Network Bias Indication Trainer
Network Neutrality has been a hot topic since the proliferation of the internet. Indeed, there have been numerous efforts by the research community to expose the Quality of Service (QoS) policies that could lead to violations of net neutrality. This paper is building upon their success and is intended to employ new methods that can assist in detecting such violations. There are many methods that an Internet Service Provider can implement to violate Net Neutrality. For the most part, we will focus on Strict Priority Queueing (SPQ). SPQ leaves a unique and interesting pattern in our detection packet trains. Our goal is to train a Machine Learning classifier that can identify whether a packet train has gone through a network that violates Net Neutrality using Strict Priority Queueing. In this paper, we will employ statistical models and Machine Learning techniques to identify the areas where ISPs are violating Network Neutrality. Our goal is to show that with Machine Learning the detection of network neutrality will require smaller and less detectable packets. Our hope is that researchers will employ more Machine Learning related techniques to identify Network Neutrality violations. One of the main classifiers in Machine Learning is Support Vector Machines. We have decided to implement this thesis using an SVM identifier. SVMs are great at identifying division lines in binary data sets. Therefore, an SVM classifier can detect whether a certain condition exists or not. This is useful for our cause because our goal is to identify whether a packet train has gone through a network that discriminates using SPQ or not. Additionally, we will use Random Forest to train a second set of classifiers and compare their results. My goal of writing this paper and doing a research in this field was not to make a political stand but rather to create transparency and provide more information to internet users.