FP-Crawlers: Studying the Resilience of Browser Fingerprinting to Block Crawlers

Abstract

Data available on the Web, such as financial data or public reviews, provides a competitive advantage to companies able to exploit them. Web crawlers, a category of bot, aim at automating the collection of publicly available Web data. While some crawlers collect data with the agreement of the websites being crawled, most crawlers do not respect the terms of service. CAPTCHAs and approaches based on analyzing series of HTTP requests classify users as humans or bots. However, these approaches require either user interaction or a significant volume of data before they can classify the traffic. In this paper, we study browser fingerprinting as a crawler detection mechanism. We crawled the Alexa top 10K and identified 291 websites that block crawlers. We show that fingerprinting is used by 93 (31.96%) of them and we report on the crawler detection techniques implemented by the major fingerprinters. Finally, we evaluate the resilience of fingerprinting against crawlers trying to conceal themselves. We show that although fingerprinting is good at detecting crawlers, it can be bypassed with little effort by an adversary with knowledge on the fingerprints collected.