Before 1.0, only the newest released 0.x version receives security fixes.

Read THREAT_MODEL.md before using the package. The precise claim is:

Given the container alone, an adversary cannot distinguish a real encrypted slot from a randomly-filled empty slot with better than negligible advantage over guessing.

The physical slot count is observable. This project does not protect secrets on a compromised or unlocked host.

If you discover a security issue, do not open a public issue with exploit details.

1. Open a private GitHub Security Advisory if this repository is hosted on GitHub.
2. If a private advisory is unavailable, use the private maintainer contact published in the repository or package metadata.
3. Include a concise description, affected version, impact, and a proof of concept only when it can be shared safely.
4. Allow time for triage and remediation before public disclosure.

Never include a real passphrase, plaintext, derived key, or production container in a report. Use synthetic fixtures.

• Treat historical rsa_private.pem, rsa_public.pem, secret.key, and crypto_log.txt files as compromised.
• Rotate any secrets that were ever protected by those files.
• Use strong, unique passphrases from a password manager.
• Avoid swap and hibernation where the host threat model requires it.
• Restrict container file permissions and securely delete plaintext source data according to the storage medium's limitations.
• Never paste secret material into issues, logs, telemetry, or crash reports.

The format and cryptographic implementation have not been independently audited. A cryptography specialist must review THREAT_MODEL.md and src/keyprism/crypto.py before a 1.0 tag.