The authors present methods to cryptographically sign finite and infinite length digital streams while reducing computational requirements of previous methods. The authors divide the stream into blocks and embed the signing information in the stream itself such that signing the information in the ith block can be used to authenticate the (i + 1) block because the signature on the first block will propagate through succeeding blocks (that is, the first block contains a hash of the second block, the second block contains a hash of the third block, and so on). Consequently, verification of succeeding blocks requires only hash computations instead of full signature computations.

The authors describe previous methods for implementing stream-based signatures including similar work and the variations they used to produce their results and a detailed security proof for off-line and on-line implementations. They discuss applications including MPEG video and audio, Java applets, broadcast, long file authentication, and content filtering at proxy servers.

The presentation is well done and appears to provide information useful to cryptographers, who will examine the author’s proof and determine whether the proposed method can be readily broken. If the methods described are secure, they appear to simplify the computational requirements for signing streaming information and could have significant benefits for several important applications.