Not uncrackable: Even quantum physical random number generators do not yet deliver absolutely perfect random sequences – that is a problem for encryption. But physicists have now developed a method that uses “random amplification” to convert such imperfect sequences into a perfectly random series of bits. This was achieved using entangled quantum bits, a Bell test and a special experimental setup. The technology could make encryption more secure, as the team explains in “Nature”.
Random numbers are indispensable for encrypting data, but also in statistics and simulation. Billions of such random numbers are generated every day by so-called random number generators. But even the most modern random number generators based on quantum physics processes do not deliver absolutely random, unpredictable series of numbers. Small systematic errors can cause individual numbers to appear slightly more often than others – the resulting number is no longer completely random.
This has consequences, especially for cryptography: “Parts of private keys that were created using such random numbers turned out to be predictable and that made it possible to reconstruct these keys,” explain Anatoly Kulikov from ETH Zurich and his colleagues.
Even more random thanks to “random amplification”
But there is a method that can convert an originally imperfectly random series of bits into a perfectly random bit sequence: random amplification. The core element of this procedure is a so-called Bell test, which physicists usually use to check whether, for example, two photons are quantum physically entangled with each other or not. In their experiment, Kulikov and his colleagues used such a Bell test to determine the measurement parameters for the readout.
“In a second step, this independent property is used for a technique known from computer technology, the two-source extractor,” explain the physicists. If this algorithm is fed with two imperfect random sequences and their modification using the Bell test, it can generate perfectly random bit sequences. “The resulting sequence of zeros and ones is now really perfectly random, and we can even certify that,” says senior author Renato Renner from ETH Zurich.
Quantum chips, entanglement and an extractor
Specifically, the experiment went like this: The experimental setup consists of two superconducting chips in special freezers 30 meters apart. Each chip contains a quantum bit that generates sequences of imperfectly random bits based on a common quantum random generator. The qubits are entangled with each other via microwave photons and a 30-meter-long pipeline. If a qubit is read out, this automatically influences the measurement result of the entangled qubit at a distance of 30 meters.
The highlight, however, is that the measurement parameters for the random sequences generated by both qubits are changed arbitrarily based on the Bell test. “The random input bit controls the state of a microwave switch, which, depending on the condition, generates a microwave pulse that rotates the qubit by π/2 around a controlled axis,” the physicists write. The resulting outputs are then combined by a two-source extractor, which creates a perfectly random bit sequence.
“Perfectly coincidental for all eternity”
To verify perfect randomness, Kulikov and his colleagues subjected the resulting bit sequence to two internationally established batteries of standardized randomness tests. The random series generated in the experiment passed all tests. “Thanks to technical improvements, we can for the first time produce random numbers that remain perfectly random for all eternity – no matter what analysis methods are used to assess the randomness,” says Renner.
According to the team, their method represents a physically certified source of random sequences that can be used for numerous applications. Applications could range from encryption of sensitive communications to digital identities to public random services for lotteries and blockchain applications.
Source: Anatoly Kulikov (Swiss Federal Institute of Technology Zurich) et al., Nature, 2026; doi: 10.1038/s41586-026-10521-8