Integral attack
An attack that exploits sets of inputs whose intermediate sums cancel to zero in a cipher with reduced rounds.
An integral attack feeds the cipher a set of plaintexts that differ in only one position. After a few rounds, the sum of the intermediate values over the whole set is zero, and that balance leaks information about the key. With few rounds, key bits can be recovered without guessing the whole key.
Daemen, Knudsen and Rijmen introduced the idea in 1997 against the Square cipher, calling it the square attack. Knudsen and David Wagner gave it the name integral cryptanalysis in 2002.