# A Survey of Game Theoretic Approaches for Adversarial Machine Learning in Cybersecurity Tasks

**Authors:** Prithviraj Dasgupta, Joseph B. Collins  
**arXiv:** [1912.02258](https://arxiv.org/abs/1912.02258)  
**Date:** 2019-12-04  
**Summary page:** https://www.summarizepaper.com/en/paper/1912.02258v1/

## Key points

- Machine learning techniques are widely used in cybersecurity
- Supervised learning algorithms can be vulnerable to adversarial attacks
- Adversaries can alter training data to misguide the learning algorithm and cause errors
- Game theoretic approaches are being explored to make machine learning algorithms more robust against attacks
- Game theory allows for strategic decision-making in situations with conflicting interests
- The paper titled "A Survey of Game Theoretic Approaches for Adversarial Machine Learning in Cybersecurity Tasks" provides a detailed survey of techniques to enhance resilience against attacks using game theory
- These approaches can help mitigate the impact of adversarial attacks and improve reliability of deep machine learning-based systems for cybersecurity tasks
- Open problems, challenges, and potential directions for further research are highlighted in the paper
- Addressing these challenges can contribute to making machine learning algorithms more robust and reliable in cybersecurity.

## Summary

Machine learning techniques are widely used in cybersecurity to automate various tasks. However, these techniques often rely on supervised learning algorithms that can be vulnerable to adversarial attacks. Adversaries can deliberately alter the training data to misguide the learning algorithm and cause classification errors, leaving critical systems vulnerable to cybersecurity attacks. To address this vulnerability, researchers have been exploring game theoretic approaches to make machine learning algorithms more robust against adversarial attacks. Game theory provides a computational framework that allows for strategic decision-making in situations where multiple entities with conflicting interests are involved. In their paper titled "A Survey of Game Theoretic Approaches for Adversarial Machine Learning in Cybersecurity Tasks," authors Prithviraj Dasgupta and Joseph B. Collins provide a detailed survey of state-of-the-art techniques used to enhance the resilience of machine learning algorithms against adversarial attacks using game theory. They discuss how these approaches can help mitigate the impact of adversarial attacks and improve the reliability of deep machine learning-based systems for cybersecurity tasks. The authors also highlight open problems and challenges in this field, as well as potential directions for further research. By addressing these challenges, future studies can contribute to making machine learning algorithms even more robust and reliable in the context of cybersecurity. Overall, this paper sheds light on the importance of considering adversarial attacks in cybersecurity and provides valuable insights into how game theoretic approaches can be leveraged to enhance the security and effectiveness of machine learning-based systems.