Skip to main content

Multipaz 0.102.0 Released

· 9 min read

Written by David Zeuthen, Multipaz project leader.

Keeping up our steady 6-to-8 week release cadence as we charge toward Multipaz 1.0, we are thrilled to announce that Multipaz 0.102.0 was officially released last week on October 22, 2026!

This release is packed with foundational capabilities spanning biometric verification, quantum-resistant cryptography, zero-knowledge proof build pipelines, and deep defense-in-depth security hardening across every layer of the stack.

You can find the full release notes and changelog on GitHub:

As always, all release artifacts are published directly to Maven Central.

Let’s dive into the headline features of Multipaz 0.102.0, followed by a tour of everything else that landed in this release!


1. Biometric Face Matching & ISO/IEC 23220-5 Portrait Image Equivalence​

When a credential is presented from a mobile wallet, traditional device user authentication (such as Android BiometricPrompt, Face ID, Touch ID, or a device passcode) proves that an enrolled user of the device unlocked the hardware-backed key. However, high-assurance relying parties often need to know something more specific: does the live human presenting the device right now actually match the portrait photo issued inside the credential—without the relying party necessarily having to receive and inspect the raw portrait image itself or operate biometric matching hardware at the point of verification? To solve this, Multipaz 0.102.0 implements the proposed protocol in ISO/IEC 23220-5 for "portrait image equivalence", enabling a relying party to request the CHV_1 data element (in the org.iso.23220.5.1 namespace) and receive back a device-signed boolean from the wallet that is set to true if and only if the device biometrically matched the live human user of the device to the credential's portrait image! This works end-to-end across both ISO mdoc (ISO/IEC 18013-5) and OpenID4VP presentments, governed by relying-party-aware FaceMatchingMode policies (NEVER, ONLY_IF_REQUESTED, and ALWAYS) and MSO device-key element authorization checks.

To power this experience across platforms, we introduced a new cross-platform biometric verification framework (FaceMatcher, FaceMatcherSession, and FaceMatcherLivenessSession) complete with full-screen interactive prompts in both Compose Multiplatform and SwiftUI featuring automatic screen brightness boosting and background illumination control. Alongside the core abstractions, the new multipaz-facenet module delivers a complete, on-device implementation across Android, iOS, and JVM using MediaPipe BlazeFace for face detection and 4-point Procrustes alignment, MobileFaceNet for embedding extraction, single-frame passive anti-spoofing (MiniFASNetV2), and configurable head-pose capture for onboarding and issuance flows. Note that multipaz-facenet is included for development and testing purposes only and does not come with any guarantees of biometric performance, accuracy, or security (see the Performance & Security Notice for details); production deployments with strict assurance requirements should plug in a certified biometric and liveness verification solution via the FaceMatcher interface.

Check out this video to see biometric face matching and ISO/IEC 23220-5 portrait image equivalence in action:


2. Readiness for Post-Quantum Cryptography (PQC)​

Preparing the digital identity ecosystem for the post-quantum era requires real-world code that standards authors and wallet engineers can build with today. In Multipaz 0.102.0, we have upgraded our multiplatform Crypto and SecureArea wrappers with full support for ML-KEM (FIPS 203 key encapsulation: ML_KEM_512, ML_KEM_768, ML_KEM_1024) and ML-DSA (FIPS 204 digital signatures: ML_DSA_44, ML_DSA_65, ML_DSA_87), as well as classic RSA (RS256/384/512 and PS256/384/512). Under the hood, our platform-specific cryptographic wrappers tap directly into native and proven cryptographic implementations on every target: Java Cryptography Architecture (JCA) and Bouncy Castle (bcprov-jdk18on, loaded dynamically via reflection so the core SDK stays lightweight without forcing a mandatory dependency) on JVM and Android; Apple's CryptoKit on iOS; and the browser's W3C Web Crypto API (SubtleCrypto) on the Web (Kotlin/JS and WasmJS). Even better, hardware-backed PQC is already wired up in AndroidKeystoreSecureArea (supporting ML-DSA on Android 17 / KeyMint 5.0+) and SecureEnclaveSecureArea (supporting ML-DSA and ML-KEM via CryptoKit in the Apple Secure Enclave on iOS 26+), alongside SoftwareSecureArea, CloudSecureArea, and our in-browser Multipaz Web Developer Tools.

With ML-KEM and ML-DSA plumbed all the way through our PublicKey, PrivateKey, and Signature abstractions, X.509 certificates, COSE/CWT, and JWS, Multipaz is ready to start end-to-end prototyping of Post-Quantum Cryptography across SD-JWT VC, ISO mdoc, OpenID4VP, and ISO/IEC 18013-5. We are actively participating in the PQC standardization discussions across the Standards Developing Organizations (SDOs)—including ISO/IEC SC17 WG10, the IETF, and the OpenID Foundation—and having working, cross-platform implementations allows us to benchmark payload sizes, test key encapsulation in session encryption, and bring concrete implementation feedback directly back to the working groups.


What Else is New in 0.102.0?​

Beyond biometric face matching and post-quantum cryptography, Multipaz 0.102.0 includes a massive set of improvements across the entire repository:

  • Key Material Memory Zeroing and Type Safety: Introduced SecureByteString and SecretKey (AutoCloseable) with deterministic memory wiping (ByteArray.secureZero()), GC safety-net cleaners (KeyDisposer), constant-time equality checks, and private key duplication (PrivateKey.duplicate()). Updated PrivateKey, AsymmetricKey, SessionEncryption, Hkdf, Crypto, and SecureArea (keyAgreement() and kemDecapsulate()) to enforce strict type safety between input keying material and symmetric cipher keys while securely zeroing sensitive buffers upon disposal.
  • Cryptographically Secure and Testable Randomness: Added Crypto.secureRandom backed by platform CSPRNGs (SecureRandom on JVM/Android, SecRandomCopyBytes on iOS, and crypto.getRandomValues on the Web) and updated security-sensitive APIs across the SDK and servers to accept a Random parameter defaulting to Crypto.secureRandom.
  • Cloud Secure Area CBOR-RPC Rework: Replaced the custom Cloud Secure Area command protocol, E2EE session layer, and manual state serialization with Multipaz CBOR-based RPC (@RpcInterface CloudSecureAreaProtocol and @RpcState server states) and device-assertion authentication (RpcAuthIssuerAssertion and RpcAuthClientSession).
  • Longfellow ZKP Rust Migration and 16 KB Page Alignment: Switched multipaz-longfellow from the legacy C++ JNI pipeline to an in-repo Rust cross-compilation pipeline (longfellow-zk) using JNA across JVM and Android, and updated JNA to 5.19.1 with 16 KB ELF segment alignment across all Android native libraries for full Android 15+ compatibility.
  • Credential Structure Validation: Added comprehensive validation routines (validate(), validateAgainstMso(), and validateVc()) and a structured ValidationResult / ValidationFinding DSL for inspecting ISO/IEC 18013-5 (IssuerSignedItem, IssuerNamespaces, MobileSecurityObject, MdocCredential) and RFC 9901 / SD-JWT VC (Disclosure, SdJwt, SdJwtVcCredential) payloads prior to certification.
  • Digital Credentials API Module Split and Credman Enhancements: Split multipaz-dcapi into platform-specific multipaz-dcapi-credman (Android) and multipaz-dcapi-ios (iOS) modules with explicit instantiation. Added support in multipaz-dcapi-credman for omitting claim values when registering credentials (Document.androidCredmanOmitClaims), matching OpenID4VP DCQL array claim paths (null wildcards and integer indices), and encoding native Android app origins with URL-safe Base64.
  • KMP NFC APDU Routing and mdoc Services: Moved mdoc NFC engagement and data-transfer services (NfcApduRouter, NfcApduService, MdocNdefService, MdocNfcV2Service, and MdocNfcDataTransferService) from multipaz-compose (androidMain) to multipaz (commonMain) with strongly-typed CommandApdu/ResponseApdu routing and NfcDeactivationReason.
  • Google Android Key Attestation Verifier on JVM: Delegated validateAndroidKeyAttestation() on JVM (jvmMain, targeting Java 21 while keeping Android targets at Java 17) to Google's com.android.keyattestation:keyattestation library with certificate revocation checking support.
  • Secure-by-Default API and Configuration Parameters: Audited and hardened default parameter values and server configuration fallbacks across device and key attestation, credential validation (now = Clock.System.now()), SdJwt.verify() / SdJwtKb.verify(), and Cloud Secure Area settings.
  • iOS Secure Enclave Key Invalidation and Unlock Reasons: Added biometric domain-state tracking in SecureEnclaveSecureArea to detect key invalidation (getKeyInvalidated() and KeyInvalidatedException) for BIOMETRY_CURRENT_SET keys, along with automatic localized unlock reason conversion via PromptModel in SecureEnclaveDefaultKeyUnlockDataProvider.
  • Doctype Updates (EU PID and Aadhaar): Aligned EUPersonalID with the EUDI PID Rulebook v1.7 (structured place_of_birth, JSON-only house_number, and updated age and portrait equivalence requests) and aligned Aadhaar attribute identifiers and request selectors with Google Wallet.
  • BLE, NFC, and Presentment Reliability Fixes: Fixed NFC engagement and BLE session cleanup on repeated taps, closed NFC tags on reader transport failure so the iOS system scan sheet dismisses promptly, fixed an iOS BLE peripheral stall when sending large responses off the main thread, failed BLE peripheral transports immediately upon central disconnection, retried Android BLE service discovery after connection, and rejected duplicate transports in waitForConnection().
  • Protocol and Cryptographic Bug Fixes: Fixed empty CRL parsing and generation per RFC 5280, fixed JsonWebEncryption.encrypt() Concat KDF input when apu/apv are absent per RFC 7518, omitted empty _sd arrays when generating SD-JWTs, fixed query-only URI parsing and improved HTTP error diagnostics in uriSchemePresentment(), and derived the MSO COSE signing algorithm from the document signer key in DocumentType.createMdocCredentialWithSampleData().

Join the Community!​

Multipaz is a proud OpenWallet Foundation project and is completely free and open-source. Whether you are building digital wallets, deploying verifiers, or helping shape international identity standards, we’d love to have you join our community! We meet every Tuesday—check the OWF community calendar for details.

Until the next release—happy coding!