Evasive bots masquerading as human beings on the web

摘要: Web bots such as crawlers are widely used to automate various online tasks over the Internet. In addition conventional approach of human interactive proofs CAPTCHAs, a more recent observational (HOP) has been developed automatically distinguish web from users. Its design rationale is that behave intrinsically differently beings, allowing them be detected. This paper escalates battle against by exploring limits current HOP-based bot detection systems. We develop an evasive system based on behavioral patterns. Then we prototype general framework and set flexible de-classifier plugins, primarily application-level event evasion. further abstract define benchmarks for measuring our system's evasion performance contemporary applications, including social network sites. Our results show proposed can effectively mimic behaviors evade detectors achieving high similarities between users bots.