Patent · US Active

Root cause analysis of vulnerability of neural networks to adversarial examples

US11443069B2 · kind B2 · utility

2Cited by

1References

20Claims

0Family size

Assignee

International Business Machines Corporation · US

Inventors

Sijia Liu · Beijing, CN
Quanfu Fan · Somerville, US
Gaoyuan Zhang · Medford, US
Chuang Gan · Cambridge, US

Key dates

Filing date	Sep 3, 2019
Grant date	Sep 13, 2022
Priority date	—
Expiry date	Mar 21, 2041

Classification

Technology area (CPC G)Physics
CPC primaryG06N3/045
WIPO fieldComputer technology
WIPO sectorElectrical engineering

Abstract

An illustrative embodiment includes a method for protecting a machine learning model. The method includes: determining concept-level interpretability of respective units within the model; determining sensitivity of the respective units within the model to an adversarial attack; identifying units within the model which are both interpretable and sensitive to the adversarial attack; and enhancing defense against the adversarial attack by masking at least a portion of the units identified as both interpretable and sensitive to the adversarial attack.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.