Abstract
By tracking changes in electromagnetic radiation footprint emitted from computers using a magnetometer on commodity mobile devices, a malicious attacker can easily learn the secrets of the computer's owner without physically peeping at or hacking into the victim's system. Targeting at Applications and Web browsers, we present MagAttack, which uses the built-in magnetometer on commodity mobile phones to infer which App is running or which webpage the user is browsing on a nearby computer, as well as finergrained information about victim's interests, habits, etc. Our preliminary results show that MagAttack is independent of the earth's magnetic filed, model of phones, and magnetometer sampling rates. We also conducted an in-The-wild evaluation where an instrumented participant uses her laptop as usual and MagAttack can detect when she opens 10 different popular Apps. MagAttack achieves a classification accuracy of up to 98%.