arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.11130cs.CRcs.LGstat.ML
收藏

部署情感分析模型的系统攻击面缩减

Systematic Attack Surface Reduction For Deployed Sentiment Analysis Models

Auburn University(奥本大学);PeopleTec, Inc(PeopleTec公司)

Josh Kalin, David Noever, Gerry Dozier

摘要:

这项工作提出了一种结构化的方法来确定模型的基线,识别攻击向量,并在部署后保护机器学习模型。这种在部署后保护每个模型的方法称为BAD(构建、攻击和防御)架构。BAD架构的两个实现进行评估,以量化的黑盒情感分析系统的对抗生存周期。作为一个具有挑战性的诊断,Jigsaw Toxic Bias数据集被选为我们性能工具的基线。架构的每个实现都将构建一个基准性能报告,攻击一个常见的弱点,并防御传入的攻击。作为一个重要的注意事项:本工作中演示的每个攻击表面都是可检测和可预防的。我们的目标是展示一种可行的方法,用于在生产环境中保护机器学习模型。

英文摘要:

This work proposes a structured approach to baselining a model, identifying attack vectors, and securing the machine learning models after deployment. This method for securing each model post deployment is called the BAD (Build, Attack, and Defend) Architecture. Two implementations of the BAD architecture are evaluated to quantify the adversarial life cycle for a black box Sentiment Analysis system. As a challenging diagnostic, the Jigsaw Toxic Bias dataset is selected as the baseline in our performance tool. Each implementation of the architecture will build a baseline performance report, attack a common weakness, and defend the incoming attack. As an important note: each attack surface demonstrated in this work is detectable and preventable. The goal is to demonstrate a viable methodology for securing a machine learning model in a production setting.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑