This paper presents a provably secure method for embedding multiple watermarks in sequential designs. A number of different watermarks signed with the IP owner's secret key from a public key cryptography system are generated. The owner's watermarks are then dissembled into the states and transitions of the original sequential design. Hiding the multiple watermarks in the states and transitions is shown to be an instance of obfuscating a multi-point function with a generalized output. We draw on the theoretical cryptographic results of provable obfuscation of this function family to build a secure sequential multi-watermark system by construction. An iterative synthesis method for integrating the collection of watermarks to the original design is introduced. Analysis of watermark properties and the attack resiliency of the new multiple watermarking construction is presented. Experimental evaluations on benchmark circuits demonstrate practicality and low overhead of the new provably secure multiple watermarks construction method.
展开▼
