Jumio deepfake detection adds ISO Level 2 liveness to combat fraud

The integration of Level 2 liveness detection addresses a critical vulnerability in remote onboarding as synthetic media becomes indistinguishable from authentic video to human reviewers. For streaming platforms and digital services, this shift moves identity verification from simple visual checks to forensic-level analysis of light scattering and micro-expressions. As 'deepfake-as-a-service' lowers the barrier for attackers, the broader ecosystem must transition toward randomized, hardware-aware challenges to maintain account integrity. Watch for regulatory bodies to move from voluntary guidelines to mandatory technical controls for KYC workflows as deepfake losses continue to scale toward the $40 billion mark.

The broader identity verification market is under intensifying pressure as synthetic media fraud scales rapidly across financial services and digital platforms. According to a 2025 report from the World Economic Forum, deepfake-related fraud losses are projected to exceed $40 billion globally by 2027, driven largely by synthetic identity attacks on KYC and account onboarding workflows (per weforum.org, January 2025). This figure has become a benchmark cited by multiple vendors and analysts to justify accelerated investment in liveness detection and biometric screening.

Jumio's move to achieve ISO/IEC 30107-3 Level 2 compliance places it among a small cohort of vendors that have passed the iBeta lab's more rigorous testing tier. Competitors such as Onfido and Veriff have also pursued Level 2 certification in recent months, signaling an industry-wide race to demonstrate resilience against injection attacks and AI-generated video (per biometricupdate.com, June 2025). The iBeta lab, accredited by NIST's National Voluntary Laboratory Accreditation Program, remains the primary independent testing body for presentation attack detection standards, and Level 2 requires defeating attacks using three-dimensional masks and high-resolution screen replays — a significantly higher bar than Level 1's printed photo and basic video replay tests.

The streaming and entertainment sector has become an unexpected frontier for identity fraud. A 2025 analysis by Arkose Labs found that media and entertainment platforms experienced a 340% year-over-year increase in synthetic identity creation attempts, with attackers using deepfake selfies to bypass account verification during free-trial signups and password recovery flows (per arkoselabs.com, March 2025). This trend has prompted several major streaming services to integrate third-party liveness checks into their onboarding pipelines, though most have not publicly disclosed which vendors they use.

Jumio's patented active illumination technique — which projects randomized color sequences onto a user's face and analyzes the reflected light patterns — represents one approach among several emerging in the market. Competing methods include Veriff's motion-based challenge-response system and iProov's Genuine User Assurance protocol, which uses colored light flashes in a similar fashion but with a different analytical framework (per iproov.com, 2024). The differentiation between these approaches is becoming a key procurement criterion for enterprises evaluating identity verification stacks, particularly in regulated industries where audit trails and certification levels carry compliance weight.

Regulatory momentum is also accelerating. The EU AI Act transparency obligations, which entered into force in May 2024, establishes requirements for qualified electronic attestation of attributes that will likely mandate certified liveness detection for certain high-assurance use cases by 2026 (per eur-lex.europa.eu, May 2024). In the United States, the CFPB has signaled interest in updating its identity verification guidance for financial institutions to address synthetic media threats, though no formal rulemaking has been announced as of mid-2025. Recent research highlights that humans fail at synthetic profile detection at alarming rates, further necessitating these automated security layers.