Liveness Detection vs Document Forgery Detection: Key Differences

Liveness detection verifies that a real person is physically present during biometric verification, while document forgery detection determines whether an identity document is genuine or fabricated. These two controls address fundamentally different attack vectors and must be combined in any robust identity verification process.

This article is provided for informational purposes only and does not constitute legal advice. Regulatory references are accurate as of publication date, June 2026.

Identity fraud attempts frequently combine both vectors: a fraudster may present a stolen authentic passport while using a printed photograph to defeat biometric verification. Understanding the distinction between these two techniques is essential for designing a KYC architecture with no blind spots.

What Is Liveness Detection?

Liveness detection refers to the techniques that verify whether the face presented to a camera belongs to a live human being — as opposed to a printed photograph, a replayed video, a 3D mask, or an AI-generated deepfake.

The ISO/IEC 30107-3 standard (Biometric Presentation Attack Detection) is the international reference for evaluating liveness detection systems, defining attack types (printed artefacts, video replay attacks, 3D masks) and standardised evaluation metrics (ISO/IEC 30107-3:2023).

Active vs Passive Liveness

Liveness detection systems fall into two main categories based on the degree of user interaction required:

• Active liveness: the user performs a prompted action (turn head, blink, read a number). More robust against static photo and video replay attacks, but introduces friction in the user journey.
• Passive liveness: the system analyses texture, depth, natural micro-movements, and compression artefacts without visible interaction. Delivers a smoother user experience but requires more sophisticated models to match active liveness robustness.

Attack vectors against liveness systems include: high-resolution printed photographs, video replay on a secondary screen, printed 3D masks, and — since 2024 — real-time deepfakes injected through virtual camera drivers.

Regulatory Framework for Liveness Detection

The EU AI Act (Regulation (EU) 2024/1689) classifies remote biometric identification systems as high-risk AI (Annex III, Category 1). Liveness detection systems used in regulated identity verification contexts are subject to mandatory conformity assessment before market placement in the EU (Regulation (EU) 2024/1689, Art. 6 and Annex III).

eIDAS 2 (Regulation (EU) 2024/1183), progressively applied since January 2025, requires a "high" assurance level for sensitive identity use cases. This level mandates online identification procedures incorporating active liveness detection (Regulation (EU) 2024/1183, Art. 8).

The UK FCA, through its guidance on digital onboarding under the Money Laundering Regulations 2017 (as amended), expects firms to implement verification measures that are commensurate with the risks associated with non-face-to-face customer interaction — which in practice means liveness checks for remote KYC (FCA Financial Crime Guide, Chapter 3).

What Is Document Forgery Detection?

Document forgery detection encompasses the techniques used to determine whether a presented document is genuine, altered, counterfeit, or AI-generated. It applies to all document types: identity cards, passports, payslips, bank statements, diplomas, and driving licences.

The ACFE estimates that only 37% of document fraud is detected by manual controls, underlining the necessity of automated checks that cover attack vectors invisible to the human eye (ACFE 2024 Report to the Nations).

Main Categories of Document Fraud

Forensic Analysis Layers

Document forgery detection relies on several complementary analytical layers:

1. Security feature inspection: MRZ (Machine Readable Zone) validation, hologram checks, secure fonts, UV watermarks
2. Error Level Analysis (ELA): detects areas altered by inpainting or digital compositing through JPEG compression inconsistencies
3. Metadata verification: consistency between creation software, date, colour profile, and the document type expected
4. Semantic cross-validation: consistency between fields within a document and across other documents in a dossier

AMLD6 (Directive (EU) 2024/1640, Art. 20) requires obliged entities to implement remote identification measures proportionate to document fraud risks, including automated controls for digital onboarding (Directive (EU) 2024/1640).

AI-generated document detection is deployed as an additional layer alongside existing structural controls, depending on client configuration. The CheckFile platform provides this capability for organisations seeking enhanced protection against AI-synthesised documents.

Side-by-Side Comparison

Why Both Controls Are Essential

Liveness detection and document forgery detection cover radically different attack angles. A robust KYC system must combine both to eliminate blind spots.

Scenario 1 — Authentic document, biometric spoof. A fraudster uses a stolen authentic passport belonging to a victim. Document verification validates the document (it is genuine). Without liveness detection, the fraudster can pass verification by presenting a high-resolution printed photograph of the victim. Liveness detection identifies the printed artefact and blocks the attempt.

Scenario 2 — Real person, forged document. A loan applicant submits an AI-generated payslip showing a fictional income. Liveness detection confirms a real person is in front of the camera (which is true). Without forensic document analysis, the fraud passes undetected. Document forgery detection identifies metadata anomalies and flags the file.

Scenario 3 — Combined double fraud. Real-time deepfake + synthetic document. This sophisticated attack vector requires both protection layers operating simultaneously. ENISA documented in 2024 a rise in attacks combining biometric deepfakes with AI-generated identity documents in financial onboarding contexts (ENISA Threat Landscape 2024).

Compliance professionals frequently raise the question on specialist forums: should liveness or document verification be implemented first? Practitioners consistently answer that both must run in the same KYC flow, because experienced fraudsters deliberately target whichever control is missing.

For a detailed breakdown of liveness detection in banking KYC contexts, see our guide on liveness detection and identity spoofing prevention.

Integration in a KYC Workflow

The optimal implementation sequence in an identity verification process typically follows these steps:

1. Document capture → real-time forensic analysis (document forgery detection)
2. Data extraction → OCR + MRZ validation
3. Selfie or video capture → liveness detection
4. Face matching → biometric comparison face-to-document
5. Global scoring → consolidated KYC decision

This sequence aligns with eIDAS 2 "high" assurance level requirements for remote identification. NIST SP 800-63A (Digital Identity Guidelines, Rev. 4, 2024) recommends a comparable architecture for digital enrolment flows (NIST SP 800-63A).

Modern identity verification solutions expose both capabilities through a unified API, enabling full KYC flow coverage without multiple integrations. Data security and decision auditability are complementary requirements in a regulated context.

For technical integration guidance, see our document fraud detection API guide.

Explore our complete document verification guide to understand the full range of controls required by sector.

Frequently Asked Questions

Are liveness detection and document verification interchangeable?

No. They cover different attack vectors. Liveness detection protects against biometric spoofing; document verification protects against forged or fabricated documents. A fraudster can pass one and fail the other — which is why both are required in a complete KYC process.

What is the international standard for liveness detection?

ISO/IEC 30107-3 defines Presentation Attack Detection (PAD) requirements. Certified systems have been tested by accredited laboratories such as iBeta Quality Assurance at standardised conformance levels (Level 1 and Level 2), measuring both the Bona Fide Presentation Classification Error Rate (BPCER) and the Attack Presentation Classification Error Rate (APCER).

Does the EU AI Act apply to liveness detection systems?

Yes. Regulation (EU) 2024/1689 classifies biometric verification systems used in regulated contexts as high-risk AI (Annex III). Providers must complete a mandatory conformity assessment before placing their system on the EU market, with requirements covering robustness, transparency, and record-keeping.

Is document verification alone sufficient for AMLD6 compliance?

Directive (EU) 2024/1640 requires identification measures proportionate to assessed risk. For remote identification (digital onboarding), the combination of document verification plus liveness detection represents the minimum standard for medium-to-high risk levels. Document verification alone may be insufficient depending on the obliged entity's own risk assessment.

How should I evaluate the quality of a liveness detection system?

Prioritise solutions with ISO/IEC 30107-3 certification from an accredited laboratory, with published BPCER and APCER metrics. Verify that models are updated regularly to cover new deepfake generation techniques (GAN, latent diffusion). Also confirm compatibility with eIDAS 2 assurance level requirements for your specific use case.

For where this fits in the CheckFile offering, see our AI and deepfake detection approach.