Fighting authorship linkability with crowdsourcing

摘要: Massive amounts of contributed content -- including traditional literature, blogs, music, videos, reviews and tweets are available on the Internet today, with authors numbering in many millions. Textual information, such as product or service reviews, is an important increasingly popular type that being used a foundation trendy community-based reviewing sites, TripAdvisor Yelp. Some recent results have shown that, due partly to their specialized/topical nature, sets authored by same person readily linkable based simple stylometric features. In practice, this means individuals who author more than few under different accounts (whether within one site across multiple sites) can be linked, which represents significant loss privacy.In paper, we start showing problem actually worse previously believed. We then explore ways mitigate authorship linkability reviewing. first attempt harness global power crowdsourcing engaging random strangers into process re-writing reviews. As our empirical (obtained from Amazon Mechanical Turk) clearly demonstrate, yields impressively sensible reflect sufficiently characteristics prior techniques become largely ineffective. also consider using machine translation automatically re-write Contrary what was believed, show decreases number intermediate languages grows. Finally, combination report results.