Sober & Transparent

Our Forensic Logic.

Veridex does not provide "truth" as a binary service. We provide a rigorous framework for decomposing digital assets into probabilistic forensic markers.

The Veridex Stance: Assistive, Not Final.

In a high-stakes legal, journalistic, or research environment, automated tools should be used for screening and signal detection, not final adjudication. Our methodology is designed to augment expert human judgment by surfacing anomalies that are invisible to the naked eye/ear.

Read the Limitations DocRequest a Technical Walkthrough

Forensic Domains

Multimodal Synthetic Analysis

Our primary detection layer focuses on identifying artifacts unique to generative models across frequency and spatial domains.

View Limitations
Technical Logic
  • 01Spectral anomaly detection in audio (e.g., GAN voice artifacts).
  • 02Error Level Analysis (ELA) for image compression inconsistencies.
  • 03Temporal coherence checks for video (frame-to-frame stability).
  • 04Frequency domain pattern identification in synthesized pixels.

Claim Decomposition Heuristics

Content is broken into 'Atomic Truth Claims' using NLP models to assess the factual density and source correlation.

View Limitations
Technical Logic
  • 01Named Entity Recognition (NER) to isolate actors and events.
  • 02Cross-referencing against primary news agency archives (AP/Reuters).
  • 03Temporal consistency audits (Checking dates vs event histories).
  • 04Source-citation validation through automated web-provenance.

Rhetorical Bias Mapping

We identify manipulation mechanisms by analyzing the linguistic structure and emotional variance of the text.

View Limitations
Technical Logic
  • 01Mapping of 'Propaganda Heuristics' (e.g., loaded language, fear appeals).
  • 02Sentiment variance analysis for synthetic text detection.
  • 03Correlation of rhetorical patterns with known disinformation playbooks.
  • 04Complexity scoring to flag machine-generated repetitive structures.

Copyright & Origin Logic

We assess the risk of training-data leakage and verbatim overlap with known copyrighted or trademarked materials.

View Limitations
Technical Logic
  • 01Embedding-based similarity threshold checks against open data sets.
  • 02Fragmented match evidence for LLM 'leakage' identification.
  • 03SHA-256 ledger registration for every audit artifact.
  • 04Source-correlating copyright risks to specific platform training sets.

Understanding the "Verity Index"

The Verity Index (0.0 to 1.0) is a weighted probabilistic score. It aggregated signals from all active modules. A low score (e.g., 0.2) does not definitively prove "falsehood," but indicates a high density of synthetic artifacts, factual inconsistencies, or rhetorical manipulation markers that require investigation.

0.0 - 0.4
High Forensic Risk
0.4 - 0.7
Intermediate Warning
0.7 - 1.0
Procedural Verity