Integrity
Also known as: data integrity
Making sure information stays correct and complete, and is not changed by accident or without permission.
Draft - this entry has not been reviewed yet.
Formal
The property that information is accurate and complete, and that any change to it is made only by those allowed to, in a way that can be noticed. Hashing and digital signatures are common ways to show that nothing has been altered.
In plain English
Like a recipe card - if someone quietly changes “one teaspoon of salt” to “one cup”, everyone who follows it ruins the dish without knowing why.
In practice
An attacker inside a supplier's mailbox changes the account number on an invoice sent to a municipality, and a finance clerk pays a large sum to the wrong account.
Why it matters
Decisions, payments and even medical treatment rely on data being right; wrong data can do more harm than missing data because nobody questions it.
Technical deep dive
ISO/IEC 27000 defines integrity simply as the property of accuracy and completeness, while the US FISMA definition used in FIPS 199 speaks of guarding against improper modification or destruction and explicitly folds in authenticity and non-repudiation. In practice the term covers two different problems: detecting unauthorised or accidental change to data, and ensuring that authorised changes are themselves correct, which is a matter of business rules, validation and process rather than cryptography.
The cryptographic toolkit forms a ladder. A checksum or CRC detects random corruption but offers no protection against a deliberate attacker, who can simply recompute it. A cryptographic hash such as SHA-256 is collision-resistant, but a bare hash only helps if the reference value itself is delivered over a trusted channel. A message authentication code such as HMAC (RFC 2104) binds integrity to a shared secret key, so anyone holding the key can verify and also forge; it therefore gives integrity and data-origin authentication between two parties but not non-repudiation. A digital signature uses an asymmetric key pair, so only the private-key holder can sign and anyone can verify, which is what makes non-repudiation possible. Modern protocols use authenticated encryption (AEAD), for example AES-GCM or ChaCha20-Poly1305 in TLS 1.3, so that confidentiality and integrity are provided together and tampered ciphertext is rejected before decryption output is used.
At the model level, the Biba model (1977) is the integrity counterpart of Bell-LaPadula, forbidding subjects from reading down to less trustworthy data or writing up to more trustworthy data, and the Clark-Wilson model (1987) describes commercial integrity through well-formed transactions, constrained data items and separation of duties. Databases enforce integrity through constraints, transactions and ACID properties; storage uses checksumming file systems and Merkle trees; backups use immutable or WORM storage so ransomware cannot rewrite them; software supply chains rely on code signing, reproducible builds and provenance frameworks such as SLSA.
Common failure modes include verifying a hash downloaded from the same compromised server as the file, signing data but never checking the signature, and logs that an administrator can edit, which undermines both forensic value and accountability. Business email compromise shows that integrity attacks often target process, not bits: a correctly transmitted invoice with a changed account number passes every technical check. GDPR Art. 5(1)(f) and Art. 32(1)(b) name integrity explicitly, and an unauthorised alteration of personal data counts as a personal data breach under Art. 4(12).
Relationships
- Part of
- CIA triad
- Unlocks
- Non-repudiation
- Implemented by
- Digital signatureHashingSecure bootTLS
- Don't confuse with
- Non-repudiationConfidentiality
Sources & further reading
Standards & official texts
- ISO/IEC 27000:2018 - Information security management systems - Overview and vocabulary · ISO/IEC
Course material
- Cyber Security Fast Track - Ordliste
Where this data comes from
This entry was drafted by an AI from the sources above and has not yet been checked by a person. Treat it as a starting point, and check anything important against the sources.
See the review queueSuggest a correction on GitHubThis term as JSON
Mentioned in
Check yourself
Loading…