What the research discovered
Researchers analysed refined fluctuations in GPS knowledge — together with sign power, noise and frequency shifts — to know a person's setting and behavior. This goes past typical navigation makes use of and faucets into steady GPS alerts that smartphones already obtain within the background.
The research was carried out by MTech pupil Soham Nag and Prof Smruti R Sarangi from IIT Delhi's Centre of Excellence in Cyber Methods and Info Assurance and the pc science and engineering division.
Their system, known as AndroCon, makes use of 9 low-level GPS parameters like Doppler shift, multipath interference and energy variations to detect person exercise and context. It could establish whether or not an individual is sitting, standing, mendacity down, strolling, travelling in a metro or on a flight, or staying in a crowded location. It could additionally inform if a room is empty or occupied.
How the system works
The group mixed conventional sign processing with machine studying to transform noisy GPS inputs into detailed insights. The framework can map indoor areas equivalent to rooms and staircases with an error of lower than 4 metres, utilizing solely GPS patterns and motion paths.
“Throughout a year-long research overlaying 40,000 sq km and a number of other smartphone fashions, AndroCon achieved as much as 99% accuracy in figuring out environment and greater than 87% accuracy in recognising human actions — even refined gestures like hand actions close to the cellphone,” mentioned Prof Smruti R Sarangi.
Privateness danger highlighted
Whereas the system may help create context-aware functions with out utilizing intrusive sensors, the researchers warning that it additionally exposes a privateness hole. Any Android app with exact location entry might probably extract contextual knowledge with out person consent.”This research reveals an unseen facet of GPS — a silent but highly effective sensing channel,” Prof Sarangi added. “AndroCon turns an odd smartphone into an unexpectedly exact scientific instrument, and a reminder that even acquainted applied sciences can maintain hidden dangers when misused.”
The research has been printed in ACM Transactions on Sensor Networks.