Several studies have shown that quantum neural networks are susceptible to adversarial attacks, which pose a significant threat to the accuracy of quantum neural networks. Adversarial attacks may exploit the input data of a network by feeding incorrect data to a model while manipulating one that has already been trained, which may cause erroneous results. In this research, we propose a novel defense model to protect quantum neural networks against adversarial attacks by using homomorphic data encryption. Homomorphic encryption allows computations to be processed on encrypted data that has not yet been decrypted by converting the data into ciphertext, which allows the data to be handled in its original form without any risks of privacy breach. By incorporating homomorphic data encryption into quantum neural networks, our proposed model tries to reduce adversarial attacks that may perturb network outputs. Taken together, our approach sheds light on the future of quantum computing and the preservation of sensitive information.
Several studies have shown that quantum neural networks are susceptible to adversarial attacks, which pose a significant threat to the accuracy of quantum neural networks. Adversarial attacks may exploit the input data of a network by feeding incorrect data to a model while manipulating one that has already been trained, which may cause erroneous results. In this research, we propose a novel defense model to protect quantum neural networks against adversarial attacks by using homomorphic data encryption. Homomorphic encryption allows computations to be processed on encrypted data that has not yet been decrypted by converting the data into ciphertext, which allows the data to be handled in its original form without any risks of privacy breach. By incorporating homomorphic data encryption into quantum neural networks, our proposed model tries to reduce adversarial attacks that may perturb network outputs. Taken together, our approach sheds light on the future of quantum computing and the preservation of sensitive information.
Abstract It takes a lot of computational resources to train a machine learning model. So does quantum machine learning. In the NISQ (Noisy Intermediate Scale Quantum) era, there is a need for users who cannot afford quantum computers to utilize quantum servers to complete quantum machine learning with protection privacy of data and model parameters. In this paper, novel homomorphic encryption methods and circuit design with hidden parameter for R z , R x and R y gate are putting forward. Based on which, we further propose quantum neural network (QNN) with privacy protection of input data and training parameters, where hiding model parameters has hardly mentioned in existing literature. To verify the validity of the proposed scheme, we conducted experiments on MNIST and Iris datasets, and obtained losses and accuracy that are almost identical to the original algorithms. Experiments show that our scheme can protect privacy well in the training and have no effect on accuracy of the algorithm. By comparison, the proposed scheme is more secure than most existing literature and requires less extra complexity O(4n).
Quantum machine learning in cloud environments requires protecting sensitive data while enabling remote computation. Here we demonstrate the first realistic implementations of a perfectly-secure quantum homomorphic encryption (QHE) scheme applied to quantum neural networks (QNN). Using efficient Clifford+T decomposition, we implement quantum convolutional neural networks for two complementary scenarios: (i) reverse delegated training, where encrypted data from multiple providers trains a user's network via federated aggregation; (ii) private inference, where users process encrypted data with remote quantum networks. Moreover, analysis of server circuit privacy reveals probabilistic model protection through Pauli gate concealment. These results establish perfectly-secure QHE as a practical framework for multi-party quantum machine learning.
Abstract Quantum machine learning is expected to utilize the potential advantages of quantum computing to advance the efficiency of machine learning. However, with the help of quantum cloud servers, ordinary users may confront the threat of privacy leakage of input data and models when performing the training or inference of quantum neural networks (QNNs). To address this problem, we present a new framework that allows the training and inference of delegated QNNs to be performed on encrypted data to protect the privacy of users’ data and models. This framework contains two models that are alternately trained: an encryptor and a predictor. The classical client first trains the encryptor defined by a classical neural network to map plaintext input data to vastly different ciphertext data. The ciphertext data is sent to the quantum cloud server to train the predictor defined by a QNN, which can indirectly predict the labels of plaintext data. With the trained encryptor and predictor, the client can send the encrypted data to the server for prediction and obtain almost equivalent prediction results. The proposed framework is applied to three types of QNN models, each dealing with low-dimensional tabular data, image data, and one-dimensional time series data, respectively. Experimental results show that the privacy protection method based on our framework can protect data and model privacy without degrading the performance of QNNs. The framework does not require users to have quantum capabilities and is suitable for protecting data and model privacy for various QNN models.
With the rapid advancement of artificial intelligence (AI) and quantum computing, cybersecurity threats have evolved, giving rise to adversarial quantum attacks. These attacks exploit the vulnerabilities of AI models using quantum algorithms, posing a significant risk to data security, model robustness, and decision-making systems. This paper presents a comprehensive review of adversarial quantum attacks on AI, analyzing their mechanisms, potential impacts, and countermeasures. It explores how quantum computing can enhance adversarial attacks by accelerating the generation of adversarial examples, breaking cryptographic protections, and undermining AI model integrity. Additionally, the study examines different attack vectors, including quantum-enhanced adversarial perturbations, quantum machine learning (QML) vulnerabilities, and quantum decryption threats. The paper also discusses defensive strategies such as quantum-resistant AI models, quantum cryptographic defenses, and hybrid quantum-classical security frameworks that can mitigate these risks. By evaluating existing research and emerging trends, this review provides insights into the growing intersection of AI security and quantum computing. The findings emphasize the urgent need for robust quantum-aware AI security frameworks to safeguard AI-driven systems in the quantum era. Future research directions include developing quantum-adaptive AI models, post-quantum cryptographic techniques, and real-world applications of quantum-safe AI architectures. This review aims to contribute to the ongoing discourse on ensuring AI resilience against adversarial quantum threats, paving the way for a secure and quantum-resistant future.
Quantum computing does present a big threat to classic cryptography and hence endangers the security of Internet of Things devices. This paper is therefore concerned with proposing a Quantum-Resistant Homomorphic Encryption (QRHE) system tailored for Internet of Things (IoT) environments. The main view of this QRHE key system is basically protection against the quantum threat in the processing of information within Internet of Things network traffic. Aside from this, the system further allows the processing of data on encrypted information without prior decryption, which guarantees the confidentiality and integrity of the data processed. The lattice-based cryptography used in the system is based on the Learning With Errors (LWE) problem, which has already shown strength against classical and quantum attacks. In this paper, homomorphic encryption algorithm was introduced that allows both addition and multiplication between ciphertexts for the assurance of privacy during secure data aggregation and analysis. The experimental results demonstrated that even after several homomorphic operations, the proposed system maintained high accuracy of %98, proving its effectiveness in preserving data confidentiality and integrity. Although the computational cost for this proposed system was a little more compared to traditional methods, it still gave an all-rounded security solution suitable for Internet of Things applications in the quantum computing era.
An Individual in the present time is surrounded with advanced computing technology in several forms. It is expected that in the very near future, Quantum Computers will be replacing the fastest computer systems of present times and will be having enough computational power to bring down the most prevailing cryptographic algorithm. The information that is being shared between different users on the Internet will be at risk in that particular scenario even after significant changes, like increasing the key size or using more complicated mathematical calculations for encrypting the text, which won’t be able to protect the integrity of the system. Several types of research are being carried out to protect the data and make systems resilient to the assaults made by the Quantum Computers of the Future. In this Research, a dynamic type of cryptography is discussed that has a lot of potential to overcome this situation in Post Quantum Scenario, namely Adversarial Neural Cryptography. Adversarial Neural Cryptography works on the architecture of Generative Adversarial Networks where two Neural Networks try to communicate with each other and protect the communication channel and a third Neural Network tries to compromise the connection. The conducted research also explores the computing applicability of the presented notion and its effect in the post-quantum Scenario.