Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2610.01574

Adversarial Robustness in Fake Quantum Simulators

Marc Maußner, Volker Reers

quant-phSubmitted Oct 1, 2026• Announced Oct 2, 2026• 0 citations
arXivHTMLPDF

Abstract

This paper investigates the performance scalability and adversarial robustness of Quantum Machine Learning (QML) models deployed on noise-model-based fake simulators. We conduct a dual-phased study, first benchmarking the computational throughput of Qiskit's Aer simulation engine across varying hardware architectures, and second, evaluating the effectiveness of Projected Gradient Descent (PGD) attacks and adversarial retraining strategies under realistic noise conditions. Our results quantify the runtime tradeoffs and scaling behavior for medium-scale simulations (projected up to 8 qubits) and demonstrate that high adversarial-to-benign retraining ratios (50/50) are essential for achieving practical model robustness for 4-qubit classifiers under realistic noise conditions.

AI Summary

The paper studies adversarial robustness of quantum classifiers under hardware-derived simulated noise, alongside the computational cost of running these experiments through Qiskit Aer and PennyLane. Attacks perturb classical image features before quantum encoding, rather than quantum states directly. Experiments use four-class mathematical-symbol images resized to $8\times8$ pixels, with 200 training and 50 test examples. The classifier has four qubits, 16 strongly entangling layers, data re-uploading, and 384 trainable parameters. White-box projected gradient descent uses an $\ell_\infty$ budget $\epsilon=0.1$, step size $0.01$, and 10 iterations. Across three fake backends, clean accuracy is 92–96%, falling to 12–40% under attack. Retraining with 10% adversarial examples yields 44–52% attacked accuracy. On FakeGuadalupeV2, balanced adversarial/benign retraining reaches 84%, versus 44% for the sparse retraining baseline; gradient regularization reaches 64%, nonlinear preprocessing 24%, and learning-rate scheduling 12%. An ideal-simulator control reaches 92% after balanced retraining. The reported 95% binomial confidence interval for the 84% result is approximately 74–94%. Runtime profiling reports approximately $1.9\times$ multithreading speedup for eight-qubit circuits and projects a $10.86\times$ runtime increase from four to eight qubits. The paper identifies density-matrix simulation costs and the PennyLane–Qiskit interface as computational bottlenecks. Robustness results are confined to small, simulated four-qubit classifiers; eight-qubit scaling is a projection, and transfer of defenses across backend noise profiles is left to future work.

Likely largely AI-generatedIncrementalSmall numerics onlyOverclaimsPossible error

Central claim may not hold: Comparing only 10% and 50/50 adversarial retraining, without defining practical robustness, does not establish the abstract's claim that high retraining ratios are essential. The 84% noisy robust-accuracy result is demonstrated only on FakeGuadalupeV2 with one dataset and one architecture, limiting its generality. The ratio comparison also changes training-set construction and size, confounding the ratio with added adversarial examples and potentially additional optimization work. The two ratio conditions have different undefended adversarial accuracies, 40% and 14%, and the paper does not document a matched ratio-only comparison. Binomial confidence intervals over 50 test examples cannot support the paper's assertion that training-run variability has been ruled out.

Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.

Comments
10 pages, 3 figures

Related papers

Towards Robust Quantum Machine Learning

Saeefa Rubaiyet Nowmi, Jesus Lopez, Md Mahmudul Alam Imon +3 more

Quantum Machine Learning (QML) fuses quantum computing with machine learning to address complex tasks with potential speedups over classical methods. While QML models such as variational quantum classifiers (QVC), quantum multilayer perceptron(QMLP) models and quantum convolutional neural networks (QCNN) have demonstrated promise in applications like image recognition and malware classification, they remain vulnerable to adversarial threats. Inspired by the vulnerabilities in classical ML, this work systematically investigates how attacks-ranging from training-time poisoning and inferencetime evasion (e.g., quantum versions of FGSM and PGD) to deployment-time trigger attacks-manifest in quantum circuits. We explore how classical adversarial strategies adapt to quantum models through parameterized encodings, hybrid architectures, and probabilistic outputs. The poster evaluates different quantum classifiers under adversarial settings, comparing their robustness, trainability, and resilience to perturbations. Defense mechanisms such as randomized encoding, engineered noise channels, and quantum adversarial training are analyzed in relation to specific threat types. In addition, we assess how quantum models behave differently than classical counterparts under attack, and how quantum-specific design constraints can either exacerbate or mitigate these effects. Our ultimate goal is to define and construct robust QML systems that are both theoretically sound and practically secure. By synthesizing results across multiple QML architectures and threat models, this work provides actionable insight into the development of resilient quantum classifiers, paving the way toward trustworthy and scalable quantum machine learning.

IEEE International Conference on Quantum Computing and Engineering (QCE)•Aug 30, 2025•2 citations

A Threat-Model–Driven Robustness Benchmark for Quantum Machine Learning Under Device Noise and Deployment Shift

Volker Reers, Marc Maußner

Quantum machine learning (QML) systems executed on noisy intermediate-scale quantum (NISQ) hardware are vulnerable to both adversarial perturbations and backend-dependent noise effects. Existing robustness evaluations are typically performed under fixed execution conditions and rarely consider deployment shift between training and inference backends. This work proposes a threat-model–driven benchmark framework for evaluating QML robustness under adversarial attacks, stochastic device noise, and backend-dependent deployment variability. Using FGSM and PGD attacks, we analyze robustness scaling under varying training budgets, encoding strategies, and fake quantum backends. Experimental results demonstrate substantial backend-dependent robustness variability and reduced robustness transfer across heterogeneous backend environments. The findings highlight the importance of backend-aware robustness evaluation for realistic NISQ-era QML deployments.

IEEE International Conference on Quantum Control, Computing and Learning (qCCL)•Jun 1, 2026•2 citations

Adversarial Robustness in Quantum Machine Learning: A Scoping Review

Yanche Ari Kustiawan, Khairil Imran Ghauth

Quantum machine learning (QML) is emerging as a promising paradigm at the intersection of quantum computing and artificial intelligence, yet its security under adversarial conditions remains insufficiently understood. This scoping review aims to systematically map empirical research on adversarial robustness in QML and to identify dominant threat models, defense strategies, evaluation approaches, practical constraints, and future research directions. Following PRISMA-ScR guidelines, four major databases were searched, resulting in 53 eligible empirical studies published between 2020 and 2026. The findings show that most research concentrates on input-level evasion attacks, particularly adversarial examples, and primarily evaluates robustness in classification-oriented models such as variational quantum circuits and quantum neural networks. Defense strategies are largely adapted from classical adversarial training and noise-based mitigation, with limited deployment on real quantum hardware. Robustness assessment is predominantly empirical, relying on accuracy degradation and attack success rate, while formal certification methods remain less common. The literature also highlights substantial constraints related to hardware limitations, NISQ noise, computational cost, and dataset scale. Overall, the evidence indicates that adversarial robustness research in QML is expanding but remains methodologically concentrated, underscoring the need for standardized benchmarking, scalable defenses, and hardware-validated robustness evaluation frameworks.

Computers•Apr 9, 2026•3 citations

SoK: Critical Evaluation of Quantum Machine Learning for Adversarial Robustness

Saeefa Rubaiyet Nowmi, Jesus Lopez, Md Mahmudul Alam Imon +2 more

Quantum Machine Learning (QML) integrates quantum computational principles into learning algorithms, offering the potential for improved representational capacity and computational efficiency. Nevertheless, the security and robustness of QML systems remain largely underexplored, particularly under adversarial conditions. We present the first comprehensive systematization of adversarial robustness in QML, integrating conceptual organization with empirical evaluation across black-, gray-, and white-box threat models. We implement five representative attacks across all three threat models: a label-flipping data poisoning attack under blackbox; an encoder-level indiscriminate data poisoning attack and a proxy-model-based clean-label backdoor attack under gray-box; and a circuit-level backdoor attack (QTrojan) and gradient-based evasion attacks (FGSM and PGD) under whitebox. We evaluate the attacks using a Quantum Multilayer Perceptron (QMLP) trained on MNIST and AZ-Class across circuit depths of 2, 5, 10 and 50 layers and two encoding schemes (angle and amplitude). Our extensive evaluations reveal a fundamental accuracy-robustness trade-off. In particular, amplitude encoding yields the highest clean accuracy (92.6 % on MNIST, 67% on AZ-Class); however, it collapses under adversarial perturbations and depolarizing noise, while shallow angle-encoded models remain substantially more stable. In addition, QUID is highly effective under noiseless conditions but is weakened by noise, whereas the proxy-model backdoor persists unless the circuit itself is overwhelmed, highlighting that noise is an asymmetric and unreliable passive defense. Furthermore, the circuit-level backdoor fails in the multi-class setting, indicating a scalability constraint. Finally, QMLP models are more robust than Classical Multi-Layer Perceptron (CMLP) models under label-flipping attacks but are substantially more vulnerable to gradient-based evasion, motivating the need for quantumspecific defenses. We conclude by proposing a threat-aware, noise-resilient design framework for secure and robust QML deployment.

arXiv:2511.14989•IEEE Symposium on Security and Privacy (S&P)•May 18, 2026•5 citations

Adversarial Attacks and Defence Mechanisms in Quantum Machine Learning Systems

Dzulfikar Ahmed, Mohamed Hafiz, SYED ABUTHAHIR +1 more

Quantum Machine Learning (QML) is a new discipline that combines quantum computing and machine learning algorithms in order to take advantage of the computational benefits which are provided by quantum computing for performing complex problems. While similar to classical machine learning systems, quantum-models tend to be prone for adversarial attacks—craft perturbations designed purposely for causing model mispredictions. The intrinsic noise and probabilistic behavior of quantum systems— in particular the (Noisy Intermediate-Scale Quantum) NISQ—is only worsening these vulnerabilities. This paper provides an overview of adversarial threats in QML and illustrates how the quantum nature of any specific system to be attacked shapes both attack and defence strategies. We study on prevalent adversarial methods including gradient based perturbations, optimization attacks and transfer attacks among others, for their resonance in the quantum settings. And, we explore the influence of quantum noise, circuit layout, and hybrid architectures on system weaknesses. In order to develop solutions for such problems, we provide a survey of defines approaches including adversarial training noise-aware learning quantum error mitigation and robust optimization By incorporating probabilistic approaches like Bayesian inference, the ability to make decisions with uncertainty informs and strengthens resilience. Fields like cybersecurity, healthcare and autonomous systems showcase implications of adversarial robustness in QML. Lastly, the paper presents open challenges and future directions for research where some of them are on building scalable secured interpretable quantum AI systems.

International Journal of Quantum Computing and Artificial Intelligence (IJQCAI)•Jan 1, 2026•0 citations

Verifying Adversarial Robustness in Quantum Machine Learning: from theory to physical validation via a software tool

Ji Guan, Mingsheng Ying

As with classical neural networks, quantum machine learning (QML) models are vulnerable to small input perturbations that can significantly alter output predictions. Certifying the robustness of QML models, particularly on NISQ hardware, is therefore a fundamental step toward trustworthy quantum AI. This chapter reviews our recently developed comprehensive formal framework for verifying adversarial robustness in QML. The core of this framework is a fidelity-based robustness lower bound computable directly from the measurement outcome distribution, which enables both formal verification and empirical estimation on real quantum devices. Additionally, the optimal bound can be computed via semidefinite programming (SDP) with full knowledge of the quantum machine learning models. We incorporate these results into: (1) an efficient formal verification framework; (2) VeriQR, the first dedicated QML robustness verification tool; and (3) the first experimental benchmark of quantum adversarial robustness on a 20-qubit superconducting processor. Together, these systematic advances enable scalable, physically grounded robustness evaluation of QML models.

quant-pharXiv:2605.29877•Quantum science and technology•May 29, 2026•0 citations
This paper investigates the performance scalability and adversarial robustness of Quantum Machine Learning (QML) models deployed on noise-model-based fake simulators. We conduct a dual-phased study, first benchmarking the computational throughput of Qiskit's Aer simulation engine across varying hardware architectures, and second, evaluating the effectiveness of Projected Gradient Descent (PGD) attacks and adversarial retraining strategies under realistic noise conditions. Our results quantify the runtime tradeoffs and scaling behavior for medium-scale simulations (projected up to 8 qubits) and demonstrate that high adversarial-to-benign retraining ratios (50/50) are essential for achieving practical model robustness for 4-qubit classifiers under realistic noise conditions.
The paper studies adversarial robustness of quantum classifiers under hardware-derived simulated noise, alongside the computational cost of running these experiments through Qiskit Aer and PennyLane. Attacks perturb classical image features before quantum encoding, rather than quantum states directly.
Experiments use four-class mathematical-symbol images resized to 8×88\times88×8 pixels, with 200 training and 50 test examples. The classifier has four qubits, 16 strongly entangling layers, data re-uploading, and 384 trainable parameters. White-box projected gradient descent uses an ℓ∞\ell_\inftyℓ∞​ budget ϵ=0.1\epsilon=0.1ϵ=0.1, step size 0.010.010.01, and 10 iterations. Across three fake backends, clean accuracy is 92–96%, falling to 12–40% under attack. Retraining with 10% adversarial examples yields 44–52% attacked accuracy. On FakeGuadalupeV2, balanced adversarial/benign retraining reaches 84%, versus 44% for the sparse retraining baseline; gradient regularization reaches 64%, nonlinear preprocessing 24%, and learning-rate scheduling 12%. An ideal-simulator control reaches 92% after balanced retraining. The reported 95% binomial confidence interval for the 84% result is approximately 74–94%.
Runtime profiling reports approximately 1.9×1.9\times1.9× multithreading speedup for eight-qubit circuits and projects a 10.86×10.86\times10.86× runtime increase from four to eight qubits. The paper identifies density-matrix simulation costs and the PennyLane–Qiskit interface as computational bottlenecks. Robustness results are confined to small, simulated four-qubit classifiers; eight-qubit scaling is a projection, and transfer of defenses across backend noise profiles is left to future work.
Quantum Machine Learning (QML) fuses quantum computing with machine learning to address complex tasks with potential speedups over classical methods. While QML models such as variational quantum classifiers (QVC), quantum multilayer perceptron(QMLP) models and quantum convolutional neural networks (QCNN) have demonstrated promise in applications like image recognition and malware classification, they remain vulnerable to adversarial threats. Inspired by the vulnerabilities in classical ML, this work systematically investigates how attacks-ranging from training-time poisoning and inferencetime evasion (e.g., quantum versions of FGSM and PGD) to deployment-time trigger attacks-manifest in quantum circuits. We explore how classical adversarial strategies adapt to quantum models through parameterized encodings, hybrid architectures, and probabilistic outputs. The poster evaluates different quantum classifiers under adversarial settings, comparing their robustness, trainability, and resilience to perturbations. Defense mechanisms such as randomized encoding, engineered noise channels, and quantum adversarial training are analyzed in relation to specific threat types. In addition, we assess how quantum models behave differently than classical counterparts under attack, and how quantum-specific design constraints can either exacerbate or mitigate these effects. Our ultimate goal is to define and construct robust QML systems that are both theoretically sound and practically secure. By synthesizing results across multiple QML architectures and threat models, this work provides actionable insight into the development of resilient quantum classifiers, paving the way toward trustworthy and scalable quantum machine learning.
Quantum machine learning (QML) systems executed on noisy intermediate-scale quantum (NISQ) hardware are vulnerable to both adversarial perturbations and backend-dependent noise effects. Existing robustness evaluations are typically performed under fixed execution conditions and rarely consider deployment shift between training and inference backends. This work proposes a threat-model–driven benchmark framework for evaluating QML robustness under adversarial attacks, stochastic device noise, and backend-dependent deployment variability. Using FGSM and PGD attacks, we analyze robustness scaling under varying training budgets, encoding strategies, and fake quantum backends. Experimental results demonstrate substantial backend-dependent robustness variability and reduced robustness transfer across heterogeneous backend environments. The findings highlight the importance of backend-aware robustness evaluation for realistic NISQ-era QML deployments.
Quantum machine learning (QML) is emerging as a promising paradigm at the intersection of quantum computing and artificial intelligence, yet its security under adversarial conditions remains insufficiently understood. This scoping review aims to systematically map empirical research on adversarial robustness in QML and to identify dominant threat models, defense strategies, evaluation approaches, practical constraints, and future research directions. Following PRISMA-ScR guidelines, four major databases were searched, resulting in 53 eligible empirical studies published between 2020 and 2026. The findings show that most research concentrates on input-level evasion attacks, particularly adversarial examples, and primarily evaluates robustness in classification-oriented models such as variational quantum circuits and quantum neural networks. Defense strategies are largely adapted from classical adversarial training and noise-based mitigation, with limited deployment on real quantum hardware. Robustness assessment is predominantly empirical, relying on accuracy degradation and attack success rate, while formal certification methods remain less common. The literature also highlights substantial constraints related to hardware limitations, NISQ noise, computational cost, and dataset scale. Overall, the evidence indicates that adversarial robustness research in QML is expanding but remains methodologically concentrated, underscoring the need for standardized benchmarking, scalable defenses, and hardware-validated robustness evaluation frameworks.
Quantum Machine Learning (QML) integrates quantum computational principles into learning algorithms, offering the potential for improved representational capacity and computational efficiency. Nevertheless, the security and robustness of QML systems remain largely underexplored, particularly under adversarial conditions. We present the first comprehensive systematization of adversarial robustness in QML, integrating conceptual organization with empirical evaluation across black-, gray-, and white-box threat models. We implement five representative attacks across all three threat models: a label-flipping data poisoning attack under blackbox; an encoder-level indiscriminate data poisoning attack and a proxy-model-based clean-label backdoor attack under gray-box; and a circuit-level backdoor attack (QTrojan) and gradient-based evasion attacks (FGSM and PGD) under whitebox. We evaluate the attacks using a Quantum Multilayer Perceptron (QMLP) trained on MNIST and AZ-Class across circuit depths of 2, 5, 10 and 50 layers and two encoding schemes (angle and amplitude). Our extensive evaluations reveal a fundamental accuracy-robustness trade-off. In particular, amplitude encoding yields the highest clean accuracy (92.6 % on MNIST, 67% on AZ-Class); however, it collapses under adversarial perturbations and depolarizing noise, while shallow angle-encoded models remain substantially more stable. In addition, QUID is highly effective under noiseless conditions but is weakened by noise, whereas the proxy-model backdoor persists unless the circuit itself is overwhelmed, highlighting that noise is an asymmetric and unreliable passive defense. Furthermore, the circuit-level backdoor fails in the multi-class setting, indicating a scalability constraint. Finally, QMLP models are more robust than Classical Multi-Layer Perceptron (CMLP) models under label-flipping attacks but are substantially more vulnerable to gradient-based evasion, motivating the need for quantumspecific defenses. We conclude by proposing a threat-aware, noise-resilient design framework for secure and robust QML deployment.
Quantum Machine Learning (QML) is a new discipline that combines quantum computing and machine learning algorithms in order to take advantage of the computational benefits which are provided by quantum computing for performing complex problems. While similar to classical machine learning systems, quantum-models tend to be prone for adversarial attacks—craft perturbations designed purposely for causing model mispredictions. The intrinsic noise and probabilistic behavior of quantum systems— in particular the (Noisy Intermediate-Scale Quantum) NISQ—is only worsening these vulnerabilities. This paper provides an overview of adversarial threats in QML and illustrates how the quantum nature of any specific system to be attacked shapes both attack and defence strategies. We study on prevalent adversarial methods including gradient based perturbations, optimization attacks and transfer attacks among others, for their resonance in the quantum settings. And, we explore the influence of quantum noise, circuit layout, and hybrid architectures on system weaknesses. In order to develop solutions for such problems, we provide a survey of defines approaches including adversarial training noise-aware learning quantum error mitigation and robust optimization By incorporating probabilistic approaches like Bayesian inference, the ability to make decisions with uncertainty informs and strengthens resilience. Fields like cybersecurity, healthcare and autonomous systems showcase implications of adversarial robustness in QML. Lastly, the paper presents open challenges and future directions for research where some of them are on building scalable secured interpretable quantum AI systems.
As with classical neural networks, quantum machine learning (QML) models are vulnerable to small input perturbations that can significantly alter output predictions. Certifying the robustness of QML models, particularly on NISQ hardware, is therefore a fundamental step toward trustworthy quantum AI. This chapter reviews our recently developed comprehensive formal framework for verifying adversarial robustness in QML. The core of this framework is a fidelity-based robustness lower bound computable directly from the measurement outcome distribution, which enables both formal verification and empirical estimation on real quantum devices. Additionally, the optimal bound can be computed via semidefinite programming (SDP) with full knowledge of the quantum machine learning models. We incorporate these results into: (1) an efficient formal verification framework; (2) VeriQR, the first dedicated QML robustness verification tool; and (3) the first experimental benchmark of quantum adversarial robustness on a 20-qubit superconducting processor. Together, these systematic advances enable scalable, physically grounded robustness evaluation of QML models.