Trust Score Explained: How Vashix Rates Photo Integrity from 0 to 100

When a photo arrives in your claims system or inspection dashboard, the first question is always the same: can I trust this? Is it a genuine capture from a real device at the right location and time? Or is it recycled, edited, spoofed, or generated by AI?

The Vashix Trust Score answers that question with a single number between 0 and 100. No ambiguity, no subjective review — a deterministic score computed from verifiable cryptographic and sensor signals. This article explains exactly how it works, what goes into the calculation, and what the score tiers mean for your verification workflows.

What Is the Trust Score?

The Trust Score is a number from 0 to 100 assigned to every photo captured through the Vashix system. It's computed deterministically — the same inputs always produce the same score. There's no machine learning model making probabilistic guesses. The score is the sum of verified signals minus fraud penalties.

The score answers five questions about the photo:

The Scoring Formula

The Trust Score starts at 0 and adds points for each verified signal. The maximum possible score is 100.

Important: The score can also go down from fraud penalties. If the system detects that a photo is an exact duplicate of a previous submission, 40 points are deducted. Failed liveness checks deduct 20 points. Invalid watermarks deduct 15. These penalties apply on top of whatever positive score the signals earned.

The Four Tiers

The raw 0-100 score maps to four tiers. These tiers determine how your organization should treat the photo in your workflow:

Why Deterministic, Not AI?

A common question: why not use machine learning to determine if a photo is genuine?

The answer is auditability. An ML model produces a confidence score based on learned patterns — but you can't explain exactly why it gave 87% instead of 92%. When you're making decisions about insurance claim payouts, legal evidence, or regulatory compliance, you need to be able to point to the exact signals that justify the score.

The Vashix Trust Score is fully decomposable. For any score, you can see: "This photo scored 84 because it received 30/40 for TEE attestation, 18/20 for GPS accuracy, 15/15 for nonce freshness, 11/15 for motion seal, and 10/10 for workflow binding." No black box. Every point is traceable to a verifiable signal.

This doesn't mean ML has no role. Server-side fraud detection uses pattern analysis for duplicate detection and geo-anomaly flagging. But the core Trust Score — the number your team acts on — is deterministic and explainable.

Why Web Uploads Can Never Score Above 38

Browsers have no access to hardware security chips, no IMU sensor data, and GPS can be spoofed in seconds via developer tools. Without hardware attestation (40 points gone), without real IMU motion seal (15 points gone), and with GPS capped at 4 points, the maximum theoretical score for a web upload is 38/100.

This ceiling is intentional and permanent. It's not a limitation we plan to "fix" — it reflects the fundamental security difference between a browser and a hardware-attested mobile device. Web uploads are useful for basic analysis (EXIF parsing, duplicate detection, AI content checks), but they can never reach the verified threshold.

Setting Your Own Thresholds

Different workflows have different risk tolerances. A routine property inspection might accept SILVER (60+) captures. A high-value motor insurance claim might require GOLD (90+). A quick delivery proof might accept BRONZE if the amount is low.

Vashix lets you set minimum tier requirements per session. When creating a capture session, you specify the minimum acceptable tier. If a capture comes in below that threshold, the dashboard flags it for review rather than auto-accepting.

What the Score Doesn't Tell You

The Trust Score verifies evidence integrity — that the photo was taken on a real device, at a real location, at a specific time, and hasn't been modified. It does not verify the content of the photo. A GOLD-scored photo proves that a specific device captured a specific scene at a specific location and time. It doesn't prove that the damage in the photo is new, or that the property in the photo is the right one.

Photo verification addresses the evidence chain. Claim validation still requires human judgment. The Trust Score removes the evidence manipulation problem so your team can focus on the business decision.