Algorithms that can fool fingerprint scanners

Algorithms that can fool fingerprint scanners

Researchers used a neural network to generate artificial fingerprints that act as biometric keys for identification systems. An artificial intelligence called DeepMasterPrints can create fake fingerprints with which it can fool fingerprint scanners.

Every security feature has a gap of some kind. Fingerprint scanners, whichóre increasingly being used to protect at least a smartphoneów, also can be fooled. This was proven by New York researchers whoóers have developed an artificial intelligence – DeepMasterPrints – capable of doing just that.

At a conference on security issues in Los Angeles, researchers from New York University presented an artificial intelligence called DeepMasterPrints. They can generate fake fingerprintsóin and without the problem ofów cheat fingerprint scanners.

The method developed by the researchers could have wide applications in fingerprint securityóIN FINGERów, as well as in the synthesis of imprintedóin fingerów. In security research, demonstrating the shortcomings of existing systemsóin authentication is considered an important part of developing a more secure replacement for theóin the future.

DeepMasterPrints takes advantage of two properties of the systemóin fingerprint-based authenticationów. Most readersóin fingerprint does not read the whole finger. This is partly due to the negligence of the users themselvesóin the systemów. Instead, it only pictures the part of the finger touching the scanner.

Most importantly, the authentication system does not combine all partial imagesóin order to poróof full fingerprints with full record. Simply porównn partial scanning with partial recording. This means that a forger only needs to match one of dozens ofów or hundreds of stored partial imagesóin the fingerprint to gain access.

The second property, with whichówhich DeepMasterPrints uses, is the fact that someóre characteristics of the imprintów fingerów are more frequent than others. This means that it is more likely to match a fake print containing many very common features.

Based on these insights, the researchers used a neural network to generate fake printsów fingeróin, whichóre match as many partial imprints as possibleów fingerów.

Based on a library of more than 6,000 printsóin the DeepMasterPrints fingerprint generated the basic skeleton of the imprint. For this pattern, algorithms match fragments of other fingerprints until they are successful. Such a method may not be able to hack a specific account, but when used on a large scale, it generates a high enough number of hits that it is worth the effort.

– Without verifying that the biometric data is from a real person, many such attacksów becomes possible – said Philip Bontrager of New York University.

SourceóSource: Engadget, The Guardian, pic. Rachmaninoff/ Wikimedia Commons/ CC BY-SA 3.0

Related Posts