A digital signature protects the integrity and authenticity of data through asymmetric cryptography. A signer uses a private key to generate a signature over a message, generally through an algorithm that incorporates a cryptographic hash. A verifier uses the corresponding public key, the message, and the signature to check validity. Changing the signed bytes causes verification to fail. The private key must remain protected, while the public key can be distributed to verifiers.
Successful verification establishes a relationship between the data and a particular key, not automatically a trusted person or organization. The verifier must obtain the public key through a trusted channel, validate an appropriate certificate chain, or use an explicitly approved key identifier. Accepting a public key supplied alongside an untrusted file without checking its provenance allows an attacker to replace the file, signature, and key together.
A practical cloud example is signing a release artifact before uploading it to object storage. A controlled build pipeline produces the artifact, requests a signature from a managed signing service or protected asymmetric key, and publishes the signature with the artifact. Before installation or deployment, a consumer verifies the signature against an approved public key. Verification must apply to the exact artifact that will be used; verifying one object version and subsequently downloading a different version creates a verification gap.
Digital signatures differ from encryption, checksums, and message authentication codes. Signing does not conceal data, and encryption alone should not be assumed to authenticate its producer. An ordinary checksum detects changes only when the expected checksum is trustworthy; an attacker who can replace both a file and its checksum can bypass that check. A message authentication code uses a shared secret, so any party able to verify with that secret can also generate valid codes. Public-key signature verification does not require possession of the signing secret.
For data at rest, signatures detect unauthorized changes but do not prevent overwrites or deletion. Amazon S3 Object Lock and S3 Glacier Vault Lock provide retention controls rather than digital signatures. S3 Versioning preserves object versions but does not authenticate their contents. These controls complement signing: retention protects evidence from removal, versioning supports recovery and precise artifact selection, and signatures support authenticity and integrity checks. Amazon S3 object checksums are useful for transfer and storage integrity, but they are not publisher signatures.
Audit validation is another application. AWS CloudTrail log file integrity validation delivers digitally signed digest files that reference log files and link successive digests. Running validation can detect changes to delivered logs and gaps in the digest chain. Enabling digest delivery alone does not continuously validate logs or stop an attacker from deleting them; retention controls and scheduled validation remain necessary.
Operational trade-offs include signing latency, service quotas, verification overhead, and key lifecycle management. Restrict signing permissions separately from verification permissions, because an attacker who can request signatures may authenticate malicious artifacts without extracting the key. Plan key rotation, trusted-key distribution, compromise response, and retention of verification material. A valid signature does not prove that software is safe, that data is current, or that an authorized signer acted correctly. Prevent replay or rollback by signing relevant context, such as artifact identity and version, and enforcing an acceptance policy.