Skip to content

Latest commit

 

History

History
49 lines (38 loc) · 2.64 KB

File metadata and controls

49 lines (38 loc) · 2.64 KB

Threat model

Protected values

Protected values are Stage 0, its embedded Ed25519 public key, confirmed slot images, redundant boot metadata and rollback floor, runtime security events, authenticated update packages, research markers, and the public device identity fingerprint. Private signing seeds are operational secrets rather than device-readable values.

Attacker capabilities

The model covers arbitrary UART bytes, malformed or replayed packages, old but correctly signed images, corrupted metadata, a compromised application image, faulty signed firmware, public knowledge of protocols and addresses, and physical access while RDP is disabled. It records the future RDP2 case, where debug and ROM-bootloader recovery are unavailable, and the consequences of a lost or stolen signing seed.

The model does not assume that a signature makes faulty code safe. A correctly signed image can still exhaust the trial budget, fault, or fail health checks.

Out of scope for Part 2

Invasive analysis, decapping, laser or electromagnetic injection, voltage or clock glitching, laboratory side channels, and a physical power-loss campaign are not tested here. Power-loss testing is a separate campaign and remains open.

Security claims and evidence

Claim Implementation/evidence Assumption or limit
Only authorized firmware boots manifest, hash, Ed25519, vector and version checks Stage 0 and public key are trusted
Only authorized firmware installs same checks before erase, then readback and commit software policy until WRP/RDP
Old firmware is rejected minimum version plus confirmed metadata floor no hardware monotonic counter
Stage 0 is not a normal update target restricted write regions; negative tests not physically WRP-protected
Metadata ambiguity fails closed CRC, commit markers, canonical-state validation corruption simulation is not power-loss evidence
Recovery does not bypass security recovery uses the same signed verification requires the valid signing key
Application writes are constrained MPU read-only region plus confirmation exception application is privileged
Test/production keys are distinct key tool, labels, offline process operators must verify fingerprints

Residual high-impact attacks

An attacker with the private signing seed can authorize arbitrary code. A compromised privileged application can disable the MPU or call the existing metadata confirmation primitive. Corruption of Stage 0, both metadata copies, or both slot images can require signed recovery or pre-RDP SWD repair. RDP2 does not repair software defects and removes ordinary repair paths.