Homomorphic encryption

Homomorphic encryption lets you evaluate functions on encrypted data without the secret key. Decrypting the result gives you the function applied to the original plaintext. NIST treats fully homomorphic encryption (FHE) as a privacy-enhancing cryptography tool: compute arbitrary functions over ciphertext, non-interactively, without knowing the key. Partial schemes only allow some operations, such as addition or multiplication. Fully homomorphic means any computable function.

The use is outsourced computation. A cloud can score a model, total a payroll, or join two datasets while the values stay encrypted. The operator sees ciphertext in and ciphertext out. That is not the same as encrypting a disk with AES. AES at rest stops a stolen drive. Homomorphic encryption is for when the machine doing the work is not allowed to read the work.

It is not a drop-in. Ciphertext is larger, operations are much slower than cleartext, and FHE is malleable by design: anyone can transform a ciphertext into a valid ciphertext for f(m). That is the feature. It is also why you do not treat FHE like ordinary encryption when you needed non-malleability. Parameters have to be chosen for a security level. This is a specialist control, still being standardised, not the default for a web app.