# Supporting AI/ML Security Workers through an Adversarial Techniques, Tools, and Common Knowledge (AI/ML ATT&CK) Framework

**Authors:** Mohamad Fazelnia, Ahmet Okutan, Mehdi Mirakhorli  
**arXiv:** [2211.05075](https://arxiv.org/abs/2211.05075)  
**Date:** 2022-11-09  
**Summary page:** https://www.summarizepaper.com/en/paper/2211.05075v1/

## Key points

- The paper aims to support professionals involved in the development and deployment of secure AI-enabled software systems.
- It focuses on AI/ML Security Workers and presents the AI/ML Adversarial Techniques, Tools, and Common Knowledge (AI/ML ATT&CK) framework.
- The framework enables workers to explore offensive and defensive tactics intuitively.
- The authors recognize that as AI technology becomes more prevalent in various industries, so do the risks associated with it.
- They provide a comprehensive guide for security workers to identify potential vulnerabilities in their systems through adversarial techniques such as data poisoning or model inversion attacks.
- They offer tools for detecting and mitigating these attacks while also providing common knowledge about the latest threats and trends in AI security.
- Overall, this paper serves as a valuable resource for those seeking to enhance their understanding of AI security and develop effective strategies for protecting their systems against malicious actors.

## Summary

This paper, authored by Mohamad Fazelnia, Ahmet Okutan, and Mehdi Mirakhorli, aims to support professionals involved in the development and deployment of secure AI-enabled software systems. Specifically, it focuses on AI/ML Security Workers and presents the AI/ML Adversarial Techniques, Tools, and Common Knowledge (AI/ML ATT&CK) framework. This framework enables these workers to explore offensive and defensive tactics intuitively. The authors recognize that as AI technology becomes more prevalent in various industries, so do the risks associated with it. Therefore, they provide a comprehensive guide for security workers to identify potential vulnerabilities in their systems through adversarial techniques such as data poisoning or model inversion attacks. Additionally, they offer tools for detecting and mitigating these attacks while also providing common knowledge about the latest threats and trends in AI security. Overall, this paper serves as a valuable resource for those seeking to enhance their understanding of AI security and develop effective strategies for protecting their systems against malicious actors.