
The Scoop
The legendary technologist known as “Darpa Dan” has joined an obscure cybersecurity startup with an ambitious goal: A biometric security product that functions without storing users’ credentials anywhere.
Dan Kaufman, whose career has spanned Dreamworks, Google and a 10-year tour at the Defense Advanced Research Projects Administration, told Semafor he believes the new project may be his most consequential endeavor yet, with the potential to render a huge category of malicious hacks useless while making life easier for consumers.
“The world is catching on that the old way of doing things is a terrible way of doing things,” he told Semafor.
The company, Badge Inc., was co-founded in 2019 by Dr. Tina P. Srivastava, a former NASA rocket scientist and head of electronic warfare programs for Raytheon. Her thesis was that the cybersecurity paradigm is fundamentally broken because of one, fatal flaw: Security credentials need to be stored somewhere, but nowhere is safe from hackers.
Srivastava and co-founder Dr. Charles Herder figured out a way to authenticate internet users without the need to store security credentials at all with a new method of encryption that uses a concept often called “fuzzy logic,” which enables biometric data like fingerprints and face scans to be used for authentication without exposing any sensitive information.
It also means that biometric authentication can be used from anywhere, not just a phone or a specific piece of hardware.
“We need to live in a world where you’re not tied to a device anymore. It works on anything,” Srivastava said.
She set out to create the method after her information was stolen by Chinese hackers in 2015 in a data breach that targeted the Office of Personnel Management. The breach exposed 22.1 million records, including government employees and others who had been subject to background checks.
Srivastava says that hack and others like it shouldn’t even be possible, because storing sensitive information is unnecessary with the new method of encryption that Badge employs.
And yet, the hacks continue. On Tuesday, a whistleblower revealed that apparent Russian hackers may have stolen gigabytes of sensitive data belonging to American businesses simply by logging in with usernames and passwords belonging to workers from the Department of Government Efficiency.
In this article:
Step Back
Kaufman, who served at DARPA from 2005 to 2015 in various roles that included director of cybersecurity, said our online accounts are not as secure as we think, despite fancy technology used in authentication. “The hackers always win,” he said.
Apple’s Face ID, for instance, is a technological marvel. The iPhone creates a 3D map of your face and stores a version of it inside the phone’s “secure enclave.” When you use FaceID, the phone maps your face again and compares it to the one stored on the device, ensuring it’s really you. Your bank can use Apple’s Face ID, too, by storing a key that can only be locked by your device. It’s essentially flawless, except for a glaring problem: if you lose your phone or get a new one, you have to use a different method.
That means your bank has to let you prove your identity in other ways, like providing your mother’s maiden name, or the town where you were born. Hackers can get that information easily.
It’s essentially like having the world’s strongest vault, but always leaving a side door open in case someone forgets the combination.
Badge’s method differs in that no device is necessary to store a copy of your face, or fingerprints or any other method. If you used facial recognition to enroll in online banking, there would be a key generated based on your face that the bank would store. But the key wouldn’t be useful on its own.
That’s not just because it’s encrypted — Badge says it is even theoretically resistant to quantum computers — but because it really only matches the person’s face approximately. It’s sort of like a bouncer who is there to let in only one special guest into a night club. But the bouncer doesn’t know who the guest is until they show up at the front of the line. (This is how the concept got the name “fuzzy”). When the bouncer sees that person, they suddenly remember the code to the door lock.
Each time a person enrolls with a new Badge credential, a new bouncer is created, sharing no similarities with any of the others. So even if a hacker were to tie one of the “bouncers” to a particular user, it couldn’t be used to identify that person in other hacks, or unlock any of the person’s other accounts.
The benefit of the method is that users aren’t bound to a single device. They can use their faces as an authentication method from anywhere, and there’s no information stored that could be useful to hackers.
Usually, better security means less convenience and more complexity. In this case, it’s much simpler.
Badge shared one real world example: One of its clients uses it in an office environment where employees share workstations. When an employee sits down at a desk, the computer automatically unlocks based only on facial recognition and loads that person’s profile. But when they log in outside the office, they use facial recognition in addition to a PIN. If they forget the PIN, they can reset it from the office.
“We’re finally able to deliver the secure and private internet envisioned more than 40 years ago,” Kaufman said.
Badge has signed deals with Microsoft, Cisco, Okta, and others, charging customers on a per-user basis.

Reed’s view
Over coffee in San Francisco, Kaufman and Srivastava explained how Badge worked. It took me a while to wrap my head around it, but once it clicked, the concept was compelling.
Most great consumer tech products are very complicated to make but very easy to use. It’s the only authentication method I know of that fits that description.
Authentication is going to become increasingly important in a world in which AI can mimic humans online. At the same time, we are sitting ducks for hackers.
It seems like every week, there’s a major breach affecting a huge company and its customers. Or, in the case of the recent DOGE hack, taxpayers all over the country.
Badge has seen early success with some big-name partners, but it’s disappointing that there isn’t more widespread adoption. The sad truth about cybersecurity is that companies have little financial incentive to put resources into innovating in security.
Using a password manager and a hardware device like Yubikey works well for the most security-conscious, but setting it up is daunting for the vast majority of people.
The alternatives aren’t much better. Authentication apps can turn a lost phone into a password nightmare. Passkeys, which were supposed to be a better alternative, can have the same issue. They are often tied to one device or platform. It’s not always clear to people that they are creating a passkey in the first place, or where it’s being stored, making it somewhat confusing to a lot of people.
The nice thing about the Badge concept is the simplicity. It would make it easy to log into accounts, but would allow companies to make it much more difficult for hackers to find a way in through the side door.
Companies need to start looking for better solutions.

Room for Disagreement
Even if no data is stored, advances in biometric technology could be abused, the FTC wrote in 2023:
“Consumers face new and increasing risks associated with the collection and use of biometric information. For example, using biometric information technologies to identify consumers in certain locations could reveal sensitive personal information about them such as whether they accessed particular types of healthcare, attended religious services, or attended political or union meetings. Large databases of biometric information could also be attractive targets for malicious actors who could misuse such information. Additionally, some technologies using biometric information, such as facial recognition technology, may have higher rates of error for certain populations than for others.”

Notable
- The Brookings Institute has a comprehensive article on why new methods of biometric authentication are needed, focusing on nightmare scenarios if databases of biometrics fall into the wrong hands.