TypingDNA raises $7 million for AI that identifies people by how they type
The risk of online accounts being compromised increases by the day — and it’s costly. A hacker attacks roughly every 39 seconds, and it’s estimated that 65% of companies have over 1,000 stale accounts. Worse still, the average expense of data breaches is expected to hit $150 million in 2020.
That’s perhaps why startups like TypingDNA have risen to prominence. TypingDNA — which was founded in Romania in 2016 by Adrian Gheara, Cristian Tamas, and Techstars alum Raul Popa, and which recently moved its headquarters to New York — provides typing biometrics authentication as a service, enabling companies to recognize people by the way they type. After raising $1.8 million in capital from the founders of companies such as UiPath and DataDog last year and attracting clients like large banks in Latin America and Romania, it’s gearing up for a major expansion in 2020.
TypingDNA announced this morning that it has raised $7 million in series A funding led by Gradient Ventures, Google’s AI-focused venture fund, with participation from Techstars Ventures and EU-based fund GapMinder. According to CEO Popa, the proceeds will fund the build-out of TypingDNA’s developer support network and tools to integrate its services with popular web development tools.
“Advancing the research and distribution of typing biometrics is of global importance. Keyboards are incorporated [into] almost any device today, making typing behavior the most widely available user biometric,” he added. “This round of funding will allow us to further our mission to provide user-friendly, non-intrusive biometrics and increased security to people around the world.”
TypingDNA’s platform records dynamic statistics about pressed keyboard keys and turns them into typing patterns, which its proprietary engine analyzes and verifies against patterns collected from real-world users. As Popa explains, the way a person types on a keyboard is unique and fairly difficult to replicate — in point of fact, it’s behaviorally rich enough to reveal biometric traits like gender and age.
Local clients train models on users’ patterns silently in the background, focusing on 44 commonly used letter and number keys. Stats like individual key presses and flight times are stored for later algorithmic ingestion, but not sequences between two or more keys — that’s to ensure text can’t be recreated.
Popa says most intruders are caught in under 100 typed characters, or on a typical sentence.
TypingDNA offers a number of client-side products, including a PC tool that continuously authenticates users based on their typing patterns. It alerts administrators or automatically locks down the system when it detects anomalous keystrokes, and it optionally signs out accounts, captures screenshots, takes webcam snapshots, and sends alerts (via email or SMS) if it detects anomalous behavior. A complementary Chrome extension serves up two-step verification codes for third-party services like Amazon and Facebook in the browser, and TypingDNA’s API and SDK provide a means of securing logins and enforcing reset passwords on mobile, as well as web apps.
The biometric authentication market is projected to generate $65.3 billion by 2024, and TypingDNA is far from the only player in the lucrative space. Pindrop, a firm developing biometric voice authentication solutions, raked in $90 million in funding in December 2018. Global identity verification provider Onfido nabbed $50 million in April. And Buguroo, which taps deep learning and behavioral biometrics to detect banking fraud, recently secured $11 million.
But Gradient Ventures general partner Darian Shirazi isn’t concerned. “We’re excited about TypingDNA’s developer-first approach to [enabling] people to authenticate securely based on how they type,” he said. “With global regulation impacting face-recognition-based authentication and hackers targeting SMS-based two-factor authentication, typing biometrics is the best form of identifying people without compromising privacy or security.”
