Some efforts to replace traditional letter-and-number passwords rely on gestures, wearable devices, or biometrics. An approach in the works from research-and-development company SRI International and Stanford and Northwestern takes a different tack: passwords that you know but don’t know you know.
Patrick Lincoln, director of SRI’s computer science laboratory and a researcher on the project, calls this “rubber-hose resistant authentication” in reference to rubber-hose cryptanalysis, in which a user is coerced or physically forced to give up, say, the passcode to a secure building. Lincoln says the approach relies on implicit learning—the sort of learning that occurs through sheer repetition, such as learning to ride a bike, that the learner can’t verbally explain—to prevent passwords from being compromised.
So far, the project has used a game interface, resembling a rudimentary version of Guitar Hero, that trains the user to enter a unique pattern. Users press a key, corresponding to a column, each time a falling ball hits the bottom of one of the columns, but because the sequence of falling balls changes each time, users can’t consciously determine what is their unique sequence, and what is simply added noise. Later, the user is authenticated by playing the game, which contains parts of the learned pattern, and the user’s superior skill at this task proves his or her identity.
It’s one of many attempts to move away from standard passwords, which can be hard to remember and insecure. And if the researchers behind the project can get it to work sufficiently well, it may eventually help workers enter high-security areas like airplane cockpits, as well as more mundane realms such as your home or bank account.
Users also might be able to learn more than one unconscious password without interference, Lincoln says—so you could have one unconscious password for your office and another for your bank account. And if one of the passwords was somehow compromised, you could be retrained on that one without wiping out the other.
The researchers’ initial findings were published in a paper last year, including a study indicating that trained users could properly enter their patterns over time but couldn’t consciously remember them. The project has received a National Science Foundation award that Lincoln says is allowing the research to move forward. So far, Lincoln says, training is time-consuming (it takes about 40 minutes per password), and the system’s accuracy needs work, since this password system is not yet even as secure as traditional passwords. Lincoln’s group is launching some new experiments that he hopes will lead to more-effective and easier-to-learn unconscious passwords.
Despite the challenges and current impracticality of such a system, David Wagner, a UC Berkeley computer science professor who studies computer security, notes that there are examples of security technologies becoming widely used despite initially seeming impractical, such as public-key cryptography, which got its start in the 1970s with the invention of the RSA encryption algorithm. “Anyone can guess whether this will ever see the light of day,” he says, “but it’s pretty inspiring to see, at least in theory, that it might be possible to have a password you don’t know but you can use.”
How a Russian cyberwar in Ukraine could ripple out globally
Soldiers and tanks may care about national borders. Cyber doesn't.
Meet Altos Labs, Silicon Valley’s latest wild bet on living forever
Funders of a deep-pocketed new "rejuvenation" startup are said to include Jeff Bezos and Yuri Milner.
A horrifying new AI app swaps women into porn videos with a click
Deepfake researchers have long feared the day this would arrive.
Meta’s new learning algorithm can teach AI to multi-task
The single technique for teaching neural networks multiple skills is a step towards general-purpose AI.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.