NNrepair: constraint-based repair of neural network classifiers

Muhammad Usman, Divya Gopinath, Youcheng Sun, Yannic Noller, Corina S. Pasareanu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

33 Citations (Scopus)
27 Downloads (Pure)

Abstract

We present NNREPAIR, a constraint-based technique for repairing neural network classifiers. The technique aims to fix the logic of the network at an intermediate layer or at the last layer. NNREPAIR first uses fault localization to find potentially faulty network parameters (such as the weights) and then performs repair using constraint solving to apply small modifications to the parameters to remedy the defects. We present novel strategies to enable precise yet efficient repair such as inferring correctness specifications to act as oracles for intermediate layer repair, and generation of experts for each class. We demonstrate the technique in the context of three different scenarios: (1) Improving the overall accuracy of a model, (2) Fixing security vulnerabilities caused by poisoning of training data and (3) Improving the robustness of the network against adversarial attacks. Our evaluation on MNIST and CIFAR-10 models shows that NNREPAIR can improve the accuracy by 45.56% points on poisoned data and 10.40% points on adversarial data. NNREPAIR also provides small improvement in the overall accuracy of models, without requiring new data or re-training.
Original languageEnglish
Title of host publication33rd International Conference on Computer-Aided Verification (CAV): proceedings
Editors Alexandra Silva, K. Rustan M. Leino
PublisherSpringer
Chapter1
Pages3-25
VolumePart 1
ISBN (Electronic)9783030816858
ISBN (Print)9783030816841
DOIs
Publication statusPublished - 15 Jul 2021
Event33rd International Conference, CAV 2021 -
Duration: 20 Jul 202123 Jul 2021

Publication series

Name Lecture Notes in Computer Science
Volume12759
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference33rd International Conference, CAV 2021
Period20/07/202123/07/2021

Fingerprint

Dive into the research topics of 'NNrepair: constraint-based repair of neural network classifiers'. Together they form a unique fingerprint.

Cite this