EQUIANGULAR TIGHT FRAME FINGERPRINTING CODES

Watermarking and Multimedia Security

Presented by: Dustin Mixon, Author(s): Dustin Mixon, Princeton University, United States; Christopher Quinn, Negar Kiyavash, University of Illinois Urbana-Champaign, United States; Matthew Fickus, Air Force Institute of Technology, United States

We show that equiangular tight frames (ETFs) are particularly well suited as additive fingerprint designs against Gaussian averaging collusion attacks when the number of users is less than the square of the signal dimension. The detector performs a binary hypothesis test in order to decide whether a user of interest is among the colluders. Given a maximum coalition size, we show that the geometric figure of merit of distance between the corresponding "guilty" and "not guilty" linear forgeries for each user is bounded away from zero. Moreover, we show that for a normalized correlation detector, reliable detection is guaranteed provided that the number of users is less than the square of the signal dimension. Moreover, we show that the coalition has the best chance of evading detection when it uses equal weights.

You need the Flash Player.

Share:

Download subtitles | Enlarge video

1. slide