{
  "slug": "key-commitment",
  "term": "Key commitment",
  "aliases": [
    "Key robustness",
    "Key collision resistance"
  ],
  "definition": "Key commitment makes it computationally difficult to construct a ciphertext that authenticates under two different keys.",
  "note": "The guarantee concerns adversarially chosen keys and inputs, not just accidental wrong-key rejection. RFC 9771 allows the nonce, associated data and plaintext to differ between the two encryption inputs. Full commitment binds all inputs and implies key commitment; the converse does not hold. [rfc9771]\n\nStorage relevance: Albertini et al. analyze malicious key import during rotation and inconsistent DEKs supplied to different envelope recipients. The attacker controls the candidate keys; this does not recover an arbitrary secret key. Ordinary authenticated encryption can satisfy its usual security goals while lacking this extra property. [albertini2022]\n\nMigration example: from version 2.0.x, the cited encryption SDK distinguishes ForbidEncryptAllowDecrypt (noncommitting writes, either read format), RequireEncryptAllowDecrypt (committing writes, either read format) and RequireEncryptRequireDecrypt (committing writes and reads only). The strict default rejects legacy messages without commitment; possession of their keys is insufficient for acceptance. [sdk-policy]\n\nBefore changing writers, upgrade every required reader. Before strict decryption, resolve retained legacy ciphertext, including recovery copies. A wrapping-key rotation is a different operation from migrating the payload format. [sdk-migration]",
  "status": "shipping",
  "advantages": [],
  "limitations": [
    {
      "text": "A successful round trip or wrong-key test is not a proof of a cryptographic commitment bound.",
      "citationIds": [
        "rfc9771"
      ]
    }
  ],
  "related": [
    "authenticated-encryption",
    "aes-gcm-siv",
    "envelope-encryption",
    "key-rotation"
  ],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "updated": "2026-11-22",
  "citations": [
    {
      "id": "rfc9771",
      "title": "RFC 9771: Properties of AEAD Algorithms, sections 4.3.2-4.3.3 (Informational, 2025)",
      "url": "https://www.rfc-editor.org/rfc/rfc9771.html#section-4.3.3",
      "accessed": "2026-11-22"
    },
    {
      "id": "albertini2022",
      "title": "Albertini et al.: How to Abuse and Fix Authenticated Encryption Without Key Commitment. USENIX Security 2022, pp. 3291-3308, sections 2-3.",
      "url": "https://www.usenix.org/system/files/sec22-albertini.pdf",
      "accessed": "2026-11-22"
    },
    {
      "id": "sdk-policy",
      "title": "AWS Encryption SDK: Setting your commitment policy",
      "url": "https://docs.aws.amazon.com/encryption-sdk/latest/developer-guide/migrate-commitment-policy.html",
      "accessed": "2026-11-22"
    },
    {
      "id": "sdk-migration",
      "title": "AWS Encryption SDK: How to migrate and deploy, staged reader and writer rollout",
      "url": "https://docs.aws.amazon.com/encryption-sdk/latest/developer-guide/migration-guide.html",
      "accessed": "2026-11-22"
    }
  ],
  "provenance": [
    "Source synthesis, not an attack reproduction or a cloud-service assessment. Historical scenarios in the 2022 paper must not be read as current product vulnerabilities. [albertini2022]",
    "The migration policy names describe the AWS Encryption SDK, not a universal KMS API. Source documentation reviewed on 2026-11-22. [sdk-policy]"
  ]
}
