Characterizing Composite User-Device Touchscreen Physical Unclonable Functions (PUFs) for Mobile Device Authentication
Ryan A. Scheel, Akhilesh Tyagi · 2015
Mobile systems have unique authentication requirements. A composite user-device identity that is computationally difficult to decompose into its user and device contribution is better suited for mobile context authentication for services such as Google wallet. We base such a composite identity in a composition of human user biometric and device silicon biometric realized as a user-device (UD-) physical unclonable function (PUF). This UD-PUF is derived from the touch screen of a mobile device. Challenge is a shape drawn on the screen, which the human user traces. The pressure values generated in the resulting touch events reflect the device level variability of the underlying transistor array. These pressure sequences can be quantized into an appropriate response. We characterize such a composite PUF for both its variability and reproducibility. We illustrate 0 bits of error in reproducibility for the (same device, same user, same challenge) scenario with the use of an innovative statistical concentrator serving the role of ECC (error correcting codes) in traditional PUFs. For the (same device, same user, different challenge), (same device, different user, same challenge), (different device, same user, same challenge), we benefit from as large a variability in the response as possible. We show 60+ bits Hamming distance in the composite UD-PUF responses of length 128 bits when variability is expected. We also demonstrate the promise of these PUFs to serve as biometric hardware pseudorandom number generators (PRGs) by putting them through Montreal TESTU01 suite of tests. Our best PUFs pass all the tests except occasionally failing 3. This PUF was implemented on Nexus 7 devices running Android.