Breakthroughs in machine learning (ML) and advances in quantum computing (QC) drive the interdisciplinary field of quantum machine learning to new levels. However, due to the susceptibility of ML models to adversarial attacks, practical use raises safety-critical concerns. Existing Randomized Smoothing (RS) certification methods for classical machine learning models are computationally intensive. In this paper, we propose the combination of QC and the concept of discrete randomized smoothing to speed up the stochastic certification of ML models for discrete data. We show how to encode all the perturbations of the input binary data in superposition and use Quantum Amplitude Estimation (QAE) to obtain a quadratic reduction in the number of calls to the model that are required compared to traditional randomized smoothing techniques. In addition, we propose a new binary threat model to allow for an extensive evaluation of our approach on images, graphs, and text.Breakthroughs in machine learning (ML) and advances in quantum computing (QC) drive the interdisciplinary field of quantum machine learning to new levels. However, due to the susceptibility of ML models to adversarial attacks, practical use raises safety-critical concerns. Existing Randomized Smoothing (RS) certification methods for classical machine learning models are computationally intensive. In this paper, we propose the combination of QC and the concept of discrete randomized smoothing to speed up the stochastic certification of ML models for discrete data. We show how to encode all the perturbations of the input binary data in superposition and use Quantum Amplitude Estimation (QAE) to obtain a quadratic reduction in the number of calls to the model that are required compared to traditional randomized smoothing techniques. In addition, we propose a new binary threat model to allow for an extensive evaluation of our approach on images, graphs, and text.

arXiv:2408.00895IEEE International Conference on Quantum Computing and Engineering (QCE)Sep 15, 20242 citations

Simulation of Quantum Computation

Tzvetan S. Metodi, Frederic T. Chong

Synthesis lectures on computer architectureJan 1, 20060 citations

A Com parison of Classical and Quantum Implement ations of a Random Number Generator: An Introduction to Quantum Computing

Charlotte Coffin

Quantum computers have potential to contribute to advancements in many disciplines including chemistry, material science, and artificial intelligence. As quantum computers become more accessible, there is a greater need for people to understand quantum computation to spearhead such advancements. This thesis discusses the basics of quantum computing, including fundamental concepts and a literature review. In order to demonstrate how quantum computation works, a random number generator is implemented on a quantum computer simulator and compared to a random number generator implemented on a classical computer. The results of the two experiments are compared. The results show that a quantum random number generator creates truly random numbers. This work contributes to the basic understanding of quantum computing by computer science novice learners, thus opening quantum computing to further exploration. 2Quantum computers have potential to contribute to advancements in many disciplines including chemistry, material science, and artificial intelligence. As quantum computers become more accessible, there is a greater need for people to understand quantum computation to spearhead such advancements. This thesis discusses the basics of quantum computing, including fundamental concepts and a literature review. In order to demonstrate how quantum computation works, a random number generator is implemented on a quantum computer simulator and compared to a random number generator implemented on a classical computer. The results of the two experiments are compared. The results show that a quantum random number generator creates truly random numbers. This work contributes to the basic understanding of quantum computing by computer science novice learners, thus opening quantum computing to further exploration. 2

Jan 1, 20180 citations

Quantum randomized encoding, verification of quantum computing, no-cloning, and blind quantum computing

Tomoyuki Morimae

Randomized encoding is a powerful cryptographic primitive with various applications such as secure multiparty computation, verifiable computation, parallel cryptography, and complexity lower bounds. Intuitively, randomized encoding $\hat{f}$ of a function $f$ is another function such that $f(x)$ can be recovered from $\hat{f}(x)$ , and nothing except for $f(x)$ is leaked from $\hat{f}(x)$ . Its quantum version, quantum randomized encoding, has been introduced recently [Brakerski and Yuen, arXiv:2006.01085]. Intuitively, quantum randomized encoding $\hat{F}$ of a quantum operation $F$ is another quantum operation such that, for any quantum state $\rho$ , $F(\rho)$ can be recovered from $\hat{F}(\rho)$ , and nothing except for $F(\rho)$ is leaked from $\hat{F}(\rho)$ . In this paper, we show three results. First, we show that if quantum randomized encoding of BB84 state generations is possible with an encoding operation $E$ , then a two-round verification of quantum computing is possible with a classical verifier who can additionally do the operation $E$ . One of the most important goals in the field of the verification of quantum computing is to construct a verification protocol with a verifier as classical as possible. This result therefore demonstrates a potential application of quantum randomized encoding to the verification of quantum computing: if we can find a good quantum randomized encoding (in terms of the encoding complexity), then we can construct a good verification protocol of quantum computing. Our second result is, however, to show that too good quantum randomized encoding is impossible: if quantum randomized encoding for the generation of even simple states (such as BB84 states) is possible with a classical encoding operation, then the no-cloning is violated. Finally, we consider a natural modification of blind quantum computing protocols in such a way that the server gets the output like quantum randomized encoding. We show that the modified protocol is not secure.Randomized encoding is a powerful cryptographic primitive with various applications such as secure multiparty computation, verifiable computation, parallel cryptography, and complexity lower bounds. Intuitively, randomized encoding $\hat{f}$ of a function $f$ is another function such that $f(x)$ can be recovered from $\hat{f}(x)$, and nothing except for $f(x)$ is leaked from $\hat{f}(x)$. Its quantum version, quantum randomized encoding, has been introduced recently [Brakerski and Yuen, arXiv:2006.01085]. Intuitively, quantum randomized encoding $\hat{F}$ of a quantum operation $F$ is another quantum operation such that, for any quantum state $\rho$, $F(\rho)$ can be recovered from $\hat{F}(\rho)$, and nothing except for $F(\rho)$ is leaked from $\hat{F}(\rho)$. In this paper, we show three results. First, we show that if quantum randomized encoding of BB84 state generations is possible with an encoding operation $E$, then a two-round verification of quantum computing is possible with a classical verifier who can additionally do the operation $E$. One of the most important goals in the field of the verification of quantum computing is to construct a verification protocol with a verifier as classical as possible. This result therefore demonstrates a potential application of quantum randomized encoding to the verification of quantum computing: if we can find a good quantum randomized encoding (in terms of the encoding complexity), then we can construct a good verification protocol of quantum computing. Our second result is, however, to show that too good quantum randomized encoding is impossible: if quantum randomized encoding for the generation of even simple states (such as BB84 states) is possible with a classical encoding operation, then the no-cloning is violated. Finally, we consider a natural modification of blind quantum computing protocols in such a way that the server gets the output like quantum randomized encoding. We show that the modified protocol is not secure.

arXiv:2011.03141Quantum Information and ComputationSep 1, 20211 citation

Alternative Models of Quantum Computation

Colin P. Williams

Texts in computer scienceDec 6, 20100 citations

\hat{f}

of a function

f

is another function such that

f(x)

can be recovered from

\hat{f}(x)

, and nothing except for

f(x)

is leaked from

\hat{f}(x)

. Its quantum version, quantum randomized encoding, has been introduced recently [Brakerski and Yuen, arXiv:2006.01085]. Intuitively, quantum randomized encoding

\hat{F}

of a quantum operation

F

is another quantum operation such that, for any quantum state

\rho

F(\rho)

can be recovered from

\hat{F}(\rho)

, and nothing except for

F(\rho)

is leaked from

\hat{F}(\rho)

. In this paper, we show three results. First, we show that if quantum randomized encoding of BB84 state generations is possible with an encoding operation

E

, then a two-round verification of quantum computing is possible with a classical verifier who can additionally do the operation

E

. One of the most important goals in the field of the verification of quantum computing is to construct a verification protocol with a verifier as classical as possible. This result therefore demonstrates a potential application of quantum randomized encoding to the verification of quantum computing: if we can find a good quantum randomized encoding (in terms of the encoding complexity), then we can construct a good verification protocol of quantum computing. Our second result is, however, to show that too good quantum randomized encoding is impossible: if quantum randomized encoding for the generation of even simple states (such as BB84 states) is possible with a classical encoding operation, then the no-cloning is violated. Finally, we consider a natural modification of blind quantum computing protocols in such a way that the server gets the output like quantum randomized encoding. We show that the modified protocol is not secure.