Jumio Go Automates Identity Proofing to Help Businesses Digitally Onboard Customers
Photo Gallery
 |
Jumio Go Automates Identity Proofing to Help Businesses Digitally Onboard Customers
Additional Info
| Company | Jumio |
| Website | http://www.jumio.com/ |
| Company size (employees) | 1,000 to 4,999 |
Overview
In 2020, Jumio made significant advancements to automating its identity proofing services through major breakthroughs in AI and machine learning. In 2019, Jumio introduced Jumio Go, the first fully automated digital identity verification solution on the market capable of defending against bots, advanced spoofing attacks and sophisticated deepfakes, which are increasingly being weaponized for fraud. Powered by AI and machine learning, Jumio Go enables modern enterprises to reliably verify remote users, ensuring that someone is who they claim to be online.
Jumio was able to revolutionize our AI models thanks to several significant breakthroughs by our AI Labs team. These breakthroughs allowed us to scale Jumio Go worldwide, and to improve the speed and accuracy of our fully automated solutions to reach more businesses frantically scrambling to digitally transform their operations.
Jumio only leverages real-world production data to build its algorithms. Jumio has processed more than 300 million verifications comprising more than 3,500 types of ID documents from over 200 countries and territories. This dataset of ID documents is the second largest in the world (second only to Interpol) and gives us a big leg up in developing our AI algorithms. Until COVID struck, all tagging was performed by trained, seasoned identity verification specialists. By tagging these historical identity verification transactions, Jumio was able to develop and refine our ML algorithms which has subsequently improved the quality and accuracy of our verification decisions.
As a result of these AI breakthroughs, Jumio successfully transitioned our key customers who were using our hybrid identity verification solutions to our fully automated identity verification solution with improved speed and greater verification accuracy in 2020. Jumio Go now supports more than 1,200 ID types and the level of verification accuracy has improved by more than 125% in less than a year.
How we are different
Ground Truth: Jumio leveraged supervised AI from the beginning, meaning Jumio employs identity verification experts who tag every identity verification based on an analysis of the security features of an ID and selfie. These experts have thousands of hours of experience in dealing specifically with government-issued IDs from all over the world. By tagging these historical identity verification transactions, Jumio could develop and refine our ML algorithms which has subsequently improved the quality and accuracy of our identity verification decisions.
A Better OCR Engine: Jumio uses OCR to extract important data from an ID document. OCR was originally intended for reading black text against a white background often using a flatbed scanner — not for extracting key data fields from ID documents using small fonts and colored backgrounds that include holograms, watermarks and microprint. Powered by informed AI, Jumio’s OCR engine overcomes many of the limitations of traditional OCR and is capable of highly accurate data extraction — data that can be used to ping third-party databases or to verify a person’s age (based on DOB). In 2020, our OCR engine actually outperformed Amazon’s OCR engine in terms of data extraction and accuracy.
Production Data: Jumio’s AI models are based on a decade’s worth of real-world production data. Jumio AI models are trained on real ID images and learn to make a decision even when the quality of images submitted by users is less than perfect. This database enables our AI models to minimize demographic bias since all types of people, nationalities and documents coming from real production environments are used. Jumio has direct communication with document-issuing authorities, meaning Jumio models are also informed by the ID document-issuing authorities in many countries. As a result, more realistic data has translated into better and dramatically less bias in our AI models.
