Blog: In practice

Three ways operators use Open Voice Shield

By the TelecomsXChange engineering team ยท

Operators run Open Voice Shield three ways: a security eye on one client, random call checks across the base, and proof when a carrier, client or regulator asks what happened. All three use the same hop, verdict and evidence. Only the volume and the follow-up differ.

01

A security eye on one client

Maybe it is a new account with thin KYC, a reseller whose minutes jumped overnight, or a client whose calls feel wrong. You do not want to re-route your whole network to find out. Route that one client through Open Voice Shield and every one of their calls comes back scored, with the recording and the transcript to read.

The setup

  1. Under Network, add the destination those calls should go on to, and a routing rule for the client's dialled prefix so their calls come back to the right place. Routing rules match the longest prefix first.
  2. On your switch, send that client's traffic to the Open Voice Shield ingress instead of straight to the route. Nothing else changes for the rest of your customers.
  3. Under Automation, create a rule: the probability threshold, the recommendation that counts, the action Block the caller, and how long the block lasts.

What happens next

Each call is recorded, transcribed and analyzed within minutes of hangup, and the verdict (block, review or allow) lands on the call record with the red flags behind it. The moment a verdict crosses your threshold the caller's number is put under control: the next INVITE from it is refused with SIP 403 before it reaches your route. Controlled numbers are listed on the Automation page with the reason and the expiry, and you can release one in a click. An alert rule can page the desk at the same time.

The Automation page of the customer portal: rules that block or divert a caller when an analyzed call crosses a probability threshold, and the list of controlled numbers with their reason and expiry.
Automation rules and the numbers they have put under control. A block answers the next call with SIP 403. A divert sends it to a destination of yours instead.

02

Random call checks across your base

You cannot listen to a thousand calls, but you can read a thousand verdicts. Send a slice of traffic through (a prefix range, one route or one busy hour) and read the results the same afternoon. The robocall campaign shows up as a cluster of near-identical transcripts. The fake card-services desk is a 95 percent verdict with the card number request quoted in the red flags. The pressure script is a compliance note that reads "no opt-out offered".

The setup

  1. Whitelist your switch's address and add your destinations under Network, once.
  2. On your switch, send the slice through the ingress. Leave the rest as it is.
  3. Under Calls, filter to high risk, sort by probability, and open the top of the list. Each record has the transcript, the verdict and the compliance note.
  4. When you have what you need, route the slice back. The call records stay.

It is prepaid and billed per second of analyzed calling plus one analysis fee per call, with no monthly minimum and nothing to cancel, so an afternoon of checks costs an afternoon of calls. Pricing has the worked examples, and why the count matters explains what a campaign costs when it is found late.

The Calls page of the customer portal: every analyzed call with its probability, category, recommendation, estimated MOS, post-dial delay and status, with filters for status, probability, category and date.
The Calls page after an afternoon of checks: filter to high risk, sort by probability, and the campaign is at the top.

03

Proof when someone asks

A carrier disputes the minutes. A client denies the campaign was theirs. A regulator asks what happened on a number. The answer is usually "we believe" until someone finds a recording. Every analyzed call in Open Voice Shield keeps its evidence: the stereo recording with the caller on the left and the callee on the right, the speaker-labelled transcript, the verdict with its red flags, the SIP ladder, the packet capture of the signalling, and the audio quality of each leg.

Two ways to hand it over

Export the report
One action on the call record zips all of it into an analyst bundle with a written report. It opens on any machine, days later, with no Open Voice Shield account.
Share a link
A share link shows the call's evidence in a read-only view. It is protected by a generated password, expires after the hours you choose, and can be rotated or revoked at any time. Put it in front of the carrier, the client or the regulator.

For anything that needs more than one person, open a case on the call: assign it, discuss it in comments, and close it as resolved or dismissed, with every change audited.

Sharing a call from the customer portal: a dialog that creates a password-protected share link with an expiry, and the read-only view a recipient sees with the recording, transcript and verdict.
A share link: generated password, chosen expiry, revoke or rotate at any time, and a read-only view for whoever opens it.

Where to start

Which one first

If one client worries you, start with the security eye: it is one routing rule and one automation rule, and it pays for itself the first time a block lands. If nothing worries you yet, run the random check for an afternoon and read what comes back. Most operators find something. Proof is not a starting point. It comes from the other two, and it is there whenever someone asks. A mobile operator has a fourth: selling scam call protection to its subscribers as a monthly add-on.

New here? Watch the six-minute tour, read getting started, or start free with $5 credit.

Questions

Frequently asked questions

How do I run only one customer's calls through Open Voice Shield?

Point that customer's traffic at the Open Voice Shield ingress from your switch, and on the platform add a destination and a routing rule for their dialled prefix so their calls come back to the right place. Nothing else on your switch changes, and every call of theirs returns with a fraud verdict, the recording and the transcript.

Can Open Voice Shield block a fraudulent caller automatically?

Yes. An automation rule watches every analyzed call. When the fraud probability crosses the threshold you set, with the recommendation you choose, the caller's number is put under control and the next call from it is refused with SIP 403 before it reaches your route, for an hour, a day, a week or for good. Controlled numbers are listed and can be released.

What does it cost to spot-check a sample of my traffic?

Open Voice Shield is prepaid and billed per second of analyzed calling plus one analysis fee per call, with no monthly minimum. Send a slice of traffic through for an afternoon, read the verdicts, and route it back. You pay for the calls that were analyzed and nothing after.

What evidence does an analyzed call keep?

The stereo recording with the caller on the left and the callee on the right, a speaker-labelled transcript, the fraud verdict with its red flags, the SIP ladder, the packet capture of the signalling, and the audio quality of each leg. One action exports all of it as a zip with an analyst report that opens without an account.

How do I share a call's evidence with a carrier or a regulator?

From the call record, create a share link. It is protected by a generated password, expires after the hours you choose, and can be rotated or revoked at any time. The person who opens it sees the call's evidence in a read-only view without an account.