Skip to content

Security: AI2TH/cryptship

Security

SECURITY.md

Security policy

Reporting a vulnerability

Please report security issues privately, not as a public issue or PR.

  • Preferred: GitHub → the repository's Security tab → Report a vulnerability (private vulnerability reporting / security advisory).
  • Include: affected version or commit, a description, and ideally a minimal reproduction (a crafted license, a sequence of commands, or a short test).

We aim to acknowledge a report within a few days and to agree on a disclosure timeline with you. Please give us reasonable time to fix before any public disclosure.

Supported versions

cryptship is pre-1.0 and moves fast. Fixes land on main and in the latest tagged release; please verify against the latest release before reporting.

Scope — what is and isn't a vulnerability

cryptship's threat model is stated plainly in the README and docs/ARCHITECTURE.md. In particular, these are known, documented limitations, not vulnerabilities:

  • Host/container root extracting the running plaintext (dumping tmpfs or /proc/<pid>/mem). Code that runs on hardware the attacker controls can always be extracted; only a confidential-compute key provider (SEV-SNP/TDX, the attested roadmap) or remote execution prevents this.
  • A patched loader skipping loader-enforced policy (expiry comparison, revocation). The crypto design makes license forgery and claim tampering fail the unwrap; it does not stop someone from modifying the open-source loader they run.
  • attested mode without a real verifier. It ships with test-only attestation; cryptship-server --require-attestation fails closed precisely so this isn't mistaken for protection.
  • Enabling CRYPTSHIP_ONLINE_GRACE on an untrusted host (documented trade-off), and software machine fingerprints being spoofable by root.

In scope and very welcome: key/secret handling bugs, signature or AEAD verification that can be bypassed, nonce/replay/downgrade flaws, path-traversal or extraction escapes, anything that lets an unprivileged in-container attacker read the decrypted app, or any way to unwrap the content key without legitimate license material.

There aren't any published security advisories