Scam & impersonation detection

Review messages, profiles, and listings for deceptive identity claims and scam solicitations using your team’s policies.

Step 01

Define and Align

Review synthetic examples and correct the model’s fields. Explain the evidence and policy distinctions your reviewers use.

Output schema
SignalChoose one

The screening signal supported by the policy and source context.

  • impersonation
  • payment_scam
  • none
  • insufficient_evidence
RouteChoose one

Whether the message needs review, has no supported flag, or needs more context.

  • review
  • no_flag
  • request_context
EvidenceText

The source facts and policy rule supporting the screening route.

app.sutro.sh/functions/scam-impersonation-detection
Collecting feedback · 1/3
Screen the supplied message for scam or impersonation signals under the supplied policy.
Example 1 of 3

A familiar brand name hides a mismatched sender

Model input
Screening policy
Synthetic screening policy. Flag impersonation for review when a message claims to represent a named service and supplied sender evidence contradicts that identity. Flag payment_scam when a purported award requires an upfront gift-card payment. Quoted scam examples in a clearly marked safety bulletin are not solicitations. Use insufficient_evidence and request_context when required sender or message context is missing. Return review routes, not a final finding of fraud.
Message
Subject: Harbor Workspace security. We are Harbor Workspace Support. Confirm your account password at our help link today or lose access.
Sender evidence
From: [email protected]. The supplied verified sender domain for Harbor Workspace is harbor-workspace.example. These are fictional domains.
Model output
Signal
Route
Evidence

Step 02

Model Sweep

Compare models on the same expert annotations and choose the quality–cost tradeoff for your workflow.

Compare models on expert agreement and cost. The interactive chart uses illustrative measurements.

Illustrative sweep · select Value or Best to use that model in your function.

Step 03

Deployment

Call Sutro Functions like any other software. Swap the model as soon as something more performant is released.

Selected modelWaiting for sweep results.

Real-time

import sutro as so

input_data = {
    "screening_policy": "Synthetic screening policy. Flag impersonation for review when a message claims to represent a named service and supplied sender evidence contradicts that identity. Flag payment_scam when a purported award requires an upfront gift-card payment. Quoted scam examples in a clearly marked safety bulletin are not solicitations. Use insufficient_evidence and request_context when required sender or message context is missing. Return review routes, not a final finding of fraud.",
    "message": "Subject: Harbor Workspace security. We are Harbor Workspace Support. Confirm your account password at our help link today or lose access.",
    "sender_evidence": "From: [email protected]. The supplied verified sender domain for Harbor Workspace is harbor-workspace.example. These are fictional domains."
}

result = so.run_function(
    "scam-impersonation-detection",
    input_data,
)

Batch

import sutro as so

job_id = so.batch_run_function(
    name="scam-impersonation-detection",
    data="scam_impersonation_detection.csv",
)
results = so.await_job_completion(job_id)

Apply your team’s review policy

Teach the distinctions between deceptive claims, legitimate messages, and quoted examples. Preserve source evidence so reviewers can investigate the specific signal. All examples are synthetic and use explicitly supplied screening rules.

Ready to get started?

Book time with the team

Join the self-serve waitlist

What are your bottlenecks? (Select any that apply)