Create 'universal fingerprint', capable of unlocking the security of current smartphones

Thanks to machine learning power, researchers created one.

Security researchers have successfully programmed artificial intelligence (AI) capable of creating universal false fingerprints that can unlock the vast majority of smartphones using biometric sensors. According to the post-research report, the security attack "has a certain success rate".

It is probably the most perfect in the modern equipment. By individual biometric features (such as fingerprints, iris, etc.), individuals with such characteristics can unlock the device. However, recent studies indicate that the vast majority of biometric identification devices on the market can be fooled.

These artificially made fingerprints work similarly with a universal key.

In many cases, unlocking the face or fingerprints that the device has conducted. But in the scientific report published on arXiv earlier this month, researchers from New York University and the University of Michigan proved they could create a machine learning algorithm that generates fake, alternative fingerprints. The number of fingerprints available in the research database.

As DeepMasterPrints , these artificially made fingerprints work similarly with a universal key. To create it, researchers introduced artificial neural networks to self-fingerprint prints of 6,000 individuals. This is the first time an organization has used the machine learning algorithm to create universal fingerprints.

This is the first time an organization has used the machine learning algorithm to create universal fingerprints.

After analyzing the huge amount of fingerprints, the system creates a fingerprint identical to the whole number of fingerprints in the database. Fake fingerprints are put into an analytical network, finding out whether it is real or fake. If found to be fake, the system will edit to make fingerprints even more real. After thousands of corrections, fake fingerprints deceived the system.

The AI-created universal fingerprint focus focuses on overcoming the majority of the fingerprint recognition systems available on current smartphones. The systems in the phone you are using only use a small part of your fingerprint to identify it, so that identifying the owner is faster. But to get a high speed of validation, security developers had to sacrifice absolute accuracy, creating a security gap that could be exploited.

The researchers created two types of fingerprint data to train the AI ​​system. One is the fingerprint printed on paper, one from the digital scanner system that gets fingerprints. There are three levels of security that must be overcome by universal fingerprints.

The purpose of this research is for future security designers to consider sacrificing security for usability.

Each level has an error rate of FMR - the proportion of fingerprint sensors that identify fake fingerprints is real - different. The highest security level has an error of only 0.01%, an average of 0.1% error and the lowest security level with an FMR error of 1%. This is the failure rate of each security level

At the lowest level of security, researchers can use universal fingerprints to deceive the system in 76% of attempts . The number is impressive, but the researchers claim that no fingerprint sensor system operates at such a low security level. At the average level of security - the most realistic and widely regarded - the universal fingerprint fool the system by 22% of the attempts.

At the highest level of security, universal fingerprints deceive the system by 1.2% of attempts.

The figures above show that the fingerprint security system can be easily bypassed - we can still trust it. The purpose of this study, as the scientists point out, is for future security designers to consider sacrificing security for convenience - allowing unlocking fingerprints faster. ignore the security of the system.