Spiceware, a Korean computer security startup, has unveiled a dynamical pseudo-random generator based on reinforcement learning and a cryptosystem ensuring security even against a quantum computer.

Reinforcement learning is a type of machine learning based on a reward-penalty method to teach an AI system.

The company claimed its invention to be the world’s first AI-based random number generator whose study results were published in international peer-reviewed journal Applied Sciences.

Random numbers should be random, unpredictable, unbiased, and independent and these true random numbers could be generated by no other means but quantum-safe security. The current random number field, however, is focused on hardware chip-based quantum random numbers, raising the need to develop a new one consisting of only software free from cost and other environmental restrictions.

Spiceware utilizes the uncertainty of artificial intelligence and its characteristic to determine the optimal direction at every moment.

Reinforcement learning consists of countless variables and formulas, and is updated by itself through learning, so that the next random number cannot be predicted with an already generated random number pattern. The study also demonstrated that the new random generating method met randomness and non-reproducibility.

The cryptographic key made of the random numbers generated from this method cannot be predicted even in a quantum computer environment, said the company’s chief executive Kim Geun-jin.

By Jung Hee-young and Minu Kim

[ⓒ Pulse by Maeil Business Newspaper & mk.co.kr, All rights reserved]