Quantum Noise Injection: Enhancing Security for Quantum Computers and AI

Sunday - March 17 2024, 07:13 UTC - 8 months ago

Quantum Noise Injection for Adversarial Defense (QNAD) is a new approach that uses intrinsic quantum noise to counteract attacks on quantum computers and their AI applications. This method introduces crosstalk into the quantum neural network, resulting in 268% higher accuracy during an attack. QNAD will be presented at a conference in 2024 and has the potential to significantly enhance the security of quantum computers.

Quantum computing is a rapidly emerging technology that uses quantum mechanics—the study of how particles behave at the subatomic level—to solve complex computational problems.

One of the most exciting applications of quantum computers is their potential to revolutionize artificial intelligence (AI). With their ability to solve problems exponentially faster than classical computers, quantum computers are expected to greatly improve AI applications in various devices, such as autonomous vehicles.

However, just like classical computers, quantum computers are vulnerable to adversarial attacks. These attacks are designed to disrupt the inference process of AI, which is the ability to make decisions or solve tasks. Imagine someone putting a sticker over a stop sign - an autonomous vehicle may not recognize the stop sign properly and fail to stop, leading to serious consequences. This highlights the need for an extra layer of protection for quantum computers and their AI applications.

To address this issue, a team of researchers from the University of Texas at Dallas, along with an industry collaborator, have developed a solution called Quantum Noise Injection for Adversarial Defence (QNAD). This approach leverages the intrinsic quantum noise and crosstalk in quantum computers to counteract adversarial attacks. In simple terms, the noisy behaviour of quantum computers actually reduces the impact of attacks.

QNAD works by introducing crosstalk into the quantum neural network (QNN), a form of machine learning that uses large datasets to train computers for tasks such as detecting objects in images. This allows the noisy behaviour of the quantum computer to actually improve the accuracy of the QNN, making it more robust against adversarial attacks.

The team will be presenting their research at the IEEE International Symposium on Hardware Oriented Security and Trust in May 2024. Their studies have shown that during an attack, applying QNAD to AI applications resulted in 268% higher accuracy compared to without it. This marks a significant improvement in the security of quantum computers and their AI applications, which will be crucial for their widespread adoption in the future.

The researchers believe that QNAD is a first-of-its-kind approach that can supplement other defenses against adversarial attacks. It provides an effective and efficient way to enhance the security of quantum computers and their AI applications, paving the way for a safer and more secure future.