Website-Targeted False Content Injection by Network Operators
Gabi Nakibly1,3, Jaime Schcolnik2, and Yossi Rubin3

1 Computer Science Department, Technion, Haifa, Israel
2Computer Science Department, Interdisciplinary Center, Herzliya, Israel
3Rafael – Advanced Defense Systems, Haifa, Israel


Over the last few years there have been numerous reports of ISPs that alter or proxy their customers’ traffic, including, for example, CMA Communications in 2013 [6], Comcast in 2012 [16], Mediacom in 2011 [9], WOW! in 2008 [27], and Rogers in 2007 [32]. Moreover, several extensive studies have brought the details of this practice to light [17, 30, 25, 35]. The main motivations of ISPs to alter traffic are to facilitate caching, inject advertisements into DNS and HTTP error messages, and compress or transcode content.

 All of these reports and studies found that these traffic alterations were carried out exclusively by edge ISPs,namely, retail ISPs that sell Internet access directly to end customers, and are their “first hop” to the Internet. This finding stems from the server-centric approach the above studies have taken. In this approach, one or a handful of servers are deployed to deliver specific content to users, after which a large number of clients are solicited to fetch that content from the servers. Finally, an agent on the clients – usually a JavaScript delivered by the server itself – looks for deviations between the content delivered by the server and that displayed to the user. Figure 1(a)illustrates the traffic monitored in this server-centric approach.

Such an approach can be used to inspect the traffic of many clients from diverse geographies who are served by different edge ISPs. The main disadvantage of this approach is that the content fetched by the clients is very specific. All clients fetch the same content from the same web servers. This allows only the detection of network entities that aim to modify all of the Internet traffic1 of a predetermined set of users and are generally oblivious to the actual content delivered to the user. Such entities indeed tend to be edge ISPs that target only the traffic of their customers.

In this work we show that the above approach misses a substantial portion of the on-path entities that modify traffic on the Internet. Using extensive observations over a period of several weeks, we analyzed petabits of Internet traffic carrying varied content delivered by servers having over 1.5 million distinct IP addresses. We newly reveal several network operators that modify traffic not limited to a specific set of users. Such network operators alter Internet traffic on the basis of its content, primarily by the website a user visits. The traffic of every Internet user that traverses these network operators is susceptible to alteration."

-- 
RR
"Through counter-intelligence it should be possible to pinpoint potential trouble-makers ... And neutralize them, neutralize them, neutralize them"