Adversary Detection in Neural Networks via Persistent Homology

2017/11/28 by Gebhart, Thomas, Schrater, Paul · 1 citation
#FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML)

paper · doi:10.48550/arxiv.1711.10056

Abstract

We outline a detection method for adversarial inputs to deep neural networks. By viewing neural network computations as graphs upon which information flows from input space to out- put distribution, we compare the differences in graphs induced by different inputs. Specifically, by applying persistent homology to these induced graphs, we observe that the structure of the most persistent subgraphs which generate the first homology group differ between adversarial and unperturbed inputs. Based on this observation, we build a detection algorithm that depends only on the topological information extracted during training. We test our algorithm on MNIST and achieve 98% detection adversary accuracy with F1-score 0.98.

Cited by

• TopoReformer: Mitigating Adversarial Attacks Using Topological Purification in OCR Models

Related

• A Topology Layer for Machine Learning
• Characterizing the Shape of Activation Space in Deep Neural Networks
• An Adversarial Robustness Perspective on the Topology of Neural Networks
• Topological Detection of Trojaned Neural Networks
• When Witnesses Defend: A Witness Graph Topological Layer for Adversarial Graph Learning