Self-testing with untrusted random number generators
Moisés Bermejo Morán, Ravishankar Ramanathan · 2026
Read Self-testing with untrusted random number generators by Moisés Bermejo Morán, Ravishankar Ramanathan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-10663.txt.
Introduction
Moisés Bermejo Morán and Ravishankar Ramanathan extend quantum self-testing to cases where the random settings generator may not be independent of the device. The question is what can still be trusted when the dice and the device might be linked. Self-testing infers a device's inner physics from black-box behavior, and past protocols assumed freely chosen settings. All pure bipartite partially entangled states can still be self-tested if the generator keeps some residual randomness. That condition is strictly weaker than full independence.
Catalog record
| CREATOR | Moisés Bermejo Morán, Ravishankar Ramanathan |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29495 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.10663
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Self-testing with untrusted random number generators Moisés Bermejo Morán, Ravishankar Ramanathan Abstract Self-testing–the attractive possibility to infer the underlying physics of a quantum device in a black-box scenario–has gained increased traction in recent years, with applications to device-independent quantum information processing. Thus far, self-testing has been done under the assumption that the settings for the requisite Bell test are chosen freely and independently of the device tested in the experiment. That is, the random number generator used to generate the settings has been assumed to have no correlations with the device tested. Here, we extend self-testing protocols beyond the independence assumption. Surprisingly, we show that all pure bipartite partially entangled states can be self-tested provided that the random number generator obeys a residual randomness constraint strictly weaker than the independence assumption. This in itself provides a semi-device-independent certification of independence between the randomness source and the device. Introduction.- The foundational phenomenon of Bell nonlocality has given rise to the possibility of device-independent quantum cryptography, specifically the possibility of obtaining secure key 1 and private, fully random bits for use in cryptographic applications 23 . An attractive feature of these applications is the fact that the users do not need to trust their devices and can instead directly verify the security by means of simple statistical tests on the input-output behavior of the device, specifically checking the violation of a Bell inequality or the success probability in a nonlocal game. This violation certifies, even in a fully adversarial scenario where an eavesdropper has tampered with the device or holds additional systems correlated with the device, that the outputs of the device could not be perfectly predicted by the eavesdropper. A particularly important property of certain Bell inequalities is self-testing, which allows the users to fully determine the underlying quantum state of the device and the measurement performed (up to local isometries), based solely on the input-output behaviour. A famous example is the Clauser-Horne-Shimony-Holt (CHSH) inequality 6 , whose maximal quantum value self-tests the state of the device to be an EPR pair |\phi_{+}\rangle=\left(|00\rangle+|11\rangle\right)/\sqrt{2} as well as particular Pauli measurements performed on the state 15 . Other well-known examples include the family of tilted CHSH inequalities 30 , the magic square game 29 , the tripartite Mermin inequality 17 , families of n -player binary XOR games 16 as well as a host of other interesting examples with foundational and practical implications 5 ; 27 . Self-testing has grown into a highly interesting research field with profound implications for cryptography 1 and quantum computational complexity 19 . The success of these self-testing protocols and practical applications heavily relies on the measurement independence assumption: the choice of measurement settings in the Bell scenario is random and independent of the device. While this assumption can be naturally justified in an ideal scenario by the free will of the people performing the experiment to choose the settings, in realistic applications the settings are chosen with random number generators (RNG). It is therefore against the spirit of device-independence to trust that these sources will be perfectly random and uncorrelated with the device. Indeed, arbitrary correlations between the choice of measurement settings and the state of the device permit to describe all measurement statistics with local hidden variables 2 ; 9 . Additional assumptions on the source are therefore necessary if one expects to extend device-independent protocols to Bell scenarios with untrusted sources of randomness. Under the assumption that the outputs of the random number generators cannot be perfectly predicted with the device, we show that it is possible to self-test any partially entangled bipartite pure state. Our assumption is strictly weaker than measurement independence and is satisfied, in particular, in the device-independent protocols for randomness amplification considered in 28 ; 26 . We finally expose a fundamental distinction between purely probabilistic Bell tests and Hardy-type tests that rely on impossibilities in realistic scenarios: Bell values fail to self-test under arbitrarily weak measurement dependence, whereas Hardy self-tests succeed under arbitrarily weak independence. Bell scenarios with untrusted sources.- A bipartite Bell scenario is an experiment with local input spaces X and Y and output spaces A and B , which possibly depend on some hidden variable \lambda\in\Lambda . The probabilities of joint events a,b,x,y , p(abxy)=\sum_{\lambda}p(\lambda)p(xy|\lambda)p(ab|xy\lambda)\,, (1) are obtained by averaging over all (unobserved) hidden variables \lambda\in\Lambda . Under the assumption of measurement independence , the input variables do not depend on the hidden variable p(xy|\lambda)=p(xy) , and are often assumed to be uniform p(xy)=1/|XY| . Therefore, all information is contained in the conditional distributions p(ab|xy) . These are classical when they decompose into local distributions p(ab|xy)=\sum_{\lambda}p(\lambda)p(a|x\lambda)p(b|y\lambda) . Quantum distributions are obtained as expectation values of local measurement effects under some shared state. That is, p(ab|xy)=\langle A_{a|x}\otimes B_{b|y}\rangle=\tr( \rho A_{a|x} \otimes B_{b|y} )\,, (2) where \rho is a state in a bipartite Hilbert space H_{A}\otimes H_{B} , A_{x}=(A_{a|x})_{a\in A} is a projective measurement on H_{A} (i.e. \sum_{a}A_{a|x}=\mathbf{1}_{A} and A_{a|x}^{2}=A_{a|x} ) and B_{y}=(B_{b|y})_{b\in B} a projective measurement on H_{B} for each x,y . We consider instead a realistic scenario that violates the measurement independence assumption, in which the input variables x\in X and y\in Y are chosen with random number generators S and T . In this scenario, depicted in Figure 1 , the joint events have probabilities p(stabxy)=\sum_{\lambda}p(\lambda)p(xy|st)p(stab|xy\lambda)\,, (3) where p(stab|xy\lambda)=p(st|\lambda)p(ab|stxy\lambda) can be understood as conditional distributions p_{st}(ab|xy\lambda) subnormalized to p(st|\lambda) . Indeed, this setting can be alternatively visualized as a Bell scenario with two additional parties with empty input spaces and output spaces S and T , wired to X and Y . A distribution is classical when it decomposes into local ones, p(stab|xy)=\sum_{\lambda}p(\lambda)p(st|\lambda)p(a|x\lambda)p(b|y\lambda) . Quantum distributions are those obtained in the four-partite Bell scenario with classical-quantum states \rho=\sum_{st}|st\rangle\langle st|\otimes\rho_{st} . That is, with expectation values of local measurement effects on the device under (subnormalized) states of the device classically correlated with the sources, p(stab|xy)=\langle A_{a|x}\otimes B_{b|y}\rangle_{st}=\tr(\rho_{st} A_{a|x} \otimes B_{b|y})\,. (4) We consider device-independent protocols in which we only have access to observable events of the form p(stab|st) , in which the settings are chosen from the sources. We say that the source has residual randomness if p(st|abxy)>0 (5) for each s,t,a,b,x,y . Operationally, this imposes that measurement statistics on the device cannot perfectly discriminate the outputs of the source. In quantum theory, this means that the measurement effects A_{a|x}\otimes B_{b|y} cannot discriminate the states of the device \rho_{st} classically correlated with the source. The assumption that the source has randomness even conditioned upon the quantum state of the device also appears in the Santha-Vazirani condition for boxes 28 or quantum Santha-Vazirani condition 26 considered in the context of randomness amplification. Notice that the residual randomness assumption is strictly stronger than the condition p(st|xy)>0 often considered in measurement dependence scenarios 24 . An immediate consequence of the residual randomness assumption is that the impossibility of events is independent of the source . This condition is strictly weaker than independence for all events, but is enough to extend quantum information protocols based on impossibility beyond the measurement independence scenario. Examples of these protocols include all-versus-nothing proofs of nonlocality 18 and pseudo-telepathy games 3 , which require stronger forms of nonlocality 13 with practical consequences in quantum computing 4