Understanding TerraLink
Learn how our voice biometrics and AI agent authentication platform protects against deepfakes and ensures identity integrity.
What is TerraLink?
In a Nutshell
In an age where AI can clone voices in seconds and chatbots can impersonate humans, traditional authentication is broken. TerraLink solves this with a dual-layer approach:
Voice Biometrics
Authenticate humans using unique vocal characteristics that can't be replicated by AI voice cloning tools like ElevenLabs or VALL-E.
AI Agent Signatures
Track and verify AI agent behavior using spectral fingerprints, detecting when an AI deviates from its expected patterns (drift).
How It Works
The TerraLink Flow
- Enrollment: User records a voice sample (minimum 3 seconds)
- Feature Extraction: Wav2Vec2 extracts 768-dimensional embeddings
- Spoof Detection: RawNet3 analyzes for AI-generated artifacts
- Verification: Challenge-response prevents replay attacks
- Decision: Real-time ACCEPT/REJECT with confidence scores
Voice Authentication Pipeline
Challenge-Response System
Each verification request includes a unique challenge nonce - a random phrase the user must speak. This prevents:
- Replay attacks: Old recordings can't be reused
- Pre-recorded fraud: Attackers can't prepare in advance
- Deepfake synthesis: Real-time generation is detectable
The .mhz Protocol
What is .mhz?
Harmonic Signatures
Each AI agent has a unique spectral fingerprint based on its behavioral traits: Logic (14.3Hz), Empathy (20.8Hz), Creativity (27.3Hz), and Assertiveness (33.8Hz).
Drift Detection
Monitors AI responses for deviations from the registered signature. High drift scores indicate potential tampering, jailbreaking, or model replacement.
Dissonance Analysis
Detects anomalous frequency patterns that indicate the AI is behaving inconsistently with its registered profile.
Entropy Binding
In Unified mode, human voice and AI agent signatures are cryptographically bound together, creating unforgeable proof of human-AI collaboration.
Drift Thresholds
Key Benefits
Deepfake Resistant
RNA-wrapped biometrics detect ElevenLabs, VALL-E, Resemble AI, and other voice synthesis attacks with 99.2% accuracy.
Real-time Processing
Sub-100ms verification latency ensures frictionless user experiences without security compromises.
Replay Attack Prevention
Time-limited challenge nonces and entropy commitments make recorded audio attacks impossible.
AI Agent Accountability
Track which AI performed which action, detect model swaps, and ensure AI compliance with organizational policies.
Behavioral Continuity
Drift monitoring ensures AI agents maintain consistent behavior over time, catching gradual manipulation attempts.
Unified Identity Proof
Cryptographic binding of human + AI creates audit trails that prove authorization for high-stakes actions.
Use Cases
Industry Applications
Banking & Financial Services
Telecom & Contact Centers
E-Commerce & Payments
Healthcare & Insurance
Getting Started
Quick Start Guide
Follow these steps to integrate TerraLink into your application:
Get Your API Key
Register for a free account and generate your API key from the dashboard.
Register NowEnroll Voice Identities
Collect a 3+ second voice sample from users during onboarding.
POST /v1/identities { identityId, audio }Verify on Demand
Challenge users to speak a phrase and verify their identity.
POST /v1/voices/verify { identityId, audio, challenge_nonce }Register AI Agents (Optional)
Create .mhz signatures for your AI agents to track their behavior.
POST /v1/agents { agentId, model, systemPrompt }Monitor & Analyze
Use the dashboard to monitor drift, view analytics, and manage identities.
View DashboardReady to Secure Your Platform?
Join leading enterprises protecting their customers from AI fraud with TerraLink's unified identity verification.