An extensible artificial intelligence attack benchmarking method and system

By using a model conversion module and a unified attack algorithm, the limitations of models and language in existing AI attack frameworks are overcome, enabling flexible attacks and efficient testing of different models, and simplifying result display and integration of new methods.

CN117312119BActive Publication Date: 2026-07-03北京银联金卡科技有限公司
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
北京银联金卡科技有限公司
Filing Date
2022-06-17
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing AI attack frameworks or systems cannot simultaneously support NLP models and image recognition models. They suffer from model and language limitations, low automation, difficulty in quickly integrating new attack methods, and inconvenient result display.

Method used

This paper presents an extensible AI attack benchmark testing method and system. The model under test is converted into a converted model that matches the attack framework through a model conversion module. A unified attack algorithm is used to generate scripts for attack. The system combines a visual interface and an extended front-end to realize flexible conversion of models and frameworks and display of results.

Benefits of technology

It enables flexible conversion and attack of different artificial intelligence models, improves attack efficiency, reduces testing costs, solves the problem of model and language limitations, and simplifies result display and integration of new attack methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117312119B_ABST
    Figure CN117312119B_ABST
Patent Text Reader

Abstract

This invention provides an scalable AI attack benchmark testing method and system. A model conversion module transforms the tested model according to its type and framework, obtaining a converted model that matches the attack framework. The dataset attacks the converted model, yielding a test result. The attack framework invokes various attack algorithms with a standardized format, generates corresponding attack scripts, and then performs several attacks on the converted model, resulting in multiple attack results. When testing the next tested model, the model conversion module again transforms its framework according to its type before launching the attack. The advantages are: it can flexibly convert for different AI models, attacking the converted model to obtain attack results, without needing to rebuild test modules for different AI models, improving attack efficiency, reducing testing costs, and solving the problem of existing testing methods (frameworks) being limited by the tested model and language.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Method and system for testing robustness of artificial intelligence model

    CN112766315A

  • Deep learning model defense method aiming at adversarial attack and deep learning model

    CN113127857A