Media integrity scanner
How uploaded logos, screenshots and feature videos are decoded and measured before code security is allowed to run.
System flow
Read actual stored bytes
Decode images/video
Measure pixels/motion/audio
OCR selected frames
Correlate package + metadata
Create report
Source map
- lib/cdn/apps/media-security/app-media-analyzer.ts
- components/cdn/apps/AppMediaIntegrityReport.tsx
Real image and video analysis
Images are decoded with Sharp and measured using dimensions, luminance, contrast, entropy, edge activity, blank ratios, perceptual hashes, histograms and visual vectors.
Videos are inspected with ffprobe and decoded with ffmpeg. The scanner samples real PNG frames, measures localized motion, black/blank conditions, frozen transitions and audio levels.
OCR and app relevance
Tesseract OCR is used on relevant image/video samples. OCR is only one signal: the scanner also correlates visual behavior, package implementation, metadata claims and runtime features.
No fake “video quality” decision
A visible app with small UI changes should not be treated as frozen simply because most screen pixels remain static. Localized tile/pixel activity is part of the motion decision so calculator/editor-style apps can pass when they show real interaction.
Silence alone is not a failure. A frozen visual combined with effective silence can produce the static-placeholder finding.
Report evidence
Findings are meant to include measurable evidence and a specific correction. A blocked result remains visible to the publisher and does not become an immutable public release.