American researchers have developed graphene chips to generate encryption keys. These keys would be impossible to predict with artificial intelligence, one of the weak points of silicon chips.
Many cryptographic systems are function-basednon-clonable (PUF), microchips that generate . Although apparently identical, each chip generates different responses caused by internal variations induced by the manufacturing process. In theory, this ensures data security, as each PUF is unique.
However, during a data breach, an artificial intelligence (AI) can analyzeand create a model that predicts the operation of a PUF by . Any new key is then compromised, forcing a change of chip, or even computer. To solve this problem, researchers at Pennsylvania State University turned to the . The process was detailed in an article published in the journal .
Low-power, reconfigurable cryptographic chips
The team created 2,000 transistors, which each have a slightly different due to the manufacturing process, and have integrated them into PUFs. The researchers then used their characteristics to create a simulation of 64 million of these chips, which was analyzed by an AI to try to predict the results, without success.
The researchers have thus shown that the graphene PUF is resistant to attacks based on artificial intelligence. In addition, these chips have a very low consumption and thanks to thegraphene, where the value of its can be changed, it is possible to reconfigure them. If someone managed to create an attack capable of predicting the keys generated in this way, there would be no need to replace the PUF.