liveBook · Manning

Selfie check

As explained in Fight Fraud with Machine Learning

A selfie check is an automated verification procedure that analyzes a submitted selfie for quality, fraud, or user-eligibility conditions.

A selfie check automates visual KYC screening by analyzing a submitted selfie for readability, tampering, and user-eligibility violations. It operates alongside document checks and selfie–ID face checks in a broader ML-based verification architecture. section 7.1.3

Selfie checks divide into two practical groups:

  • Quality checks verify that the face is visible and the image is usable, including brightness, blur, glare, rotation, and face occlusion. section 7.1.3
  • Fraud checks detect manipulation such as deepfakes or photographing a photo, while also screening duplicate sign-ups, under-age users, and gender mismatches. section 7.1.3

Failed selfie submissions

A selfie can fail because glare obscures the face, the image is rotated or blurry, the face is covered, the submission is a deepfake, or it depicts a photograph rather than a live person. section 7.1.3

How Fight Fraud with Machine Learning covers it

How deep it goes
Distinguishes quality from fraud checks and lists failure conditions, but does not detail their individual model architectures or training procedures.
Examples
Uses brightness, blur, glare, rotation, occlusion, deepfakes, photo-of-photo attacks, duplicate sign-ups, age, and gender mismatch.
What it assumes
Assumes a selfie-based KYC context and distinguishes selfie, ID, and textual user data as separate verification inputs.
What you can do after
Evaluate selfie submissions against image-quality, tampering, and eligibility conditions within an automated KYC workflow.
How it connects
Connects quality checks and fraud checks with document checks, selfie–ID face checks, and face matching.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Fight Fraud with Machine Learning. The section links point to the parts of the book it draws on.