Harnessing AI and ML for Identity Verification Using Amazon SageMaker and Amazon Rekognition with Jumio

Founded in 2010 and based in Sunnyvale, California, Jumio has pioneered techniques that are now a mainstay across the IDV space. For example, the company offers a seamless end user experience that makes it simple for people to verify their identities by photographing their IDs and taking a selfie. Businesses around the world now rely on this method to verify that people are who they say they are.

According to the NSA and the FBI, the generative AI market is expected to hit $100 billion by 2030, putting advanced capabilities into the hands of cybercriminals. This prediction makes state-of-the-art IDV procedures critically important for Jumio and its customers. “We have over 50 ML models involved in the IDV process, determining whether the people on-screen are live humans or some type of deepfake or screen attack,” says Huff.

Jumio built its solution on AWS, and it offers customers a range of solutions to meet their specific use cases. For example, Jumio offers both ID verification and IDV through the Jumio KYX Platform. However, if customers need only ID verification, they don’t have to pay for IDV. This flexibility makes it simple for customers to optimize the tradeoff between security and the end user experience. Banks, for example, might choose stringent security procedures, knowing that their end users value the added protection. A university, however, might assess that top-tier security measures don’t justify the added friction for their end users.

Jumio innovates to stay one step ahead of cybercriminals, using two services from AWS to help power its offerings.

Building 50-Plus ML Models at Scale Using Amazon SageMaker

To validate the authenticity of government-issued IDs, Jumio uses Amazon SageMaker, a service to build, train, and deploy ML models for any use case with fully managed infrastructure, tools, and workflows. “We build and train models of all sizes and purposes using Amazon SageMaker,” says Huff. “Given the variety of systems that AWS offers, we tune our development training environment to match the needs of each model in a cost-efficient way.”

The company’s engineers are constantly experimenting and enhancing its solutions using Amazon SageMaker. “We’re starting to use some of the large language models for new purposes,” says Huff. “We have a battery of proprietary models and analysis techniques that look for inconsistencies in the extracted data and other signs of tampering or forgery of the identity document. We’re also extracting text from IDs and comparing it against government databases.”

Providing Secure Face Verification Using Amazon Rekognition

Many Jumio customers require biometric verification as part of their IDV process. As part of its overall solution, Jumio also deploys Amazon Rekognition, which offers pretrained and customizable computer vision capabilities to extract information and insights from images and videos. By applying computer vision to the verification process, Jumio strengthens security while delivering a virtually seamless user experience.

Using Amazon Rekognition, Jumio also advances its mission of providing fair IDV. “One of our core value propositions is to offer solutions that are free from bias and work across all different genders, age ranges, and ethnicities,” says Huff. “We find Amazon Rekognition to be highly accurate, and it’s key to advancing our mission.”

When it comes to face verification, subtle details matter. Amazon Rekognition helps Jumio catch those details. “You can hold a passport for 10 years, and at the end of that period, you’re 9 years older than you were in your passport photo. A critical part of our business is to make sure that the older person in front of the camera matches the image from the government-issued document. And that’s where Amazon Rekognition 1:1 and 1:n matching helps,” says Huff.