Skip to content

Security: Be-bibek/sentinelmark

Security

SECURITY.md

Security Policy 🔒

Supported Versions

Only the current active release stream receives security updates and constant-time algorithmic audits.

Version Supported Notes
0.1.x ✅ Yes Active development and pre-publication branch.
< 0.1.0 ❌ No Deprecated proof-of-concept drafts.

Reporting a Vulnerability

We treat our research and production infrastructure security with absolute urgency. If you discover a cryptographic side-channel, timing leakage, memory safety gap, or protocol authentication bypass within SentinelMark, please report it privately.

Do NOT open a public GitHub issue.

Contact Information

Please email your findings directly to Bibek Das at: 📧 bibekdas1055@gmail.com

Disclosure Process

  1. Acknowledgement: You will receive confirmation of receipt within 48 hours.
  2. Triage & Patch: We will analyze the submission, reproduce the side-channel or vector, and develop a constant-time or memory-safe mitigation.
  3. Notification: Once the patch is merged into our upstream security branch, coordinated public disclosure will take place with full attribution to the researcher.

🛡️ Scope & Threat Model Covered

SentinelMark's protocol architecture actively guarantees protection against the following adversarial vectors:

1. Replay Attacks

  • Vector: An attacker intercepts a previously verified telemetry packet over the wire and retransmits it verbatim.
  • Mitigation: Bounded sliding-window nonce caching combined with arrival timestamp drift verification ($\pm 30\text{s}$) guarantees immediate rejection (DuplicateNonce or Expired).

2. Forged Telemetry

  • Vector: A compromised node or man-in-the-middle attempts to fabricate an arbitrary telemetry event without possessing the static hardware secret.
  • Mitigation: The watermark $W_i$ binds directly to the payload's content via HKDF-SHA256. Altering a single bit of the payload changes the internal canonical pre-image, invalidating the derived signature.

3. Log Reordering & Deletion

  • Vector: An internal adversary selectively drops specific forensic events to conceal malicious access.
  • Mitigation: Strict append-only hash chains ensure every single event's commitment relies on its predecessor's output. A dropped event permanently breaks all future evaluations (PrevHashMismatch).

4. Memory/Stack Scraping

  • Vector: Post-compromise memory dumps attempting to pull long-term static signing keys.
  • Mitigation: All secret arrays explicitly implement the zeroize trait to actively overwrite registers and stack bounds immediately after watermarking.

Out of Scope

  • Physical hardware level side-channel attacks (e.g., differential power analysis or direct bus-probing) on the static key storage module before it is ingested by the Rust Engine.
  • Compromise of the OS Kernel CSPRNG source (/dev/urandom or Windows BCryptGenRandom). If the operating system's root entropy is poisoned, CSPRNG assumptions fail globally.

There aren't any published security advisories