The advent of GenAI has increased the sophistication of attacks that identity verification
vendors must defend against. GenAI tools are capable of producing seemingly real
content in voice, video and image format with minimal technical input, and deepfake
misuse can subvert the verification process. Though deepfakes have existed for some
time, the proliferation of user-friendly tools has made their creation more readily
accessible, even to individuals with limited technical proficiency. The number of
deepfakes detected worldwide in 2023
1
was 10 times the number detected in 2022.
Gartner estimates the time to reach the early majority (i.e., more than 16% target market
adoption) for deepfakes is one to three years because deepfakes go hand-in-hand with the
GenAI advances that underpin their creation. (See Emerging Tech Impact Radar: Artificial
Intelligence). This requires that identity verification vendors take a multipronged approach
to safeguard against these rising deepfake attacks.
Market Definition
Gartner defines identity verification as the combination of activities during a remote
interaction that brings a real-world identity claim within organizational risk tolerances.
Identity verification capabilities, delivered as SaaS or on-premises, provide the assurance
that a real-world identity exists and that the individual claiming the identity is its true
owner and is genuinely present during a remote interaction.
This typically involves a person capturing a real-time image of their photo identity
document, which the tool inspects for signs of counterfeit or forgery. Once the authenticity
of the document is established, the person is prompted to capture a photo or a video clip
of their face. During this step, the tool establishes the genuine presence of the person
using liveness detection (or, formally, presentation attack detection), followed by biometric
facial comparison with the photo from the identity document.
Some identity verification vendors also capture voiceprint during the verification process,
to be leveraged in the future for the purposes of contact-center voice authentication. That
user flow is susceptible to voice deepfakes, and some vendors are investing in real-time
voice deepfake detection capabilities as well. However, for the purpose of this research
note, we have focused on deepfakes targeting selfies or the video capture process during
identity verification. Figure 1 summarizes the critical insights for deepfake detection
discussed in this document.
This research note is restricted to the personal use of liuyang17@qianxin.com.