Abstract Internal site search is an integral part of how users navigate modern sites, from restaurant reservations to house hunting to searching for medical solutions. Search terms on these sites may contain sensitive information such as location, medical information, or sexual preferences; when further coupled with a user’s IP address or a browser’s user agent string, this information can become very specific, and in some cases possibly identifying. In this paper, we measure the various ways by which search terms are sent to third parties when a user submits a search query. We developed a methodology for identifying and interacting with search components, which we implemented on top of an instrumented headless browser. We used this crawler to visit the Tranco top one million websites and analyzed search term leakage across three vectors: URL query parameters, payloads, and theRefererHTTP header. Our crawler found that 512,701 of the top 1 million sites had internal site search. We found that 81.3% of websites containing internal site search sent (orleakedfrom a user’s perspective) our search terms to third parties in some form. We then compared our results to the expected results based on a natural language analysis of the privacy policies of those leaking websites (where available) and found that about 87% of those privacy policies do not mention search terms explicitly. However, about 75% of these privacy policies seem to mention the sharing of some information with third-parties in a generic manner. We then present a few countermeasures, including a browser extension to warn users about imminent search term leakage to third parties. We conclude this paper by making recommendations on clarifying the privacy implications of internal site search to end users.

Users struggle to adhere to expert-recommended security and privacy practices. While prior work has studied initial adoption of such practices, little is known about the subsequent implementation and abandonment. We conducted an online survey (n=902) examining the adoption and abandonment of 30 commonly recommended practices. Security practices were more widely adopted than privacy and identity theft protection practices. Manual and fully automatic practices were more widely adopted than practices requiring recurring user interaction. Participants' gender, education, technical background, and prior negative experience are correlated with their levels of adoption. Furthermore, practices were abandoned when they were perceived as low-value, inconvenient, or when users overrode them with subjective judgment. We discuss how security, privacy, and identity theft protection recommendations and tools can be better aligned with user needs.

We present a new treebank of English and French technical forum content which has been annotated for grammatical errors and phrase structure.This double annotation allows us to empirically measure the effect of errors on parsing performance.While it is slightly easier to parse the corrected versions of the forum sentences, the errors are not the main factor in making this kind of text hard to parse.