> ## Documentation Index
> Fetch the complete documentation index at: https://docs.dataspike.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

> AI-powered media authenticity for images, documents, video, and audio — fast, accurate, and privacy-first.

The **Dataspike Deepfake Detection** system safeguards high-trust workflows by identifying AI-generated and manipulated media across **images**, **documents/IDs**, **videos**, and **audio**.\
It’s built for identity verification, KYC/AML, and digital onboarding where authenticity and compliance are critical.

***

## What it is

Dataspike combines **neural network ensembles**, **forensic signal analysis**, and **AI signature tracing** to detect:

* **Selfie & liveness manipulation** (face swaps, morphing, compositing)
* **Fake or tampered documents/IDs** (edited fields, forged MRZ/barcodes, altered holograms/laminates, reprint/recapture)
* **Video deepfakes** (synthetic face swaps, recomposed scenes)
* **Audio deepfakes** (AI-generated/cloned voices, splice & resynthesis artifacts)

Models are continuously updated to keep pace with new generators (e.g., Veo, Sana, Flux) and emerging attack patterns.

***

## Why it’s powerful

* **Multi-modal analysis** — Supports evaluation of images, identity documents, video, and audio files.
* **Robust detection models** — Designed to operate reliably across varied resolutions, compression levels, and device types.
* **Data-minimization approach** — Media is processed transiently and not stored after analysis.
* **Structured output** — Each analysis produces a consistent response schema with scores and indicators suitable for automation.
* **Flexible deployment** — Available through public API, private cloud, or on-premise environments.

***

## Why it’s fast

* **Optimized inference pipeline** — Uses asynchronous processing and batched GPU workloads to maintain high throughput.
* **Adaptive frame selection** — For video streams, only key frames are analyzed to balance accuracy and latency.
* **Lightweight result delivery** — Responses include compact, pre-aggregated indicators for real-time handling.
* **Edge and mobile support** — Inference components can run locally for latency-sensitive

***

## Ecosystem & Tools

* **📱 Mobile SDK (Android)** — Provides on-device video analysis without requiring API calls.\
  Ideal for mobile verification or offline scenarios where low latency is critical.
  * [View SDK on GitHub](https://github.com/dataspike-io/Deepfake-MobileSDK-Android)
  * [Download Demo App (APK)](https://github.com/dataspike-io/Deepfake-MobileSDK-Android/releases/download/v1.0.0/deepfake-demo-app-1.0.0.apk)

* **🎥 LiveKit Integration** — Enables real-time deepfake detection in live video sessions.\
  The integration performs adaptive frame sampling and analysis directly within LiveKit rooms, providing detection results with minimal latency and no persistent storage.

***

## Summary

Dataspike provides **real-time authenticity intelligence** for selfies, **documents/IDs**, video, and **audio**, helping you block synthetic identities, detect voice clones, and catch forged documents — without sacrificing performance or privacy.