Real-Time Audio Deepfake Detection for Trustworthy Voice AI

CandyVoice develops multilingual real-time audio deepfake detection capable of analysing speech streams with or without a reference speaker.

The technology detects synthetic, cloned or manipulated voices in real time and can operate locally on edge devices or distributed infrastructures.

It is designed for trustworthy conversational AI, secure voice services and next-generation AI communication platforms.

CONTRIBUTION TO RELEVANT HORIZON EUROPE TOPICS

1. DIGITAL-EMERGING-52
Ensures trusted human-AI voice interaction by verifying authenticity of incoming speech.

Contribution:

• secure conversational AI

• trusted AI interfaces

• fraud prevention

• authentication support

2. DATA-03
Embedded deepfake detection running locally before AI inference.

Contribution:

• Edge AI security

• trustworthy embedded AI

• local inference

• secure speech processing

3. DATA-08
Distributed authenticity verification service deployed across Telco Edge infrastructures.

Contribution:

• telecom security services

• distributed trust services

• AI service protection

• voice fraud detection

-----------------------------------------------

KEYWORDS:
Deepfake Detection
Synthetic Speech Detection
Trustworthy AI
Responsible AI
Voice Security
AI Safety
Fraud Detection
Conversational AI
Edge AI
Distributed AI
Telco Edge
Cybersecurity