Quantum random number generators should be monitored, attested and designed to show where their entropy comes from, rather than relying on claims that their outputs merely look random, according to a new European Telecommunications Standards Institute report.
The report, published in March by ETSI’s Cyber Security technical committee, offers implementation guidelines for quantum random number generators, or QRNGs. These devices use quantum processes, such as the path of a photon through a beam splitter or fluctuations measured in an optical vacuum, to produce unpredictable numbers used in cryptographic keys and other security systems.
ETSI’s central message is that quantum random number generators are not a simple property a vendor can declare. A device may produce output that passes statistical tests while still containing patterns, bias or information an attacker could use to improve predictions. The report calls for systems that verify the quantum source, measure their entropy during operation, protect against tampering and document how random data reaches the applications that consume it.
That matters as governments and companies prepare cryptographic systems for a period when quantum computers may threaten some existing public-key encryption. Post-quantum cryptography is designed to resist such attacks, but its security still depends on unpredictable keys, nonces and other random values. A weak quantum random number generator can undermine otherwise strong encryption.
“Unpredictable” and “uniform” are not the same thing, the report says. A stream of bits may contain roughly equal numbers of zeros and ones, making it appear balanced in a basic test, yet remain partially predictable to someone who has access to information about the device or its environment.
A Quantum Source is Only the Starting Point
A QRNG has two main parts, according to the report. The first is a quantum entropy source, which measures a process governed by quantum mechanics. The second is a classical randomness extractor, which processes the raw measurements into a shorter output intended to be close to a uniform, independent bit stream.
In a simple photon-based system, a light source sends photons toward a beam splitter. Each photon can take one of two paths, and detectors convert those outcomes into zeros and ones. Other systems measure vacuum fluctuations, the small quantum effects that remain even in apparently empty space. More experimental approaches use entanglement, seeking to certify randomness with fewer assumptions about the devices involved.
But practical hardware is not ideal as detectors can drift, electronics can introduce noise, a sensor may be biased toward one result, and components may fail to recover fully before the next measurement. Those effects can leave patterns in a raw stream even when the underlying quantum event is unpredictable.
The extractor is meant to remove those imperfections. It takes raw output that may be slightly biased or correlated and produces fewer bits with stronger statistical properties. The report notes that a raw output in which one appears 51% of the time and zero 49% of the time is unsuitable for cryptographic use, even if the measurements came from a quantum source.
This goes beyond a statistical technicality, according to ETSI. A cryptographic QRNG should account for “side information” — information outside the control of the manufacturer that could be available to an adversary. That can include noise in the device, behavior of its sensors or outside interference.
The report uses the concept of conditional min-entropy to describe this problem. In plain language, it measures how difficult it is for an attacker to guess the next output when the attacker has relevant outside information. A device should not be judged solely by whether its output appears random to a user without that information.
Statistical test suites might be limited, according to the report. Passing a standard randomness test can help detect faults and obvious bias, but it does not by itself prove that an output is unpredictable to an attacker. ETSI says vendors should support their security claims with a physical model explaining how the quantum source works, which operating conditions affect it and how much usable entropy it generates.
Monitoring Quantum Random Number Generators
The report recommends that QRNGs estimate conditional min-entropy while operating, rather than only during laboratory evaluation. That would allow a system to determine whether the entropy source is still producing enough unpredictability for its extractor to meet a target security level.
ETSI also recommends continuous statistical monitoring, generally at a minimum rate of once per second. The system should sample the raw entropy stream, assess short-term patterns over windows such as one second, 10 seconds and one minute, and watch for unusual bias or drops in entropy.
If monitored values fall outside defined limits, the report says the quantum random number generators should log the condition, isolate or pause the affected stream, alert connected systems or administrators, and, where appropriate, switch to a validated fallback source or cache.
The guidance reflects a broader view of QRNG security. A quantum process may be fundamentally unpredictable, but a commercial device includes lasers, detectors, power supplies, processing electronics, firmware and interfaces. Those conventional components can create weaknesses that do not violate quantum mechanics but can still compromise a real product.
For example, photonic QRNGs can be affected by electromagnetic interference. An attacker could potentially use radio-frequency injection to manipulate components or create subtle output changes. Temperature swings, vibration, aging components, laser fluctuations and external magnetic fields can also affect a system, depending on its design.
The report says these risks warrant shielding, tamper resistance, self-diagnostics and environmental monitoring. It also points to the growing possibility that machine-learning systems could analyze large amounts of output and side-channel data, such as power use or electromagnetic emissions, to find correlations a human analyst might miss.
Such attacks would not “break” quantum physics. Instead, they would seek predictable behavior in the classical parts of the system. The recommended defenses include better hardware design, stronger extraction methods and continuous monitoring for small biases or correlations.
A Zero-Trust Approach
ETSI proposes an “Entropy Zero Trust,” or EZT, model for QRNGs used in regulated or high-security settings. The proposed approach treats randomness as a security supply chain that needs evidence at each stage.
The first layer is a quantum entropy source backed by measurable physics, such as photon paths or vacuum fluctuations. The second is real-time health monitoring. The third is a hardware root of trust, such as a trusted platform module or other protected component that can support secure boot, cryptographic attestation and firmware protections.
The fourth layer covers the route from the QRNG to the rest of the system. ETI recommends authenticated, auditable paths using hardened drivers, access controls and secure interfaces. The fifth addresses cloud and other multitenant systems, where separate users may need logically isolated entropy pools, logs and failure boundaries.
The report also recommends that QRNG output carry provenance information. A system should be able to show which source generated entropy, when it was produced, which firmware was in use and which processing path handled it. In high-assurance environments, that evidence could help users determine whether the random data entering a hardware security module, certificate authority or encryption system is genuinely from the intended quantum source.
ETSI draws a line between QRNGs and physical random-number generators that rely on thermal noise, semiconductor instability or classical chaos. Such devices may be useful entropy sources, but the report says they should not be presented as quantum systems unless their quantum contribution is measurable, isolated from classical noise and validated through suitable integrity procedures.
The report also cautions against broad labels such as “quantum-enhanced” without supporting evidence.
Standards Gap Remains
The report is guidance, not a binding technical specification. It identifies several areas where ETSI and the wider industry may need more formal standards, including device interfaces, commands, performance classes, logging and runtime attestation.
Today’s QRNG products often use proprietary interfaces, the report says. That can make it harder for customers to switch vendors, integrate devices into cryptographic systems or build redundant deployments. ETSI outlines possible common controls for enabling and disabling devices, setting output rates, requesting one-time or continuous output, configuring monitoring and establishing logging channels.
It also proposes classifications based on throughput, power, size and weight. Suggested throughput categories range from 100 kilobits per second or less for embedded systems to at least 1 gigabit per second for high-performance computing, cloud data centers and real-time quantum-key-distribution systems.
The report acknowledges that QRNGs still face cost, power and integration constraints. Classical hardware generators and software-based pseudo-random generators remain easier to deploy at scale, while QRNGs depend on specialized components and measurement systems. High-throughput uses may continue to rely on hybrid designs that use quantum entropy to seed or strengthen conventional deterministic generators.
