EchoDepth: Emotion Recognition AI for Defence & Security detecting stress, deception, fatigue & insider risk in real time
EchoDepth analyses 44 facial Action Units in real time — detecting stress and deception, cognitive overload and operator fatigue. No wearables. No contact. Works on your existing cameras.
Every defence system monitors everything
except the human.
Networks, devices, and perimeters are covered. The cognitive and emotional state of your people — the variable most likely to precede an incident — is entirely invisible.
You Cannot Vet What You Cannot See
Security clearances and annual reviews catch nothing in real time. Stress indicators, behavioural anomalies, and emotional changes that precede incidents go entirely undetected by existing TSCM and access control infrastructure.
// COST PER INSIDER INCIDENT: £3.2M AVERAGE
Completion Rates Measure Nothing
74% of compliance failures stem from poor knowledge retention — not skipped training, but training that was attended while disengaged. You have no measure of whether it landed.
// 74% OF COMPLIANCE FAILURES: KNOWLEDGE RETENTION, NOT ATTENDANCE
Fatigued Operators Make Critical Errors
Drone pilots, SOC analysts, control room operators, and intelligence reviewers face sustained cognitive load with no real-time monitoring. Fatigue is invisible until an incident occurs.
// FATIGUE FACTOR IN 23% OF MILITARY NEAR-MISSES — NATO REVIEW
Polygraph Is Pseudoscience. You Know It.
Courts and procurement officers are aware that polygraph lacks scientific validity. Border screening, interview assessment, and source credibility require a defensible, evidence-based alternative.
// POLYGRAPH FALSE NEGATIVE RATE: UP TO 47% IN CONTROLLED STUDIES
Four Vectors. One Engine.
A single camera-based AI platform covering the four critical human intelligence gaps in defence and security operations. View all solution areas →
Know if Training is Actually Landing
Real-time engagement and comprehension scoring during mandatory training. DSAT-compatible audit records. LMS integration. Objective evidence that training is retained, not just completed.
Explore Solution → 02 Insider Threat & VettingContinuous Emotional Anomaly Monitoring
Behavioural baseline profiling per individual. Anomaly scoring when patterns deviate. SIEM integration. Earlier signal detection for earlier intervention before incidents occur.
Explore Solution → 03 Operator ReadinessQuantify Human Readiness. Live.
Pre-mission readiness scoring, live fatigue and cognitive load monitoring, and post-incident timeline reconstruction for UAS, SOC, and control room operators.
Explore Solution → 04 Deception DetectionA Defensible Alternative to Polygraph
44 FACS-compliant Action Units per frame. Timestamped per-question stress and suppression mapping. Structured analytical output for intelligence and legal review.
Explore Solution →Facial Action Units.
Decoded at Scale.
Standard RGB camera. 68-point facial landmark detection. 44 FACS-compliant Action Units extracted per frame. VAD mapping. REST API or on-premise deployment. Works with your existing camera infrastructure — no specialist hardware required.
Full Technical Overview →Processing pipeline · API & integration · Deployment options
🔒 Air-Gap Ready
Full on-premise Docker deployment. SCIF-compatible. Zero external data transmission at any stage.
⚡ ~700ms Latency
Sub-second emotional state output. Suitable for live interview and real-time operator alerting.
🌐 API-First
REST & WebSocket. Integrates with SIEM, LMS, C2. SDK: Python and Node.js.
🇬🇧 UK Data Residency
All data processed within UK borders. UK GDPR compliant. ISO 27001 pathway.
What procurement teams ask first.
What is EchoDepth, and how does it work?+
EchoDepth is a camera-based emotion recognition AI platform developed by Cavefish AI in Cardiff, Wales. It analyses 44 FACS-compliant facial Action Units per frame in real time — detecting stress, deception, cognitive overload, and operator fatigue. It requires only a standard RGB camera and deploys fully on-premise with no cloud dependency.
Is EchoDepth a genuine polygraph alternative?+
Yes. The National Academy of Sciences found in 2003 that polygraph lacks scientific validity for security screening, with false negative rates up to 47%. EchoDepth uses 44 peer-reviewed FACS Action Units to produce timestamped, structured analytical output — a documented, evidence-based record that withstands legal and procurement scrutiny. It does not claim to detect lies; it surfaces involuntary physiological markers correlated with deception attempts. See the full credibility assessment capability overview.
Can EchoDepth operate in a SCIF or air-gapped environment?+
Yes. EchoDepth is fully deployable on-premise via Docker with zero external data transmission at any stage. No cloud dependency, no telemetry, no outbound network calls. All inference runs locally on standard server hardware. See full deployment specifications or request an air-gapped demo via the briefing request form.
What hardware is required?+
EchoDepth requires only a standard RGB camera at minimum 720p. Existing CCTV, interview room cameras, or laptop webcams qualify. No infrared, no thermal imaging, and no specialist hardware is required. This makes it viable in interview settings, training environments, and control rooms where attaching physical sensors is impractical.
Is EchoDepth compliant with UK GDPR and defence data standards?+
EchoDepth processes biometric data with pseudonymisation by default, role-based access controls, and full audit logging. All data is processed within UK borders as standard. The platform is aligned with NCSC principles, on the ISO 27001 certification pathway, and produces DSAT-compatible audit records. A full data processing agreement is available under NDA on request.
See What Your Security Stack
Is Missing.
Structured technical briefings for defence procurement, security leadership, and intelligence teams. NDA available. Air-gapped demo environment on request.
DEFENCE@CAVEFISH.CO.UK · CARDIFF, WALES · UK DATA RESIDENCY STANDARD