Security of an ill-posed operator for image authentication

This letter analyzes the security of an image authentication scheme which exploits the instability of an ill-posed operator. Since the ill-posed operator produces only a limited number of authentic images regardless of the number of watermarks, an attacker can impersonate an image owner to generate authentic images at a high probability. Our experiments demonstrate that our attack is both practical and effective.