Four layers, four boundaries. A failure in one does not reach the next.

Skinshells Voice in layers Request access
Security

The security model

Published in full, with the defaults stated and the exceptions described, because a security posture that cannot be inspected is a claim rather than a posture.

01

What is on by default

Encrypted signalling and media wherever the endpoint supports both. Address-based authorisation for fixed sites, which removes the credential rather than protecting it. Destination policy that denies everything and permits only what the organisation actually calls.

Concurrency and spend thresholds that contain automatically rather than notify. Role-based administrative access with individual, attributable accounts and no shared credentials.

Turning any of these off is a recorded decision attributable to a named person, reviewed at the next service review. It is not something that quietly happens because an endpoint made it inconvenient.

02

Why the thresholds contain rather than alert

Toll fraud is automated, non-targeted and continuous, and it begins outside working hours because the variable it exploits is detection latency rather than technical difficulty.

An alert into an unattended inbox on a Friday evening is not a control. It is a record of what happened, delivered after it finished.

So the thresholds act: concurrency ceilings, rolling spend limits and destination deviation, each capable of stopping traffic on their own. We accept occasional false positives as the price of a control that works while nobody is watching.

03

What each control produces

Every control here generates an artefact rather than only a behaviour. Configuration records for the encryption and authorisation settings. A policy record plus change history for destination policy. Threshold configuration and an alert log for containment. An access log and a periodic review record for administrative access.

The reason is practical: a control you cannot evidence is one you will have to demonstrate from memory during an audit, and memory is a poor evidential basis.

04

What we do not claim

This page describes a technical posture. It does not assert any certification, accreditation, regulatory approval or carrier authorisation, and nothing elsewhere on this site does either.

Where a specific status bears on your procurement process, ask directly and the answer will be precise and in writing, including where the precise answer is no.

Reference

Protection by boundary

Boundary Control Evidence produced
Connectivity TLS signalling, SRTP media Per-trunk configuration record
Connectivity Address-based authorisation Authorised address list
Connectivity Destination policy, deny by default Policy record and change history
Control Role-based administration Access log, periodic review record
Control Recording access and retention Access log, retention configuration
Numbering Presentation authority Authority evidence per range
Numbering Port-out verification Authorisation record
Operation Change attribution Before and after state per change
Questions

Asked about this page

Questions answered

Yes, where an endpoint genuinely cannot support something. It becomes a recorded exception attributable to a named person and is revisited at the next review rather than becoming permanent by inattention.

This site makes no certification or accreditation claim. If a specific standard matters to your procurement, ask and you will get a precise written answer, including where that answer is no.

Next

Ask about anything on this page.

Including where you disagree with it. That conversation is usually more useful than the agreeable one.