|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/threat-landscape-of-ai-systems/
课程评论：没有评论
课程名称：人工智能系统的威胁环境 课程概述：随着人工智能（AI）系统在医疗、金融等关键行业的不断应用，它们面临着来自对手攻击和系统漏洞的日益增长的安全挑战。《人工智能系统的威胁环境》课程深入探讨了现代AI系统所面临的安全威胁，包括各种攻击类型，如规避攻击、中毒攻击、模型反演等。该课程系列旨在为学习者提供理解和防御AI系统对抗各种对手攻击的知识和工具。 参与者将深入研究： - **规避攻击**：如何通过微妙的输入操控误导AI系统并导致错误分类。 - **中毒攻击**：攻击者如何破坏训练数据以操控模型行为并降低模型准确性。 - **模型反演攻击**：如何从模型的输出重建敏感输入数据，从而导致隐私泄露。 - **其他攻击向量**：包括数据提取、成员推断和后门攻击等。 此外，本课程还涵盖： - **对抗性攻击的影响**：这些威胁对人脸识别、自动驾驶汽车、金融模型和医疗AI等行业的影响。 - **缓解技术**：保护AI系统的策略，包括对抗训练、差分隐私、模型加密和访问控制等。 - **真实案例研究**：分析著名对抗攻击的实例及其缓解措施。 通过讲座、案例研究、实践练习和评估的结合，学生将获得关于AI系统当前和未来威胁环境的扎实理解，并学习如何应用先进的安全实践来保护AI模型免受攻击。
Artificial intelligence (AI) systems are increasingly integrated into critical industries, from healthcare to finance, yet they face growing security challenges from adversarial attacks and vulnerabilities. Threat Landscape of AI Systems is an in-depth exploration of the security threats that modern AI systems face, including various types of attacks, such as evasion, poisoning, model inversion, and more. This course series provides learners with the knowledge and tools to understand and defend AI systems against a broad range of adversarial exploits.Participants will delve into:Evasion Attacks: How subtle input manipulations deceive AI systems and cause misclassifications.Poisoning Attacks: How attackers corrupt training data to manipulate model behavior and reduce accuracy.Model Inversion Attacks: How sensitive input data can be reconstructed from a model's output, leading to privacy breaches.Other Attack Vectors: Including data extraction, membership inference, and backdoor attacks.Additionally, this course covers:Impact of Adversarial Attacks: The effects of these threats on industries such as facial recognition, autonomous vehicles, financial models, and healthcare AI.Mitigation Techniques: Strategies for defending AI systems, including adversarial training, differential privacy, model encryption, and access controls.Real-World Case Studies: Analyzing prominent examples of adversarial attacks and how they were mitigated.Through a combination of lectures, case studies, practical exercises, and assessments, students will gain a solid understanding of the current and future threat landscape of AI systems. They will also learn how to apply cutting-edge security practices to safeguard AI models from attack.
