Randomness forms a crucial backbone of modern society, where every encryption key, secure transaction and digital signature depends on random numbers that no adversary can predict. But every random number generator ever deployed, classical or quantum, has asked its users to take the hardware's honesty on faith. Manufacturers characterise each component, certify that it behaves as modelled and assume it will continue to do so indefinitely. If a detector develops a defect or is deliberately tampered with, the output can become predictable without triggering any alarm, thus compromising the security of the system.
All physical devices degrade over time, and quantum components are no exception. With the advent of quantum computing, the consequences are stark: a quantum-equipped adversary could exploit hardware weaknesses that would be invisible to classical testing.
A team led by Associate Professor Charles Lim from the Department of Electrical and Computer Engineering at the College of Design and Engineering , National University of Singapore (NUS CDE), has built a quantum random number generator (QRNG) chip that breaks this pattern. Published in PRX Quantum on 5 June 2026, the work demonstrates a chip that uses quantum physics not only to produce random numbers but to verify the integrity of its own measurement hardware in real time, providing security guarantees that hold even against a quantum-equipped attacker. This advance could strengthen security infrastructure across fields that depend on certified randomness, from cryptography and financial services to artificial intelligence (AI) and healthcare. NUS spin-off Squareroot8 Technologies, which specialises in quantum communication, contributed to the protocol design and security certification.
Equipped with self-checking abilities
Today's QRNGs operate on what is known as a "trusted-device" model, where every component, from the laser to the modulator to the detector, is assumed to match its theoretical blueprint. The approach works much like trusting a set of weighing scales without ever placing a known weight on them — if the scales drift, there is no mechanism to catch the error.
The team's chip removes this blind spot. It implements a measurement-device-independent (MDI) protocol, which means the user needs to trust only the light signals being sent into the device, not the detector that reads them out. In each run, the chip prepares a set of known quantum light states (the one part of the system the user does trust) and measures them using an on-chip optical detector whose behaviour is not assumed. A scoring rule then compares the detector's output against what quantum theory predicts, effectively placing a "known weight on the scales" every time the device operates. If the scores check out, the raw data is distilled into certified random bits. If they do not, the protocol halts, and no randomness is released.
"The measurement unit in quantum random number generators has traditionally been very difficult to characterise, making its real-world reliability hard to guarantee. Our solution removes the need to trust that this unit is operating as specified during use," said Associate Professor Charles Lim.
Built for real-world application
The chip packs both the signal encoder and the optical detector onto a single silicon platform, fabricated using an eight-inch wafer process — the same kind of production line that produces commercial semiconductor devices. Unlike quantum systems that need cryogenic cooling or specialised single-photon detectors, this one operates at room temperature.
Getting the required precision out of silicon posed its own challenge. In silicon-based light modulators, adjusting the timing of a light wave also changes its brightness, as though turning a radio's tuning dial also moved the volume knob. Left unaddressed, this coupling could distort the quantum states and open security loopholes. The team devised a driving scheme that exploits the modulator's own nonlinear response to cancel the effect, keeping the light's brightness constant while its phase is shifted.
The on-chip detector achieved a total efficiency of 69.1 per cent, above the protocol's minimum threshold of 67 per cent. In each experimental run, the chip produced more certified random bits than it consumed as input seed, confirming that the device generates genuinely fresh randomness rather than recycling existing random numbers.
Highest chip-level security demonstrated to date
The chip achieves the highest security level demonstrated to date on a QRNG chip. Its security analysis assumes the worst-case scenario: an adversary who may hold quantum correlations with the detector itself — a far more serious threat than anything classical testing could guard against.
That security comes at a cost in speed. The experimental rate of 64 bits per second is modest compared to trusted-device QRNGs that exceed 100 gigabits per second. But those devices assume the entire hardware chain is perfectly characterised. The trade-off is inherent to the field: the fewer components a user must trust, the stronger the security guarantee, but the lower the throughput.
The chief constraint on speed is the detector's efficiency. The team has already fabricated improved photodiodes in the lab that reach 92.4 per cent efficiency, and simulations of the same protocol with these upgraded components project a rate of 68 megabits per second — more than five orders of magnitude above the current experimental figure.
"This chip paves the way towards integrating practical self-testing quantum random number generators into compact, secure systems," added Assoc Prof Lim. "Provable secure randomness matters wherever decisions depend on numbers that cannot be predicted, from AI and financial services to healthcare and the Internet of Things.
