Friendsourcing to detect network manipulation

Censorship;Network-Neturality;Network Manipulation;Traffic Differentiation;Friendsourcing;Social Networks

Shankesi, Ravinder

关键词: Censorship;    Network-Neturality;    Network Manipulation;    Traffic Differentiation;    Friendsourcing;    Social Networks;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/45321/Ravinder_Shankesi.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Traditionally, security on the Internet has been concerned with threats posed by edge nodes (hosts) to other hosts and to the middle nodes of the network (routers).An example of the former is the spread of viruses and an example of the latter is congestion caused by packet floods. However, in recent years, we have seen a rising number of instances in which hosts seek to defend themselves against the middle of the network. This has been particularly evident with Internet censorship, where countries seek to suppress political unrest by closing off access to social networks, particular web sites, or even the Internet itself. Another areaof controversy and concern arises with violations of network neutrality, in which an Internet Service Provider (ISP) prejudices in favor of some types of packetscompared to others, typically for some commercial competitive advantage or remuneration.Fresh types of abuse are emerging as well, such as ISPs who alterclick traffic or advertising. All of these fall into a general category that one can describe as network manipulation.Many methods have been investigated to detect and circumvent network manipulation. For instance, researchers have devoted projects to determining which words trigger various national censorship firewalls and strategies for deploying proxies to enable tunnels that prevent such firewalls from recognizing that hosts are accessing forbidden sites. Other research has deployed nodes that can be used for testing network neutrality violations. One of the key challenges that faces many of these techniques is the requirement to get “help on the inside”, that is, to get cooperation from nodes that are subject to suspected network manipulation.This thesis proposes to explore this specific problem with the aim of addressing it through friendsourcing. Friendsourcing is a kind of crowdsourcing in which individuals use their social networks to gain help from their friends. For instance, if an individual is having trouble with a network link he may ask his friend if she is also having trouble with that link. The results could point to problems in various places, such as a server outage, a problem with a client, an accidental networkoutage, or deliberate network manipulation.Our thesis is that, social networks are an effective and efficient way for a user to acquire a sufficiently distributed sensor network required for detecting the source of censorship on the underlying communication network.Proving that friendsourcing is a viable strategy to detect network manipulation required three primary features. First, we needed to show that is possible to carry out recruitment that finds a collection of hosts in the right network locations; we show this through coverage and redundancy of multiple datasets. Second, we need to prove that friendsourcing is an efficient strategy to detect network manipulation;we do this through developing and validating optimizations required for recruitment. Third, we must show that it is a practical; we show this through areal-life field study outside a laboratory setting.Our contributions in this thesis are:1. We were the first to show the effectiveness of friendsourcing to detect network manipulation.2. We developed optimization algorithms that can makefriendsourcing viable in practice.3. We showed the practicality of friendsourcing, by deploying it in a real-life experiment which found a number of web-based manipulations in India.