Read local signals
Android reports the security checks available on the device. These are client-reported signals, not verified facts.
Valkirie Shield turns Android threat signals into a server-side risk decision. The full assessment stays private; a compact cryptographic commitment can be anchored to Solana Devnet so later changes are detectable.
Explore the approachHackathon prototype · Real rules and audit code · Simulated attestation in the demo
The phone supplies evidence. The backend makes the decision. The audit record makes later changes detectable without publishing a private report.
Android reports the security checks available on the device. These are client-reported signals, not verified facts.
Versioned, deterministic rules turn the submitted evidence into a risk score and triggered findings on the server.
PostgreSQL retains the complete assessment and canonical evidence needed to reproduce its hash.
When configured, a worker submits compact hash evidence to Solana Devnet and tracks confirmation separately from the verdict.
The Android app reports local checks, but it does not get to assign its own security verdict. Versioned rules on the backend calculate risk and record which findings led to the result.
Why blockchain here? It offers an external place to compare an assessment’s cryptographic fingerprint later. Valkirie Shield stores the complete report privately, hashes its canonical representation with SHA-256, and prepares a compact Solana Devnet Memo commitment when the audit worker is configured.
The assessment returns independently of the chain. A submitted transaction is only an attempt; the audit state becomes confirmed after the worker verifies a successful finalized transaction with the expected Memo and signer.
A versioned SHA-256 digest represents the stored assessment.
The backend signer submits the compact commitment; Android does not hold an audit key.
The transaction reference is exposed only after confirmation checks succeed.
The chain can help detect a changed report. It cannot prove that client signals were true, the risk decision was correct, or a device was safe.
A working backend path exists today. The demo makes its simulated integrations explicit so a mock result cannot be mistaken for device verification.
The backend can register devices, evaluate security signals, persist assessments, and queue audit work in the same database transaction.
The end-to-end staging flow simulates authentication and Play Integrity. Results are marked mock; they do not establish verified device integrity.
A hash on-chain can show that a committed assessment has not changed. It does not independently prove that a device was safe.