Stratir case file STR-2025-0847

Project Cerberus

Multi-platform romance fraud network attribution, mapped into a reviewable evidence package.

Designed by Stratir
The Bahamas & North America

Cerberus shows how messy victim reports can become source-backed intelligence: scoped, preserved, correlated, and ready for escalation.

Romance fraud is not only a story problem. It is an infrastructure problem.

The case began when concerned individuals reached out after family members were targeted by romance scams linked to West African operating patterns. They had scattered clues, emotional urgency, and no clear way to understand whether the accounts were isolated or coordinated.

Stratir treated the inquiry as an attribution and evidence problem: preserve the original observations, resolve selectors across platforms, map the network, label confidence, and package the findings for people who needed to act.

The investigation produced a usable map, not a pile of screenshots.

17

Connected fraudulent accounts

Profiles were linked through repeated imagery, behavioral overlap, and shared operational patterns.

72h

Investigation window

Collection, correlation, review, and client delivery were completed inside a three-day sprint.

5+

Platforms mapped

The inquiry moved across social surfaces, archive traces, public records, and supporting selector context.

47

Profile captures preserved

Screenshots and source observations were retained so the final findings could be reviewed and escalated.

Project Cerberus findings

Romance-fraud reports became reviewable intelligence.

Project Cerberus began with scattered family concerns, suspicious profiles, and repeated romance-fraud behavior across platforms. Stratir turned that uncertainty into a structured intelligence package.

The work connected profile imagery, account behavior, communication patterns, payment clues, and source captures so readers could understand how the operation worked and what evidence supported each conclusion.

Victim reports preserved

Screenshots, profile links, chat context, and source observations were captured before accounts could be renamed, deleted, or hidden.

Persona clusters mapped

Repeated faces, handles, scripts, timing patterns, and payment references were correlated into a clear view of related accounts.

Confidence clearly labeled

Observed facts, high-confidence links, and analytic judgments were separated so every reader could see what was known and why.

Escalation made practical

The final package gave families, platforms, and investigators the context needed to report, review, and act without starting over.

Attribution was built as a chain, not a leap.

01

Family concern becomes a requirement

Clients arrived with names, social profiles, and concern for relatives being targeted. The first task was to convert worry into a scoped intelligence requirement.

02

Profile images become selectors

Repeated visual identity material was treated as a selector. Face-image correlation helped expose accounts that appeared separate to victims but linked in the evidence record.

03

Accounts become a network

Profiles, handles, payment hints, communication threads, and timing patterns were mapped as relationships rather than isolated screenshots.

04

Findings become an evidence package

The final output separated observed facts from analytic judgments and gave clients a structured package for platform reporting and law-enforcement engagement.

Signals that made the network visible.

01

Shared profile imagery across multiple personas

02

Central hub account behavior managing peripheral identities

03

Reusable payment and communication infrastructure

04

Timing overlap between outreach, grooming, and payment pressure

05

Common language patterns across supposedly separate accounts

06

Account clusters that converged around the same victim pathway

Evidence was classified by type and confidence.

Profile screenshots

47

Verified

Communication threads

12

Verified

Payment references

3

High confidence

Face-image matches

17

High confidence

Device or behavior clues

2 clusters

Medium confidence

02 / 72-hour rhythm

Fast work still needs gates.

01

0 to 24 hours

Collect and preserve

Initial targets were recorded, profile material was captured, and early account links were preserved before content could be deleted or renamed.

02

24 to 48 hours

Correlate and map

Image reuse, handles, communication artifacts, payment clues, and platform overlap were connected into a network model.

03

48 to 72 hours

Review and deliver

Findings were checked, confidence was labeled, and the evidence package was prepared for client action.

Case outcome

The network was documented clearly enough for clients to act without becoming investigators themselves.

Cerberus demonstrates Stratir's practical value in fraud intelligence: reduce uncertainty, preserve the record, expose coordination, and turn scattered public signals into a defensible evidence package.