Defence against adversarial attacks using classical and quantum-enhanced Boltzmann machines

Abstract

We provide a robust defence to adversarial attacks on discriminative algorithms. Neural networks are naturally vulnerable to small, tailored perturbations in the input data that lead to wrong predictions. On the contrary, generative models attempt to learn the distribution underlying a dataset, making them inherently more robust to small perturbations. We use Boltzmann machines for discrimination purposes as attack-resistant classifiers, and compare them against standard state-of-the-art adversarial defences. We find improvements ranging from 5% to 72% against attacks with Boltzmann machines on the MNIST dataset. We furthermore complement the training with quantum-enhanced sampling from the D-Wave 2000Q annealer, finding results comparable with classical techniques and with marginal improvements in some cases. These results underline the relevance of probabilistic methods in constructing neural networks and highlight a novel scenario of practical relevance where quantum computers, even with limited hardware capabilites, could provide advantages over classical computers. This work is dedicated to the memory of Peter Wittek.

arXiv comment15 pages, 1 figure. V2: Updated to match published version

References in corpus (26)

• Explaining and Harnessing Adversarial Examples
• Supplementary information for "Quantum supremacy using a programmable superconducting processor"
• Intriguing properties of neural networks
• Ensemble Adversarial Training: Attacks and Defenses
• Quantum computational advantage using photons
• Feature Squeezing: Detecting Adversarial Examples in Deep Neural Networks
• Simulated Tempering: A New Monte Carlo Scheme
• One pixel attack for fooling deep neural networks
• Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples
• Deep Neural Networks as Gaussian Processes
• Generating Natural Adversarial Examples
• On the (Statistical) Detection of Adversarial Examples
• Quantum Boltzmann Machine
• Formal Guarantees on the Robustness of a Classifier against Adversarial Manipulation
• What is the Computational Value of Finite Range Tunneling?
• A practical heuristic for finding graph minors
• Entanglement in a quantum annealing processor
• MagNet and "Efficient Defenses Against Adversarial Attacks" are Not Robust to Adversarial Examples
• Adversarial Examples, Uncertainty, and Transfer Testing Robustness in Gaussian Process Hybrid Deep Networks
• Quantum-Assisted Learning of Hardware-Embedded Probabilistic Graphical Models
• Quantum-assisted Helmholtz machines: A quantum-classical deep learning framework for industrial datasets in near-term devices
• Bayesian Deep Learning on a Quantum Computer
• Benchmarking Quantum Hardware for Training of Fully Visible Boltzmann Machines
• Boltzmann Encoded Adversarial Machines
• Efficient training of energy-based models via spin-glass control
• How Wrong Am I? - Studying Adversarial Examples and their Impact on Uncertainty in Gaussian Process Machine Learning Models

Cited by in corpus (7)

• Towards quantum enhanced adversarial robustness in machine learning
• Benchmarking Adversarially Robust Quantum Machine Learning at Scale
• Drastic Circuit Depth Reductions with Preserved Adversarial Robustness by Approximate Encoding for Quantum Machine Learning
• Predominant Aspects on Security for Quantum Machine Learning: Literature Review
• Adversarial Robustness Guarantees for Quantum Classifiers
• Accelerating the training of single-layer binary neural networks using the HHL quantum algorithm
• Tensorization of neural networks for improved privacy and interpretability