Andrey Finkelshtein, Rami Puzis, Asaf Shabtai, Bronislav Sidik
SBP-BRiMS (2016)
The meteoric progress of Internet technologies and PDA (personal digital assistant) devices has made public Wi-Fi hotspots very popular. Nowadays, hotspots can be found almost anywhere: organizations, home networks, public transport systems, restaurants, etc. The Internet usage patterns (e.g. browsing) differ with the hotspot venue. This insight introduces new traffic profiling opportunities. Using machine learning techniques we show that it is possible to infer types of venues that provide Wi-Fi access (e.g., organizations and hangout places) by analyzing the Internet traffic of connected mobile phones. We show that it is possible to infer the user’s current venue type disclosing his/her current context. This information can be used for improving personalized and context aware services such as web search engines or online shops, without the presence on user’s device. In this paper we evaluate venue type inference based on mobile phone traffic collected from 115 college students and analyze their Internet behavior across the different venues types.