In cryptography, a known-key distinguishing attack is an attack model against symmetric ciphers, whereby an attacker who knows the key can find a structural property in cipher, where the transformation from plaintext to ciphertext is not random. There is no common formal definition for what such a transformation may be. The chosen-key distinguishing attack is strongly related, where the attacker can choose a key to introduce such transformations.^[1]

These attacks do not directly compromise the confidentiality of ciphers, because in a classical scenario, the key is unknown to the attacker. Known-/chosen-key distinguishing attacks apply in the "open key model" instead.^[1] They are known to be applicable in some situations where block ciphers are converted to hash functions, leading to practical collision attacks against the hash.^[2]

Known-key distinguishing attacks were first introduced in 2007 by Lars Knudsen and Vincent Rijmen^[1] in a paper that proposed such an attack against 7 out of 10 rounds of the AES cipher and another attack against a generalized Feistel cipher. Their attack finds plaintext/ciphertext pairs for a cipher with a known key, where the input and output have s least significant bits set to zero, in less than 2^s time (where s is fewer than half the block size).^[3]

These attacks have also been applied to reduced-round Threefish (Skein)^[4][5] and Phelix.^[6]