Safeguard
Application Security•

ASVS 5.0 V17.2.7: Keep Call Recording Alive During an SRTP Flood

ASVS V17.2.7 requires recording components tied to the media server to keep working during a flood of legitimate SRTP traffic. Design and test.

Safeguard Security Team
2 min read

ASVS 5.0 V17.2.7 requires that any audio or video recording mechanism associated with the media server keeps processing incoming media during a flood of SRTP packets from legitimate users. It is a Level 3 requirement.

This extends V17.2.5 to the recorder. Recorders write to disk, encode, and upload, and they are often slower than the forwarding path.

Why it matters

If the recorder falls behind, memory grows, the process dies, and the recording is lost. In regulated settings, a missing recording can itself be a compliance failure. A flood from one participant should not cost the whole room its recording.

How to implement it

  • Run the recorder as a separate process or container from the forwarding path, so a recorder crash does not drop live media, and the reverse.
  • Use a bounded queue between the media server and the recorder. Decide the overflow policy: drop packets from the flooder first, never block the live path.
  • Apply the same per participant rate limits as in V17.2.5 before packets reach the recorder.
  • Write to local disk first, upload after, with free space alerts.
  • Alert when queue depth or lag crosses a threshold.

How to test it

Start a recorded room with several normal participants. Have one participant send at 10 times the normal rate for a few minutes. Afterwards check that the recording exists, plays back, covers the whole period for the normal participants, and that recorder memory stayed bounded.

Storage, retention and privacy

Recordings are sensitive data. Encrypt them at rest, restrict access, and apply retention rules, and make sure the recorder's own failure modes do not leave plaintext fragments in temporary directories. A disk full condition should stop the recording cleanly and raise an alert, not crash the host.

Common mistakes

  • Recorder in the same process as the forwarder.
  • An unbounded in memory buffer.
  • No alert when the recorder lags.

Quick check

Which should win under pressure, live media or the recorder queue?

Never block live media. Bound the recorder queue and drop from the offending stream first.

Sources

Never miss an update

Weekly insights on software supply chain security, delivered to your inbox.

Self-healing security runs on Safeguard.

Your first fix PR is minutes away.

No sales call required, even your agent can complete the purchase over MCP.