What is format-preserving encryption (FPE)?
Format-preserving encryption (FPE) is a type of encryption whose output keeps the same format and length as the input. A 16-digit card number encrypts to another 16-digit number, and a nine-digit Social Security number stays nine digits. Anyone with the key can decrypt the value, so the key must be tightly protected.
For example, a bank wants to give its analytics team a copy of a customer table. The card number column in the old database only accepts 16 digits. Standard encryption would produce a long string of mixed characters that breaks the schema and the reports built on it. FPE produces a different 16-digit number that fits the column, so the copy works without code changes.
How format-preserving encryption works
FPE uses a standard cipher, usually AES, in a special mode that maps each input to an output from the same set of allowed characters. NIST Special Publication 800-38G describes approved FPE methods, including the FF1 mode. Each encryption uses a secret key and can also use a “tweak,” an extra input that makes the same value encrypt differently in different contexts.
- The output has the same length and character set as the input.
- With the same key and tweak, the same input always gives the same output, so joins still work.
- Very short inputs give weak protection, so standards set a minimum input size.
Format-preserving encryption vs tokenization
| Format-preserving encryption | Tokenization | |
|---|---|---|
| How the value is protected | Math with a secret key | Replaced with a token; the real value is kept in a vault |
| Who can reverse it | Anyone with the key | Callers the vault authorizes |
| If the key or vault is exposed | Every value encrypted with that key is exposed | Values in the vault are exposed; tokens alone reveal nothing |
| What you run | Key management | A token vault and its access policy |
Both keep data usable in existing systems. The choice usually comes down to key management, scale and how access to real values should be controlled. This post compares format-preserving encryption and tokenization in more depth.
Why format-preserving encryption matters for AI systems
FPE solves a schema problem, but it can create a reading problem for AI. An encrypted card number still looks like a card number, and an encrypted name may look like random letters. A model cannot tell that the value is protected or what kind of value it is. Typed tokens carry a label, so the model knows it is looking at a person or an email and can reason around it. FPE output also stays personal data under laws like the GDPR, because the key can reverse it.
How Protecto helps
Protecto Vault uses tokenization with context-preserving, format-preserving masking. Tokens can keep the shape of the original value and still carry an entity tag, for example <PHONE>8349201756</PHONE>, so schemas keep working and AI models know what each value is. Authorized users and tools receive real values only where policy allows.
Frequently asked questions
Is format-preserving encryption secure?
It can be when it uses an approved mode such as FF1 from NIST SP 800-38G, a strong key and inputs above the minimum size. Its security depends heavily on protecting the key.
Is format-preserving encryption the same as tokenization?
No. FPE computes the output from the value and a key, so anyone with the key can reverse it. Tokenization replaces the value with a token and keeps the original in a vault that controls access.
Which standard defines format-preserving encryption?
NIST Special Publication 800-38G describes approved methods for format-preserving encryption, including the FF1 mode built on AES.
Related terms
Data encryption · Tokenization · PCI data · Context-preserving masking