A fail-stop signature scheme is a type of signature devised by van Heyst and Pederson [VP92] to protect against the possibility that an enemy may be able to forge a person's signature. It is a variation of the one-time signature scheme (see Question 7.7), in which only a single message can be signed and protected by a given key at a time. The scheme is based on the discrete logarithm problem. In particular, if an enemy can forge a signature, then the actual signer can prove that forgery has taken place by demonstrating the solution of a supposedly hard problem. Thus the forger's ability to solve that problem is transferred to the actual signer.
The term ``fail-stop'' refers to the fact that a signer can detect and stop failures, that is, forgeries. Note that if the enemy obtains an actual copy of the signer's private key, forgery cannot be detected. What the scheme detects are forgeries based on cryptanalysis.