AONT

An all-or-nothing transform (AONT) is a reversible data transformation that converts input data into a collection of blocks such that no proper subset of blocks reveals information about the input; only the full set can be used to recover the original data. AONTs are not encryption by themselves but are designed to be used together with encryption to strengthen security for stored data.

In typical AONT constructions, a random seed and a cryptographic primitive such as a hash or a pseudorandom function are used to bind the output blocks together. The transform ensures that each block depends on the entire input and on the random seed, so knowledge of any subset of blocks leaves the plaintext indistinguishable from random. Recovery of the plaintext requires accessing all blocks and applying the inverse transformation.

A primary use of AONTs is secure data outsourcing and storage on untrusted servers. The usual workflow is to apply an AONT to the data, then encrypt the AONT output with a conventional cipher. An attacker who obtains only a portion of the ciphertext cannot learn the plaintext. Only after decrypting the entire AONT output can the original data be recovered. AONTs also help mitigate offline dictionary attacks on passwords or keys when paired with password-derived keys.

Limitations include added computational and storage overhead and the fact that AONTs do not automatically provide data integrity. For strong security, they are often combined with authenticated encryption or separate integrity checks. History traces the concept to Ronald L. Rivest in the 1990s, with ongoing research to improve efficiency and formal security properties.