Anonymity Properties of Stored or Transmitted Data Taken from Bluetooth Scans

Modern consumer wireless devices are increasingly powerful, making them attractive to use as wireless sensor nodes. At the same time, many use protocol suites such as Bluetooth which require devices to reveal data that may make for unique device identifiers. This paper explores this quantitatively through scans covering several thousand devices indifferent urban locations. In measuring the anonymity afforded by the elements of a Bluetooth device profile, we find that attributes such as the device class are poor for linking sightings of the same device; the device name can provide a surprising amount of anonymity but when it does not it can be a very effective key to link devices with individuals; and frequently users exhibit privacy-adverse behavior, such as placing telephone numbers in device names or using nicknames that are statistically rare.