A group of researchers at the University of the Chinese Academy of Sciences have managed to embed and hide malicious code directly in the artificial neurons that make up a machine learning system, reports Motherboard.
The researchers were even able to embed the malicious code in as much as half of an AI model’s nodes while still being able to function normally.
Some of the models will be tested against 58 common antivirus programs without the malicious code being detected.
“In connection with neural networks becoming more common, this method will be universal for delivering malware in the future,” the researchers write.
The researchers now hope that the scenarios they have presented can be used as a reference to defend neural networks in the future.
Also read: Five unpopular decisions that every IT manager must make
– .
