Skip to main content
A security product is only as trustworthy as its honesty about limits. This threat model states what ZeroKeyUSB defends, against whom, and where the boundaries are. It is grounded in the actual firmware, which is open-source and verifiable.

Assets

Adversaries and outcomes

What it protects against

  • Browser and cloud password leaks — credentials never live in a browser or a cloud vault.
  • Malware that scrapes stored passwords — the vault is off-host and encrypted with a key the host never sees.
  • Credential theft at rest — dumping the EEPROM yields only ciphertext.
  • Cloning — the AES key is device-bound inside the secure element.
  • Unsigned/rogue firmware — rejected at boot.
  • Online PIN brute force — exponential backoff, re-applied at every boot so it cannot be skipped by power-cycling; the vault is never destroyed by wrong PINs.

What it does not protect against

  • A compromised host during use. Keystrokes typed to an infected computer can be captured there. ZeroKeyUSB reduces storage exposure, not the risk of typing into a machine that is already owned by an attacker.
  • Plaintext USB backup on an untrusted machine. The export sends credentials in clear text to the host. Perform backups only on a trusted, offline computer.
  • Offline PIN cracking after physical bus access. The PIN hash is readable over I²C (a known trade-off of the secure-element SKU). An attacker who removes the epoxy and reaches the bus can attempt an offline SHA-256(PIN ‖ serial) search. Mitigations: the epoxy encapsulation and using a long PIN. There is no destructive self-wipe.
  • Coercion / shoulder-surfing of the PIN. Standard operational-security concerns apply.

Design consequences for buyers

  • Use ZeroKeyUSB on managed or personal machines you trust for the moment of typing; it is strongest exactly where stored-password managers are weakest (shared, infected, unmanaged hosts).
  • Treat the backup file as sensitive and generate it offline.
  • Choose a long PIN for high-value deployments; it is the last line against a lab-grade physical attack.
This analysis maps directly onto the control statements in ISO 27001/27002 and NIST SP 800-63B.