Scholarlink

This chapter presents adversarial defense strategies and privacy-preserving technologies for securing AI systems. It begins with discussing various defense methods against adversarial attacks and privacy-preserving techniques, including adversarial training, certified robustness, gradient masking, adversarial detection, data anonymization, homomorphic encryption, federated learning, and secure multi-party computations. It also explores the importance of model robustness, privacy preservation, explainability, and safe deployment. Additionally, this chapter examines explainable AI (XAI) methods, including SHAP, LIME, and feature attribution, to improve model interpretability. Finally, it discusses the challenges of applying these defense methods in real-world AI deployments for secure and transparent AI systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Defense Methods for Adversarial Attacks and Privacy Issues in Secure AI

  • 章节

  • 发表日期：

    2025年

查看全文
  • Dilli Prasad Sharma,
  • Arash Habibi Lashkari,
  • Mahdi Daghmehchi Firoozjaei,
  • Samaneh Mahdavifar,
  • Pulei Xiong

摘要

This chapter presents adversarial defense strategies and privacy-preserving technologies for securing AI systems. It begins with discussing various defense methods against adversarial attacks and privacy-preserving techniques, including adversarial training, certified robustness, gradient masking, adversarial detection, data anonymization, homomorphic encryption, federated learning, and secure multi-party computations. It also explores the importance of model robustness, privacy preservation, explainability, and safe deployment. Additionally, this chapter examines explainable AI (XAI) methods, including SHAP, LIME, and feature attribution, to improve model interpretability. Finally, it discusses the challenges of applying these defense methods in real-world AI deployments for secure and transparent AI systems.