Who called me with a claim that my identity had been stolen? That question usually gets handled as a consumer worry. It maps more cleanly onto a problem security teams already know: an alert arrives from an outside party you cannot authenticate, and the alert asks you to engage with it on its own terms.

Affiliation and AI note: This is promotional editorial from the website whocallme.co.uk, drafted with AI assistance. The playbook and snippet below are an illustrative manual design. They are not a deployed product or a study of real incidents, and the author did not take or reproduce the call described.

The source material

Our page for 0113 328 1099 lists a landline with an allocation hint of WAVECREST NETWORKS LIMITED and a numbering area of Leeds. Those labels describe a number range. They do not identify the caller or where a call came from. When fetched for this post, the page showed 13 searches and one user report, with a “first seen: 5 days ago” page label. That label is not a timestamp for the note.

The note is imported, its original publication date is unknown, and the visitor filed it as Scam/Fraud. That category is the visitor’s label, not an established finding about the number holder. Caller ID can be spoofed.

In the visitor’s words, the callers said they “knows information about me”, that “somone stole my ID and sensitive information and took loans on me”, and that the visitor was “in a huge debts now”. The visitor declined “Without confirming any information they has” and said they would not speak to anyone. The caller “dropped the call by themselves”. The visitor added a view of their own: this kind of debt and credit problem is not “resolved on phone call”.

Everything below works from that text. No outcome beyond it is assumed.

Model it as an unauthenticated alert

Strip the scenario down and you get a familiar shape:

  • An external party asserts that a compromise has happened.
  • The party claims to hold data about you as evidence.
  • The channel was opened by them, so you cannot authenticate them.
  • The implied next step is to engage now, on that channel.

Security teams see the same shape in cold emails reporting “your credentials are on a leak site”, in unsolicited bug reports demanding payment, and in fake abuse notices. Mature handling has a few consistent rules. The alert is treated as a lead, not as a fact. Its claims are corroborated with your own telemetry. Any response happens over channels you choose.

Confirmation is the payload

The visitor’s decision not to confirm anything is the most important line in the note, and it deserves a precise explanation.

A caller who says “we know information about you” is often inviting a correction or a confirmation. Every “yes, that’s my address” or “no, my date of birth is…” turns a guess into a verified record. In security terms, you become an oracle: the target validates the attacker’s data for free. That holds whether the caller has a real dataset, a partial one, or nothing at all.

So the safe default is not to test what they know. Do not ask them to read your details out “to prove it”, either. Hearing the right answer proves only that the data exists somewhere. It does not show that the speaker is entitled to it, and the conversation it starts usually leads straight back to confirmation.

Corroborate with your own telemetry

For an organisation, telemetry means logs, sign-in records and alerts from systems you operate. For an individual, the equivalent sources are records you can reach without the caller:

  • your credit file, requested directly from the UK credit reference agencies;
  • statements and letters from banks and lenders you already deal with;
  • account security pages reached by typing the address yourself.

If none of those show unfamiliar credit, the alert stays uncorroborated. If they do, you now have an incident with a first-party source, and you can act on it through official routes. In neither case does the original caller become part of the remediation path.

A minimal playbook

Here is the handling written as a small YAML playbook. It is a template to adapt, not a standard.

playbook: inbound-compromise-claim trigger: channel: inbound_voice # or email, sms, chat  initiated_by: external_party claim: "identity stolen / credit taken in your name" rules: - id: no-oracle action: do_not_confirm_or_correct_any_personal_data - id: no-proof-request action: do_not_ask_caller_to_recite_your_data - id: end-channel action: end_conversation_without_commitment - id: corroborate sources: [credit_file, own_bank_and_lender_records, account_security_pages] access: self_initiated_only disposition: uncorroborated: record_and_close corroborated: open_incident_via_official_channels record_fields: [date_reviewed, number_displayed, claim_text, data_caller_offered, data_you_disclosed, sources_checked, disposition] 

And a tiny triage function that makes the disposition rule explicit. It uses only the standard library.

from dataclasses import dataclass, field @dataclass class InboundClaim: claim_text: str initiated_by_external: bool = True data_disclosed_by_you: list = field(default_factory=list) corroborating_sources: list = field(default_factory=list) def disposition(c: InboundClaim) -> str: if c.data_disclosed_by_you: return "review-exposure: you confirmed data; check accounts via your own channels" if c.corroborating_sources: return "open-incident: corroborated by first-party records" if c.initiated_by_external: return "record-and-close: unauthenticated and uncorroborated" return "manual-review" note = InboundClaim(claim_text="ID stolen, loans taken (visitor note)") print(disposition(note)) # record-and-close: unauthenticated and uncorroborated 

The ordering matters. Disclosure is checked first, because once data has left your hands the question shifts from “is this real?” to “what did I expose?” Corroboration comes second. An external claim with no corroboration and no disclosure closes cleanly.

Applied to the public note, the visitor disclosed nothing, the caller ended the call, and no first-party check is described. That sits squarely in the record-and-close branch. Whether the visitor’s credit file was clean is unknown, and this post does not guess.

Record fields, kept honest

When you file a note like this, keep the fields separate. claim_text holds what the caller said, attributed. data_caller_offered holds anything they recited, but do not write your own correct values beside it. data_you_disclosed should usually read “none”. date_reviewed is the day you filled the record. It is not the page’s first-seen label and not a guessed call date.

Two things do not belong in the record: speculation about who the caller “really” was, and the number holder’s identity. The page does not establish either.

Where a lookup fits

A lookup is enrichment, not authentication. The anonymous note about a stolen-ID call can tell you whether someone else described a similar pitch from the same displayed number. That is useful context for the claim_text field and nothing more.

If you keep running into this pattern, a UK site that indexes visitor reports by number can act as a quick reference before you write a record. It does not bind a voice on the line to any lender or agency. The rule stays the same: treat the claim as a lead, confirm nothing, and corroborate only through channels you opened yourself.