Can detection be relied on as proof?

No. Detection produces a confidence signal, and generation methods change faster than detectors do. Treat the output as one input to an investigation rather than an answer.

Synthetic media and deepfake analysis

Synthetic media has become an impersonation problem rather than a novelty one: a convincing video or voice clip of an executive is a social-engineering payload, and it is aimed at the same people a look-alike domain is aimed at.

What this page checks

Facial consistency — whether the geometry and lighting of a face stay coherent across frames, which is where generated video most often falls apart.
Audio and video alignment — whether speech matches mouth movement, and whether the audio itself carries the artefacts of synthesis.
Compression and generation artefacts — traces left behind by the generation pipeline that survive re-encoding.
Temporal anomalies — discontinuities between frames that a real camera would not produce.

No detector returns a verdict. Every one of these signals produces a probability, and a probability is an input to a human decision — which is why the output of any such analysis belongs in an investigation, not in an automated block.

How it runs

This page is an interactive demonstration of the check rather than a live assessment of your estate — it shows the shape of the finding and how it is presented. For a real, continuous answer across every asset you own, the same check runs inside Maphra.

Where this sits in Maphra

Synthetic media is the same category of risk as the rest of Maphra's brand work: an attack against your customers and staff that never touches your network. Credential exposure in breach corpora, dark-web mentions, look-alike domains and impersonating apps — the attack paths that never touch your network. Look-alike domain detection, impersonating application discovery and continuous brand surveillance run alongside each other, because an impersonation campaign rarely uses only one of them.

Questions

Can detection be relied on as proof?: No. Detection produces a confidence signal, and generation methods change faster than detectors do. Treat the output as one input to an investigation rather than an answer.

A one-off check tells you about one asset today. Maphra runs this continuously across everything it discovers, and a walkthrough uses your own domain rather than a sample.