Hacker News Reader

Comment by _v7gu | Hacker News Reader

How probable is it that my fingerprint matches with someone else's, given that there is some acceptable variance allowing my fingerprints to work every time?

View on HN

Depends entirely on the implementation, of course, but modern implementations have extremely low probabilities of false matches.

These are tested by running the algorithm against large databases of fingerprints and looking for false matches between any two different fingerprints.

Note that this scenario is vastly worse than any real-world scenario, as you can’t simply have a million different people try to unlock a device.

The practical question is to consider the probability of a specific random person being able to unlock your phone, which is orders of magnitude lower probability than considering if any person can unlock your phone.

Any secure fingerprint system would also have exponential back off for excessive tries, further limiting the practicality of a random attacker matching.

Reminds me of when I was a kid. We left the grocery store, unlocked the car with a keyless remote and got in. My dad put the key in the ignition and couldn't turn the car on. He was confused, I was confused, I had felt a general sense of wrongness. Looked around and noticed, this wasn't our car. Same Make, Model, Year and Color and the keyless also matched, though the interior color was different. Our car was two cars further down the aisle. The chances of that happening are so very slim, but also not.

Same with a mid-90s Audi with my grandmother picking up someone from the airport. Took us a minute to figure out we weren’t in her car because both were immaculate inside. Though it was a while ago so I’m thinking they both just shared the same physical key to unlock the doors and start the engine.

For a long time (maybe still?) it was common for police forces to buy cars which were keyed alike. What happened when those cars were gradually replaced? Well, the new cars had to be keyed to match the old ones, and taxis or anyone else in the market for old police cars would unwittingly be able to unlock the entire police fleet.

What is the probability that this happened to my friends mom as well? We noticed because there was some stuff in the back that didn't belong to them.

What is the probability that yours was the same type of car as theirs? Unfortunately I don't remember for sure. But it was a small car. Sorry I can't be more specific. It's been like 30 years :)

It was a Honda Accord Special edition with the Purpleish paint job from the 90s. Don't remember exact year, but a very specific, not generic looking, car.

It all depends on a given implementation's choice of a threshold to use. Two objects will give a score which indicates similarity/dissimilarity and then some further logic needs to say if that score is good enough for a match. Some NIST certifications for instance end up placing requirements with false-match-rates at 1/10,000 (e.g. one left index finger matching someone else's left index finger at that frequency when trying a lot of random pairs).

Reply on news.ycombinator.com