Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cloud-based XAI Services for Assessing Open Repository Models Under Adversarial Attacks

Zerui Wang, Yan Liu

TL;DR

A cloud-based service framework that encapsulates computing components as microser-vices and organizes assessment tasks into pipelines and generates aggregated analysis that showcases the quality attributes of AI models across computer vision and tabular cases is proposed.

Abstract

The opacity of AI models necessitates both validation and evaluation before their integration into services. To investigate these models, explainable AI (XAI) employs methods that elucidate the relationship between input features and output predictions. The operations of XAI extend beyond the execution of a single algorithm, involving a series of activities that include preprocessing data, adjusting XAI to align with model parameters, invoking the model to generate predictions, and summarizing the XAI results. Adversarial attacks are well-known threats that aim to mislead AI models. The assessment complexity, especially for XAI, increases when open-source AI models are subject to adversarial attacks, due to various combinations. To automate the numerous entities and tasks involved in XAI-based assessments, we propose a cloud-based service framework that encapsulates computing components as microservices and organizes assessment tasks into pipelines. The current XAI tools are not inherently service-oriented. This framework also integrates open XAI tool libraries as part of the pipeline composition. We demonstrate the application of XAI services for assessing five quality attributes of AI models: (1) computational cost, (2) performance, (3) robustness, (4) explanation deviation, and (5) explanation resilience across computer vision and tabular cases. The service framework generates aggregated analysis that showcases the quality attributes for more than a hundred combination scenarios.

Cloud-based XAI Services for Assessing Open Repository Models Under Adversarial Attacks

TL;DR

A cloud-based service framework that encapsulates computing components as microser-vices and organizes assessment tasks into pipelines and generates aggregated analysis that showcases the quality attributes of AI models across computer vision and tabular cases is proposed.

Abstract

The opacity of AI models necessitates both validation and evaluation before their integration into services. To investigate these models, explainable AI (XAI) employs methods that elucidate the relationship between input features and output predictions. The operations of XAI extend beyond the execution of a single algorithm, involving a series of activities that include preprocessing data, adjusting XAI to align with model parameters, invoking the model to generate predictions, and summarizing the XAI results. Adversarial attacks are well-known threats that aim to mislead AI models. The assessment complexity, especially for XAI, increases when open-source AI models are subject to adversarial attacks, due to various combinations. To automate the numerous entities and tasks involved in XAI-based assessments, we propose a cloud-based service framework that encapsulates computing components as microservices and organizes assessment tasks into pipelines. The current XAI tools are not inherently service-oriented. This framework also integrates open XAI tool libraries as part of the pipeline composition. We demonstrate the application of XAI services for assessing five quality attributes of AI models: (1) computational cost, (2) performance, (3) robustness, (4) explanation deviation, and (5) explanation resilience across computer vision and tabular cases. The service framework generates aggregated analysis that showcases the quality attributes for more than a hundred combination scenarios.
Paper Structure (15 sections, 5 equations, 11 figures, 8 tables)

This paper contains 15 sections, 5 equations, 11 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Introduction of XAI Methods
  4. XAI Tools Libraries and Frameworks
  5. Adversarial Attacks Types
  6. Overview of AI System's Quality Attributes
  7. Pipelines of XAI Centric Assessment of Quality Attributes
  8. Define Pipelines for Each Quality Attribute
  9. Define Services Architecture for the Pipeline Configuration
  10. Comparative Evaluation and Service Integration of Existing XAI Frameworks
  11. Assessment Scenarios and Quality Attribute Analysis
  12. The Assessment Scenario of Vision Models
  13. The Assessment Scenario of Tabular Models
  14. Conclusions and Insights on Research Questions
  15. Conclusion

Figures (11)

  • Figure 1: Five CAM-based Visual Explanations from Vision Transformer Model with One Image Example.
  • Figure 2: Top 10 out of 83 SHAP Feature Importance Explanations from FT Transformer on RT-IoT Cybersecurity Threats Dataset.
  • Figure 3: Taxonomy of Adversarial Attacks. References: FGSM FGSM, C&W CW, JSMA JSMA, AdvGAN AdvGAN, DeepFool DeepFool, UAP UAP, DaST DaST, Houdini Houdini, ZOO ZOO, One-Pixel One-Pixel, ImageNet-C imagenetC, ImageNet-P imagenetC, Fooling LIME and SHAP fooling.
  • Figure 4: Assessment Pipelines for Open-source AI Model Quality Attributes.
  • Figure 5: Cloud-based XAI Service Architecture.
  • ...and 6 more figures