Skip to content

IDProva launches April 7 — Registry packages coming at launch. Build from source now.

Australian ISM

The Australian Information Security Manual (ISM) is the Australian Signals Directorate’s (ASD) framework for securing Australian Government ICT systems. IDProva’s identity, delegation, and audit capabilities directly address ISM controls across multiple security domains.

This mapping targets OFFICIAL: Sensitive and PROTECTED maturity levels and is relevant to agencies deploying AI agent systems subject to ISM compliance obligations. It is also relevant for IRAP assessors evaluating the security posture of such deployments.


Access control is one of the most directly addressed ISM chapters. DATs provide cryptographically enforced, scope-limited, time-bounded access tokens — a significant improvement over traditional API key or role-based approaches for agent access.

ISM-0430 — Identification of Users and Processes

Section titled “ISM-0430 — Identification of Users and Processes”

Requirement: Users and processes accessing systems are uniquely identified.

IDProva Coverage: ✅ Full

Every AI agent is assigned a unique did:idprova: identifier — a W3C Decentralized Identifier that is globally unique, cryptographically bound to the agent’s public key, and published in a verifiable registry. Unlike service accounts or shared API keys, an agent DID cannot be reused across agents.

Evidence for IRAP assessment: AID Document for each agent, available via GET /v1/aid/:id on the agency’s IDProva registry.


ISM-0431 — Authentication of Users and Processes

Section titled “ISM-0431 — Authentication of Users and Processes”

Requirement: Users and processes are authenticated before being granted access to systems and their resources.

IDProva Coverage: ✅ Full

Authentication in IDProva is cryptographic. A DAT presented by an agent carries an Ed25519 signature. The verifier checks this signature against the agent’s public key from their AID Document — no shared secrets, no password transmission. An agent that cannot produce a valid signature is rejected.

For MCP integrations, authentication occurs at session establishment. For HTTP APIs, it occurs on every request via the Authorization: IDProva <token> header.


ISM-0432 — Authorisation of Users and Processes

Section titled “ISM-0432 — Authorisation of Users and Processes”

Requirement: Access to systems and their resources is limited to what is required for users and processes to undertake their duties.

IDProva Coverage: ✅ Full — via DAT scope enforcement

DAT scope grammar (namespace:resource:action) enables fine-grained, resource-level authorisation that is:

  • Explicit — agents only have permissions that appear in their DAT
  • Enforceable — verifiers check scope before executing any action
  • Auditable — scope is recorded in every Action Receipt
  • Time-bounded — authority automatically expires

The scope reduction rule ensures sub-agents can never acquire permissions broader than their parent delegation.


Requirement: Privileged access to systems is controlled and monitored.

IDProva Coverage: ✅ Full

Privileged agent actions are constrained at the DAT level:

  • Privileged scopes (e.g., idprova:aid:write, mcp:tool:admin:*) require explicit inclusion in the issuer’s DAT
  • The maxActions constraint limits blast radius of a compromised privileged agent
  • Every privileged action produces a signed Action Receipt, creating an immutable audit trail
  • Privileged DATs should use short expiry windows (1–4 hours) for high-risk operations

Evidence for IRAP assessment: DAT tokens with privileged scopes + corresponding Action Receipt logs showing privilege usage.


Requirement: Approved cryptographic algorithms are used to protect the confidentiality, integrity, and authenticity of information.

IDProva Coverage: ✅ Full

IDProva uses ASD-approved cryptographic algorithms:

AlgorithmUseStandard
Ed25519Signing AIDs, DATs, receiptsFIPS 186-5 / RFC 8037
BLAKE3Receipt hash chainingModern, NIST-compatible
SHA-256Config attestation, interoperability hashingFIPS 180-4
ML-DSA-65Post-quantum signature (hybrid, planned)FIPS 204

The hybrid Ed25519 + ML-DSA-65 scheme provides defence in depth: an attacker must compromise both classical and post-quantum keys simultaneously to forge signatures.


Requirement: Cryptographic keys are generated, distributed, stored, used, and destroyed in a secure manner.

IDProva Coverage: ✅ Partial (full lifecycle management planned for v0.2)

Key Lifecycle PhaseIDProva Support
Generationidprova keygen — OS CSPRNG, Ed25519
DistributionPublic key published in AID Document via registry
StoragePrivate key stored locally; HSM recommended for high-assurance
UseSigning DATs and receipts; verification via AID Document
RotationAID update with new verification method
Destruction/RevocationAID deactivation — all DATs issued by agent become invalid

Implementation guidance for PROTECTED: Store private keys in a FIPS 140-2 Level 2+ HSM. Do not store private keys on shared infrastructure.


Requirement: TLS is used to protect information in transit.

IDProva Coverage: ✅ Complementary

IDProva protocol messages (DATs, receipts) include cryptographic integrity at the application layer regardless of transport. This means a compromised TLS session cannot result in undetected tampering of identity tokens. TLS remains required for confidentiality. Use TLS 1.3 with ASD-approved cipher suites for all registry and API communications.


Requirement: Audit logs are created for significant events and are protected from unauthorised modification.

IDProva Coverage: ✅ Full

Action Receipts are signed audit records — each receipt is protected from modification by:

  1. An Ed25519 signature over the entire receipt payload (the agent cannot deny authorship)
  2. A BLAKE3 hash chain linking each receipt to its predecessor (insertion, deletion, or modification of any receipt breaks the chain)

Any party with access to the agent’s public key can independently verify the integrity of the receipt chain.

Evidence for IRAP assessment: Receipt log files (JSONL format) + output of idprova receipt verify receipts.jsonl.


Requirement: Audit logs contain sufficient detail to identify the user or process that performed each action.

IDProva Coverage: ✅ Full

Every Action Receipt contains:

  • agent.did — the cryptographically verifiable identity of the acting agent
  • delegation.issuer — the DID of the human or system that authorised the agent
  • delegation.jti — the specific token ID, traceable back to the full delegation chain
  • agent.trustLevel — the verified trust level of the agent at time of action

This provides complete attribution from action → agent → authorising principal.


ISM-0988 — Logging of Privileged Actions

Section titled “ISM-0988 — Logging of Privileged Actions”

Requirement: Privileged actions are logged, including the identity of the user performing the action.

IDProva Coverage: ✅ Full

Privileged actions (those using privileged scopes in DATs) automatically produce Action Receipts. The receipt includes the scope used, the delegation that granted it, and the agent identity. Logging is enforced via the requireReceipt: true constraint in privilege-granting DATs.


Requirement: Audit logs from multiple sources are collected and correlated in a centralised system.

IDProva Coverage: ✅ Partial

Receipt logs are produced per-agent in JSONL format. Centralised log aggregation (shipping to a SIEM) is the responsibility of the deployment environment. IDProva receipts are structured (JSON) with consistent field names, making aggregation and SIEM correlation straightforward.

Implementation guidance: Ship receipt JSONL files to your SIEM (Splunk, Microsoft Sentinel, etc.) using standard log forwarders. Use agent.did as the correlation key across systems.


Requirement: Changes to system configurations are logged.

IDProva Coverage: ✅ Partial

The configAttestation field in AID Documents stores a hash of the agent’s configuration at deployment time. Changes to the agent’s configuration (system prompt, tool definitions, model version) require a new attestation hash and an AID update, which is logged in the registry.


ISM-1698 — Post-Quantum Cryptography Readiness

Section titled “ISM-1698 — Post-Quantum Cryptography Readiness”

Requirement: (Emerging) Systems implement or plan to implement post-quantum cryptographic algorithms.

IDProva Coverage: ✅ Roadmap committed

IDProva’s hybrid signature scheme is designed for PQC readiness from day one. The architecture accommodates ML-DSA-65 (FIPS 204) hybrid signatures with Ed25519, with the PQ component activated in v0.2. The pqalg and pqkid JWS header fields are already reserved in the protocol specification.


ISM ControlChapterIDProva FeatureCoverage
ISM-0430Access Controldid:idprova: unique agent identifiers✅ Full
ISM-0431Access ControlEd25519 signature authentication✅ Full
ISM-0432Access ControlDAT scope enforcement + scope reduction✅ Full
ISM-1508Access ControlPrivileged DAT constraints + receipt audit✅ Full
ISM-0336CryptographyEd25519, BLAKE3, SHA-256, ML-DSA-65 (planned)✅ Full
ISM-0457CryptographyKey generation, rotation, revocation lifecycle✅ Partial
ISM-0459CryptographyApplication-layer integrity + TLS guidance✅ Complementary
ISM-0580MonitoringSigned, hash-chained Action Receipts✅ Full
ISM-0585MonitoringAgent DID + delegation chain in every receipt✅ Full
ISM-0988MonitoringPrivileged action receipts via requireReceipt✅ Full
ISM-1405MonitoringStructured JSONL receipt logs for SIEM ingestion✅ Partial
ISM-0859MonitoringconfigAttestation hash in AID Documents✅ Partial
ISM-1698VulnerabilityML-DSA-65 hybrid PQC roadmap✅ Roadmap

For IRAP assessors evaluating systems that include AI agents using IDProva, the following evidence artefacts are relevant:

Assessment AreaEvidenceSource
Agent identity registerAID Documents for all agentsRegistry GET /v1/aid/:id
Access authorisation baselineDAT tokens issued to each agentidprova dat inspect <token>
Audit log integrityReceipt chain verification reportidprova receipt verify receipts.jsonl
Privilege usageReceipts for privileged-scope actionsReceipt log filtered by action.type
Config baselineconfigAttestation values for all agentsAID Documents
Key managementKey rotation logRegistry AID update history
Cryptographic complianceAlgorithm metadataGET /v1/meta on registry

Scope boundary note: IDProva addresses identity, delegation, and audit evidence for the AI agent layer. The underlying compute infrastructure, network controls, and data classification controls remain in scope for the broader system assessment.