Product 02 / 02  ·  Case study

See through digital deception.

AI-powered forensics that verifies images, videos, screenshots, and documents in seconds — built for the people AI manipulation actually targets, not just cybersecurity labs.

● Early access CV forensics · ML · template matching Built solo
Scroll to explore

Sora, VEO, Midjourney, and watermark-removal tools have made manipulation instant and invisible. Existing AI detectors only catch the narrow slice they were trained for — AI faces, deepfake audio, generated images. They miss the content that actually does damage in ordinary life: a forged UPI payment slip, an edited WhatsApp screenshot, a fabricated school notice.

The world built tools that can generate anything. Nobody built the tool that can verify anything.

The world now has tools that can generate anything — but no tool that can verify anything.

TrustLens closes that gap with a multilayer forensic engine — classical vision forensics, AI artifact detection, structural document analysis, and origin-path reconstruction — built for students, teachers, small sellers, and anyone who's ever squinted at a screenshot wondering if it's real.

01 — Real or fake, in seconds
SCAN
SCAN

Authenticity score plus anomaly highlights — lighting, shadows, pixel artifacts, text inconsistencies.

01
VISION FORENSICS ENGINE4 checks
Error Level Analysis
Measures compression difference to isolate edited regions
Chromatic aberration check
Real camera lenses fringe slightly — edits break the pattern
FFT texture analysis
Catches repeating AI-style textures in the frequency domain
Illumination field estimation
Checks whether lighting and shadows match physical reality
2–5starget latency per image
≤10%target false-positive rate
6–12sper short video clip

Most detectors stop at a real/fake label. This is the layer underneath it — the evidence, not just the verdict.

02 — How was this actually made?
TRACE
TRACE

Reality Trace reconstructs the most probable creation history of a file, step by step.

02
REALITY TRACE · ORIGIN-PATH RECONSTRUCTIONexample
Camera
Original capture, EXIF-consistent
Edit
Compression-layer jump detected
AI pass
Diffusion fingerprint present
Resave → screenshot
Interpolation artifacts, screenshot lineage confirmed
7 signalsediting + diffusion + pixel-pattern fingerprints
Not binarya pathway, not a real/fake coin-flip

"AI or not" is the wrong question for most real-world fakes — most damage comes from content that's part real, part edited, part AI-touched-up. Reality Trace answers the actual question.

03 — What happens if this is trusted?
RISK
RISK

TruthScore+ rates real-world consequence, not just authenticity.

03
TRUTHSCORE+ · RISK & CONSEQUENCE4 examples
Fake UPI payment slip
High · Financial
Fake school notice
Medium · Institutional
AI-generated video
High · Reputation
Fake chat screenshot
Medium · Social
Low / Med / Highrisk output, not a raw score
Context-weightedcategory, financial impact, known scam patterns

A fake UPI slip and a fake meme aren't the same threat. TruthScore+ explains the danger, not just whether something's doctored.

04 — Does this exist anywhere else?
SHIELD
SHIELD

Source Match Shield cross-checks templates, formats, and public sources before anything gets trusted.

04
SOURCE MATCH SHIELD · CROSS-VERIFICATION5 checks
Perceptual hashing
pHash similarity against known content
Reverse image / video search
Checks for prior public appearance
Template matching
UPI slips, notices, receipts against reference formats
OCR keyword indexing
Cross-references extracted text against known formats
20+reference templates targeted for MVP
No match → flaggedfails closed, not open

If content can't be matched to anything real, it doesn't get the benefit of the doubt.

TrustLens isn't positioned as another AI detector — it's the category underneath: classical forensic imaging, statistical analysis, AI artifact detection, template matching, and risk scoring, brought together for people who aren't cybersecurity researchers.

The world is busy building swords. TrustLens is building the shield.

Digital forensics,
for everyone.

TrustLens is in early access. The core forensic engine is being built layer by layer — this case study tracks the real architecture, not a pitch.

Try the MVP ↗ ← Back to portfolio