Overview

FARx is a proprietary machine learning platform designed to recognise humans the way humans recognise one another - continuously, naturally, and without friction. Its core mission is to protect organisations and their users against fraud while delivering a seamless authentication experience.

Unlike traditional authentication methods that require deliberate user action (such as entering a PIN or scanning a fingerprint on demand), FARx works passively and continuously throughout an entire session. It combines two of the most reliable biometric signals - voice and facial recognition - into a single, unified verification layer.

Multi-Factor Authentication Identity & Verification (MFA ID&V)

  • Banking: PSD2 Strong Customer Authentication (SCA) mandates multi-factor verification for online payments and account access across the EU and UK. FARx’s continuous biometric MFA delivers SCA compliance without adding friction to the customer journey.
  • Financial Services: FCA regulations and KYC/AML obligations require firms to continuously verify the identity of clients, particularly during high-value transactions or sensitive account operations. FARx provides an auditable, continuous biometric identity record throughout each session.
  • Healthcare: Access to patient records and clinical systems is governed by frameworks such as HIPAA (US), DSPT and NHS Digital standards (UK), and GDPR. These require robust identity assurance for anyone accessing sensitive health data, including practitioners conducting remote consultations.
  • Government: Digital government services and citizen-facing platforms are subject to identity assurance frameworks such as UK Gov.UK One Login, eIDAS (EU), and NIST SP 800-63 (US). These define levels of identity assurance that increasingly require biometric verification to reach the higher assurance tiers.
  • Fraud-prevalent applications: Insurance, telecommunications, online gaming, and e-commerce platforms face significant exposure to account takeover, identity fraud, and synthetic identity attacks. Even where formal mandates do not yet apply, regulators and industry bodies strongly recommend MFA ID&V as a baseline control. FARx enables these organisations to implement continuous, biometric-grade MFA with minimal integration effort.

By delivering MFA ID&V continuously and passively - through voice and facial biometrics rather than disruptive one-time challenges - FARx allows regulated organisations to satisfy their compliance obligations while simultaneously improving the user experience.

Mission

Protect against fraudsters using the same instinctive, multi-modal recognition that humans naturally apply to one another - in real time, without interrupting the user.

Core Technology

Proprietary Machine Learning Engine

Key characteristics of the ML engine include:

At the heart of FARx is a purpose-built machine learning engine trained to model the unique biometric signature of each individual user. Rather than relying on static template matching, the system builds a dynamic, evolving model of who the user is - adapting over time as natural variations occur in voice and appearance.

  • Multi-modal fusion: voice and facial biometrics are processed jointly, producing a confidence score that is more robust than either signal alone.
  • Continuous learning: the model refines itself during authenticated sessions, improving accuracy without additional user enrolment steps.
  • Liveness detection: built-in anti-spoofing mechanisms distinguish live users from recorded audio or static images.

Multi-Factor Biometric Recognition

FARx delivers multi-factor authentication (MFA) without requiring the user to take any deliberate action. The two factors - voice and face - operate simultaneously and continuously:

Voice Recognition

Analyses vocal characteristics including tone, cadence, pitch, and unique micro-patterns. Works even in noisy environments through adaptive signal processing.

Facial Recognition

Monitors facial geometry, micro-expressions, and presence in frame. Handles natural head movement and varied lighting conditions.

Behavioural Signals

Monitors facial geometry, micro-expressions, and presence in frame. Handles natural head movement and varied lighting conditions.

Behavioural Signals

Combines all active signals into a single, continuously updated trust score for the current session.

How It Works in a Session

Continuous Authentication

Traditional authentication validates a user once - at login. FARx re-authenticates continuously throughout the entire session. This means that even if a fraudster gains access to a valid session token, FARx will detect the biometric mismatch and act immediately.

The authentication loop runs silently in the background:

  • Biometric signals are captured at regular intervals via the device camera and microphone.
  • Each capture is scored against the enrolled user profile.
  • The session trust score is updated in real time.
  • If the score drops below a configurable threshold, a protective action is triggered automatically.

Real-Time Fraud Detection

FARx is designed to detect a broad range of fraudulent behaviours, including:

  • Session hijacking: a different person taking over an authenticated session.
  • Impersonation: an attacker attempting to mimic the enrolled user's voice or appearance.
  • Synthetic media attacks: deepfake audio or video used to spoof biometric checks.
  • Shared credentials: multiple users operating within a session intended for a single individual.

Protective Actions

When suspicious behaviour is detected, FARx can trigger a range of configurable protective responses without waiting for a human to intervene:

  • Screen blur: the application display is obscured, preventing data exposure while the system reassesses.
  • Session pause: interactive functionality is suspended pending re-verification.
  • Step-up challenge: the user is prompted for an explicit verification step (e.g. a spoken passphrase).
  • Session termination: the session is ended immediately for high-confidence fraud detections.
  • Alert dispatch: security or compliance teams are notified in real time via webhook or API event.
 Deployment & Integration

Infrastructure

FARx is packaged as a Docker image, making it straightforward to deploy across a range of environments:

Cloud Deployment

Deploy on any major cloud provider (AWS, Azure, GCP) using standard container orchestration tools such as Kubernetes or ECS.

On-Premises

Install FARx on dedicated on-premises infrastructure for organisations with strict data residency or regulatory requirements.

Hybrid

Operate core processing on-premises while leveraging cloud scalability for peak-load periods.

VM Compatible

Runs on any standard Virtual Machine - no specialised hardware or GPU is required for standard deployments.

REST API

FARx exposes a self-describing REST API, documented and explorable via a Swagger UI interface at: [insert API URL]

The API follows RESTful conventions and supports:

  • Enrolment: register a new user's biometric profile.
  • Session management: open, update, and close authenticated sessions.
  • Event webhooks: receive real-time notifications of authentication events and alerts.
  • Score queries: retrieve the current trust score for an active session.
  • Configuration: set per-tenant thresholds and protective action rules.

All API endpoints are versioned, secured with TLS, and require authenticated access tokens. The Swagger UI provides live interactive documentation, enabling rapid integration and testing by third-party developers.

Financial Services

Deploy on any major cloud provider (AWS, Azure, GCP) using standard container orchestration tools such as Kubernetes or ECS.

On-Premises

Install FARx on dedicated on-premises infrastructure for organisations with strict data residency or regulatory requirements.

Hybrid

Operate core processing on-premises while leveraging cloud scalability for peak-load periods.

VM Compatible

Runs on any standard Virtual Machine - no specialised hardware or GPU is required for standard deployments.

how it works

Key Use Cases

FARx is applicable across any sector where identity assurance and fraud prevention are critical:

How It Works Icon

Financial Services

Continuous authentication of customers during online banking sessions, reducing account takeover fraud without adding friction.

How It Works Icon

Contact Centres

Real-time voice biometric verification of callers, replacing knowledge-based authentication (KBA) questions.

How It Works Icon

Video Conferencing

Verify participants are who they claim to be throughout meetings - particularly important for sensitive negotiations or regulated discussions.

How It Works Icon

Actionable Insights

The integration of AI and ML is leading to the creation of intelligent systems

How It Works Icon

Government & Public Sector

Secure access to citizen-facing digital services with continuous biometric assurance.

How It Works Icon

Insurance

Detect fraudulent claimant impersonation during digital claims submissions or remote interviews.

Security & Privacy

FARx is designed with security and data privacy as foundational principles:

  • Biometric data is processed and stored in encrypted form; raw audio and video are not retained beyond the processing window.
  • On-premises deployment options ensure biometric data never leaves the organisation's own infrastructure.
  • The session trust score is updated in real time.
  • The system is designed to support compliance with GDPR, PSD2 Strong Customer Authentication (SCA), and sector-specific regulations.
  • Anti-spoofing and liveness detection prevent replay and synthetic media attacks.
  • All API communications are secured with TLS 1.2 or higher.

Synthetic Voice Detection

As AI-generated voice technology becomes increasingly sophisticated, the ability to reliably distinguish between a natural human voice and a synthetic one is a critical component of any robust fraud prevention platform. FARx includes a dedicated synthetic voice detection capability, providing an additional layer of assurance that the voice being presented is genuinely human in origin.

How the Model is Built

FARx is trained on extensive, curated databases containing both synthetic and natural voice samples. During training, the machine learning models learn to identify and internalise the acoustic characteristics that distinguish the two groups - patterns that are often imperceptible to the human ear but are reliably detectable at the signal level.

The trained models capture a range of distinguishing features across both categories:

  • Natural voice characteristics: organic micro-variations in pitch, breath patterns, vocal tract resonance, and natural disfluencies (e.g. hesitations, coarticulation effects).
  • Synthetic voice artefacts: subtle spectral inconsistencies, over-regularised prosody, generation model fingerprints, and compression or vocoding artefacts introduced during speech synthesis.
  • Channel and environment signals: acoustic room characteristics, background noise profiles, and microphone response patterns that natural and synthesised audio tend to exhibit differently.

Classification & Output

Given a voice sample, FARx classifies it as most likely synthetic or most likely natural, returning a calibrated probability score alongside the binary verdict. This score can be used by the integrating application to apply risk-based decision logic - for example, flagging borderline cases for human review or immediately terminating a session when confidence of synthetic origin is high.

Continuous Model Improvement

Synthetic voice generation technology evolves rapidly. New text-to-speech and voice cloning systems emerge regularly, each with distinct acoustic signatures that earlier models may not have encountered. FARx addresses this challenge through continuous model updating - a core feature of its ML/AI architecture:

Emerging threat tracking:

as new synthetic voice generation systems are identified, samples are incorporated into the training corpus and the models are retrained to recognise their signatures.

Active learning pipeline:

uncertain or novel classifications are flagged for expert review, with confirmed labels fed back into the training data to incrementally improve model performance.

Version-controlled deployment:

updated models are versioned and can be rolled out to deployed FARx instances without downtime, ensuring protection keeps pace with the threat landscape.

The Speech Quality API

Synthetic voice detection is exposed through the FARx Speech Quality API. This endpoint goes beyond a simple synthetic/natural binary check: it analyses a submitted voice sample against a range of quality attributes simultaneously, returning a comprehensive quality assessment in a single API call.

Quality attributes assessed by the Speech Quality API include:

  • Synthetic origin probability: the core classification score indicating the likelihood that the audio was generated by a text-to-speech or voice cloning system.
  • Signal-to-noise ratio (SNR): measures audio clarity relative to background noise, indicating whether the sample is of sufficient quality for reliable biometric analysis.
  • Clipping and saturation detection: identifies audio distortion caused by overloaded recording levels, which can mask or mimic synthetic artefacts.
  • Score queries: retrieve the current trust score for an active session.
  • Speech duration and coverage: verifies that sufficient active speech content is present in the sample for a confident assessment to be made.
  • Codec and compression artefacts: detects the presence of encoding degradation introduced through telephony codecs or audio compression, providing context that assists in interpreting the synthetic origin score.

By bundling synthetic detection alongside broader quality analysis, the Speech Quality API gives integrating applications the full context needed to make well-informed, risk-appropriate decisions about how to handle a given voice sample. Full API documentation and interactive testing are available at the FARx Swagger UI.

Get in touch with out expert team

Discover how FARx can strengthen identity assurance across your organisation.

Get in touch

Summary

FARx delivers the next generation of identity assurance: frictionless, continuous, and multi-modal. By combining voice and facial biometrics through a proprietary ML engine, it enables organisations to detect and respond to fraud in real time - without disrupting the user experience.

Its flexible deployment model (cloud, on-premises, or hybrid), self-describing REST API, and streaming FARx Broker make it straightforward to integrate into existing platforms and communication environments. Whether protecting a banking portal, a contact centre, or a video conferencing platform, FARx provides a robust and scalable biometric security layer.