FraudBuilt
Fraud hides in what people share.
A fraudster can forge a document. It is much harder to forge the device, the network, the rhythm of typing and the web of things one identity has in common with others. Aegis observes those signals at the verification surface and folds them into a graph where shared attributes become edges.
Simulation
Runs entirely in your browser.
Shared attributes become edges. Shapes become cases.
A device behind many applications, a face behind many names, and a decline at another organisation surfacing here — without revealing who.
Scenarios
Applications arrive
The reasoning
Several applications arrive from one device in a short window. How does the platform see it, and who decides what it means?
- ApplicationsEach application is a node. Nothing is suspicious yet; every value is sealed at capture and only its keyed index reaches the graph.
Signals
Observed at capture, sealed on arrival.
The capture edge reads device, browser, network and behavioural signals from a closed vocabulary. Every raw value is sealed in the vault, and only its keyed index travels on.
- Lies, not just facts
- Signals that contradict each other — a spoofed browser, a forged timezone — say more than any one signal alone. Fraud and evasion are scored separately.
- Built
- How, never what
- Typing rhythm, never the content typed. No precise location. No tracking across sites.
- Built
- A lawful basis first
- Nothing is collected without a recorded lawful basis.
- Built
- Assistive technology exempt
- Behavioural checks never penalise a person for using assistive technology.
- Built
The graph
Pure detectors, human verdicts.
The platform’s events are folded into a graph where every attribute two parties share is an edge. Detectors are pure functions over that graph; their thresholds are configuration, not code.
- Velocity and farms
- Device velocity, identity farms, bursts and unusual fan-in or fan-out.
- Built
- Reuse
- One face across several identities, reused artefacts, synthetic-identity shapes.
- Built
- Movement
- Impossible travel and SIM-swap recency.
- Built
- Shopping
- Declined at one organisation, approved at another; provider shopping; operator anomalies.
- Built
- Learned models
- Supervised models, once there is labelled data to learn from honestly.
- Planned
Across organisations
Presence, never counterparties.
Reads across organisations stay shut until the regulator’s prior authorisation is held. Even then, an organisation learns that something exists elsewhere — never which organisation, never who, never the record. And no detector designates anyone: only a person may mark an identity as fraudulent.
Build on identity that is proven, not assumed.
Aegis is built by Protoworks and offered to partners on request. Tell us what you need to verify, issue or accept, and we will show you the platform running.