Data integrity means records stay complete and unaltered except when an authorized person or process changes them on purpose.
It is distinct from secrecy, because a file can be fully readable and still be wrong, missing, or silently edited.
How it works
Integrity fails through hostile change, software defects, or ordinary mistakes such as a partial copy or a restored backup that overwrites newer work.
The sign of trouble is often a disagreement between two systems that should tell the same story.
Controls include limited write access, checks that detect unexpected change, and logs that show who altered a record.
None of those controls matter if the organization cannot explain which copy is the one it trusts.
Concrete examples
- A payment ledger and a warehouse system disagree about how many items were shipped on the same order.
- An attacker changes grades in a database, and a checksum comparison shows the table no longer matches the last approved export.
- A restore brings back yesterday's file and quietly discards corrections that staff made this morning.
Protective measures
- Name the trusted copy of each important record set and the only jobs allowed to change it.
- Alert when a protected record changes outside the application path you designated for updates.
- Test restores often enough to prove that you can return a known good copy without losing later authorized edits.
- Investigate disagreements between systems before staff manually force the numbers to match.
Related articles
Keeping data unreadable is a different goal from keeping it unaltered, and that goal is covered in What is data encryption?.
Showing that a record was preserved without silent change is part of the work in What is digital forensics?.
Copying data without permission can leave the original looking intact, which is why What is data theft? is a separate harm.