research work. In this case we had to perform a scaling of the continuous features and a one- hot encoding of the discrete features which was needed to apply a VAE to this problem. Doctoral Thesis: Novel applications of Machine Learning to NTAP - 61PDF 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 | RSS | AMP |