New revenue for mobile operators: scam call protection your subscribers pay for

Voice revenue per user has been flat or falling for a decade while the cost of handling scam-call complaints, regulator filings and churn has not. Caller-ID authentication such as STIR/SHAKEN, mandated in Canada since 2021, was a necessary step, and the Canadian Centre for Cyber Security's own guidance is clear about its limit: it authenticates the number but does not hear the call. The scammer with a legitimately obtained number, the "bank" calling from a real mobile, the robocall that passes every signalling check, all still reach the subscriber.

Subscribers install third-party blocking apps, complain to the operator, and a growing share would pay the operator itself for protection that works on every call, not only the ones a reputation list has seen before. That is a value-added service with three properties operators rarely get together: high perceived value, a cost per unit the operator controls, and no new handset, SIM or app store dependency.

• A verdict on every call, in and out, minutes after hangup: block, review or allow, the category (bank impersonation, government or IT impersonation, robocall, pressure sales), and the red flags quoted from what was said. Delivered through the operator's app, SMS or push from the operator's own systems, fed by a signed webhook or the API.
• The next attempt stopped. A number scored above the operator's threshold is blocked or diverted before it rings the subscriber again, for an hour, a week or for good, with no action from anyone. Most voice scams take more than one call.
• Evidence on request. The recording, the transcript and a PDF report for a subscriber disputing a charge with a bank, or for the operator's fraud team.
• Optional, per number. Only enrolled numbers are routed through the node. Everyone else's traffic is untouched, so cost and consent apply only to the subscribers who enrolled.

The verdict engine is the same one carriers use on wholesale routes: a stereo recording with each party on its own channel, a speaker-labelled transcript, and an analysis that returns a probability from 0 to 100 with its evidence. The operator chooses the class of models behind it, and therefore the cost per call.

A worked example, not a forecast. The price is illustrative and the attach rate is the operator's to earn through bundling, default-on trials and the moment a subscriber gets their first verdict. The figures are gross add-on revenue before the operator's cost per analyzed call, which is agreed per licence and scales with call volume, not with subscriber count.

Two levers move the result more than the price. First, what share of enrolled numbers' calls the operator analyzes: every call, or a sample with the rest recorded and analyzed on demand, which the platform supports per route. Second, which model tier runs the analysis. The open-weight tier costs a fraction of a frontier model per call and returns the same verdict shape. Both are settings and need no engineering work.

Beyond the line item: fewer scam-call complaints to handle, a reason for a subscriber to stay, and a product the operator's own brand delivers rather than a third-party app.

Open Voice Shield is a SIP hop. In a mobile network it sits as a node the core routes enrolled numbers' calls through, on the way to and from the interconnect. An IMS core or a softswitch reaches it over SIP directly. A network still on SS7 signalling for the paths in question adds an SS7-to-SIP gateway node in front of it. The media follows the same path either way. Handsets, numbers and the core stay as they are.

Capacity is added by node, per region, and the platform already runs media relays and storage per region so a subscriber's recordings stay in-country. The operator can run the whole thing on its own hardware with an on-premises licence, with its own speech and analysis models and keys, so no call content leaves its network.

1. Route one cohort. A few thousand enrolled numbers through one node, verdicts into the operator's app or by SMS. Measure the attach rate the first verdict produces and the complaint volume on the cohort against a control group.
2. Set the threshold and the price. Block above the probability the fraud team is comfortable with. Price against the churn and handling cost the pilot shows.
3. Bundle. Default-on in a premium plan, opt-in elsewhere, a default-on trial for new activations. The moment a subscriber sees their first flagged call is the marketing.

Operator licences, node sizing and the cost per analyzed call are scoped with us. Send your subscriber count and signalling (SIP, SS7 or both) and you get a node plan and the numbers for your case back. Or read how the same hop works on a wholesale route and for a business phone line, and the three ways operators already use it on their own traffic.