Binary Abstractions » Research http://www.binaryabstractions.com Research and Other Diatribes Thu, 04 Nov 2010 22:25:28 +0000 en hourly 1 http://wordpress.org/?v= sFlow Content Analysis http://www.binaryabstractions.com/2009/05/05/sflow-content-analysis/ http://www.binaryabstractions.com/2009/05/05/sflow-content-analysis/#comments Tue, 05 May 2009 19:26:35 +0000 Justin http://www.binaryabstractions.org/?p=61 I recently began exploring a new research project to determine the effectiveness of various classification schemes on sFlow v5 data.  Previous efforts using a Naive Bayesian Classifier to identify unique patters within 802.11 wireless header information, which is similarly incomplete, has shown promise.  The objective here is to compare an NBC to other classification techniques such as an Artificial Neural Network, specifically a Multi-Layer Perceptron.

The challenge is that both of these classification schemes require some degree of learning supervision to effectively classify this type of data.  The MLP NN uses the backpropagation algorithm for supervised training which generates a scaling factor to determine the error in each output node and influences nodes in previous or hidden layers accordingly.

In my supposed spare time, I plan on using the Matlab Neural Network Toolbox to implement both classifiers and compare their effective abilities to properly identify unique events of interest within a sample set of sFlow data.

Analysis of the Wireless Covert Channel Attack:
Carrier Frequency Selection: http://www.iu.hio.no/nik07/bidrag/Dyrkolbotn.pdf

A Dynamic Trust Model Based on Naive Bayes
Classifier for Ubiquitous Environments: http://uclab.khu.ac.kr/resources/publication/J_56.pdf

]]> http://www.binaryabstractions.com/2009/05/05/sflow-content-analysis/feed/ 0