Patent · US Active

Detecting trojan neural networks

US11394742B2 · kind B2 · utility

0Cited by
1References
20Claims
0Family size

Assignee

Inventors

Key dates

Filing dateAug 17, 2020
Grant dateJul 19, 2022
Priority date
Expiry dateFeb 27, 2041

Classification

  • Technology area (CPC H)Electricity
  • CPC primaryH04L63/145
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

One or more computer processors generate a plurality of adversarial perturbations associated with a model, wherein the plurality of adversarial perturbations comprises a universal perturbation and one or more per-sample perturbations. The one or more computer processors identify a plurality of neuron activations associated with the model and the plurality of generated adversarial perturbations. The one or more computer processors maximize the identified plurality of neuron activations. The one or more computer processors determine the model is a Trojan model by leveraging one or more similarities associated with the maximized neuron activations and the generated adversarial perturbations.

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