Hardware Fingerprinting Using HTML5
This addresses the issue of unreliable device identification for web security and tracking, though it is incremental as it builds on existing fingerprinting concepts with new hardware focus.
The paper tackles the problem of device fingerprinting by proposing hardware-based features using HTML5, which are harder to mask than software-based methods, and presents an initial experiment to fingerprint a device's GPU.
Device fingerprinting over the web has received much attention both by the research community and the commercial market a like. Almost all the fingerprinting features proposed to date depend on software run on the device. All of these features can be changed by the user, thereby thwarting the device's fingerprint. In this position paper we argue that the recent emergence of the HTML5 standard gives rise to a new class of fingerprinting features that are based on the \emph{hardware} of the device. Such features are much harder to mask or change thus provide a higher degree of confidence in the fingerprint. We propose several possible fingerprint methods that allow a HTML5 web application to identify a device's hardware. We also present an initial experiment to fingerprint a device's GPU.