Image authentication and tamperproofing for noisy channels

We present a new method for authenticating and tamperproofing images. It is based on the recently developed NT-MAC (noise tolerant message authentication code). The NTMAC is designed for multimedia transmitted over noisy channels. Unlike ordinary cryptographic codes, the NT-MAC fails gradually in the presence of noise or bit errors. In this work, we show how a combination of structural and random partitions allows the NTMAC to identify regions which might represent forgeries while distinguishing between forgeries and minor errors caused by communications noise. The paper concludes with a simple example of a forged image and the NTMAC's estimate of the changes.