Semantics Preserving Adversarial Learning

2019/03/10 by Ousmane Amadou Dia, Ousmane Dia, Elnaz Barshan +4 · 1 citation
Computer Science · Mathematics · #Adversarial Robustness in Machine Learning #Anomaly Detection Techniques and Applications #Digital Media Forensic Detection #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1903.03905

openalex publication_date 2019/03/10 · arxiv created 2019/12/21 · arxiv updated 2019/12/24 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

While progress has been made in crafting visually imperceptible adversarial examples, constructing semantically meaningful ones remains a challenge. In this paper, we propose a framework to generate semantics preserving adversarial examples. First, we present a manifold learning method to capture the semantics of the inputs. The motivating principle is to learn the low-dimensional geometric summaries of the inputs via statistical inference. Then, we perturb the elements of the learned manifold using the Gram-Schmidt process to induce the perturbed elements to remain in the manifold. To produce adversarial examples, we propose an efficient algorithm whereby we leverage the semantics of the inputs as a source of knowledge upon which we impose adversarial constraints. We apply our approach on toy data, images and text, and show its effectiveness in producing semantics preserving adversarial examples which evade existing defenses against adversarial attacks.

Citations

• Auto-Encoding Variational Bayes
• Backpropagation Applied to Handwritten Zip Code Recognition
• Early Visual Concept Learning with Unsupervised Deep Learning
• Conditional Image Synthesis With Auxiliary Classifier GANs
• Understanding Neural Networks through Representation Erasure
• InfoVAE: Information Maximizing Variational Autoencoders
• Adversarial Examples for Evaluating Reading Comprehension Systems
• Symmetric Variational Autoencoder and Connections to Adversarial Learning
• Variational Gaussian Dropout is not Bayesian
• Adversarial Examples Are Not Bugs, They Are Features
• On Certifying Non-uniform Bound against Adversarial Attacks
• Universal adversarial perturbations
• A large annotated corpus for learning natural language inference
• Provable defenses against adversarial examples via the convex outer adversarial polytope

Cited by

• Towards Explainable Fact Checking

Related

• Transcending Adversarial Perturbations: Manifold-Aided Adversarial Examples with Legitimate Semantics
• Constructing Semantics-Aware Adversarial Examples with a Probabilistic Perspective
• Textual Manifold-based Defense Against Natural Language Adversarial Examples
• Generating Natural Adversarial Examples
• On the Geometry of Adversarial Examples