A Nearly Four-Year Longitudinal Study of Search-Engine Poisoning

We investigate the evolution of search-engine poisoning using data on over 5 million search results collected over nearly 4 years. We build on prior work investigating search-redirection attacks, where criminals compromise high-ranking websites and direct search traffic to the websites of paying customers, such as unlicensed pharmacies who lack access to traditional search-based advertisements. We overcome several obstacles to longitudinal studies by amalgamating different resources and adapting our measurement infrastructure to changes brought by adaptations by both legitimate operators and attackers. Our goal is to empirically characterize how strategies for carrying out and combating search poisoning have evolved over a relatively long time period. We investigate how the composition of search results themselves has changed. For instance, we find that search-redirection attacks have steadily grown to take over a larger share of results (rising from around 30% in late 2010 to a peak of nearly 60% in late 2012), despite efforts by search engines and browsers to combat their effectiveness. We also study the efforts of hosts to remedy search-redirection attacks. We find that the median time to clean up source infections has fallen from around 30 days in 2010 to around 15 days by late 2013, yet the number of distinct infections has increased considerably over the same period. Finally, we show that the concentration of traffic to the most successful brokers has persisted over time. Further, these brokers have been mostly hosted on a few autonomous systems, which indicates a possible intervention strategy.