The LLM Memory Vault is designed with a "paranoid" security model. We assume that:
- All input data may contain PII.
- The storage medium could be inspected.
- External APIs are untrusted for sensitive data.
| Level | Description |
|---|---|
| PUBLIC | Can be shared freely. |
| INTERNAL | Personal context, no external sharing. |
| SENSITIVE | Must be redacted before any processing. |
| SECRET | Purge immediately, never store. |
- Detection: All inputs pass through PII detection (Presidio + LLM Guard).
- Redaction: Detected PII is masked or tokenised before storage.
- Encryption: All content blobs are encrypted at rest using AES-256 (Fernet).
| Version | Supported |
|---|---|
| 0.1.x (dev) | Active development |
Note: Project Memento is in active development. No stable release has been published yet.
If you discover a security vulnerability within this project, please send an e-mail to tbc@richardham.co.uk. All security vulnerabilities will be promptly addressed.
Please include the following in your report:
- Type of issue (e.g., key exposure, encryption bypass, PII leak, etc.)
- Location/path of the issue
- Steps to reproduce
- Potential impact
We will acknowledge receipt of your vulnerability report within 48 hours and will send you updates during our remediation process.
- We will not disclose the vulnerability to third parties without your consent.
- We will provide a fix in a timely manner.
- We will publicise the vulnerability (with your credit) once it is patched.