What is adversarial machine learning and why does it matter for security professionals?

Adversarial machine learning refers to techniques that manipulate, deceive, or exploit AI systems — for example, feeding corrupted training data to degrade model performance, or crafting inputs that cause a model to misclassify a threat. As AI is embedded deeper into security operations, understanding these attack techniques is essential for anyone building, deploying, or auditing AI-powered security tools.

Does this course cover AI regulation and compliance?

<text variant="body1">Yes. The final module covers responsible AI principles, global governance frameworks, and regulatory compliance requirements — including considerations around fairness, bias, transparency, and data ethics — ensuring you can assess AI deployments against both technical and legal standards.

When will I have access to the lectures and assignments?

To access the course materials, assignments and to earn a Certificate, you will need to purchase the Certificate experience when you enroll in a course. You can try a Free Trial instead, or apply for Financial Aid. The course may offer 'Full Course, No Certificate' instead. This option lets you see all course materials, submit required assessments, and get a final grade. This also means that you will not be able to purchase a Certificate experience.

What will I get if I subscribe to this Specialization?

When you enroll in the course, you get access to all of the courses in the Specialization, and you earn a certificate when you complete the work. Your electronic Certificate will be added to your Accomplishments page - from there, you can print your Certificate or add it to your LinkedIn profile.

Is financial aid available?

Yes. In select learning programs, you can apply for financial aid or a scholarship if you can’t afford the enrollment fee. If fin aid or scholarship is available for your learning program selection, you’ll find a link to apply on the description page.

Adversarial AI: Attacking, Defending & Governing ML Systems

Adversarial AI: Attacking, Defending & Governing ML Systems

本课程是 AI-Powered Cybersecurity 专项课程的一部分

位教师：Matt Bushby

包含在中

了解更多

5个模块

深入了解一个主题并学习基础知识。

中级等级

推荐体验

1 周完成

在 10 小时一周

灵活的计划

自行安排学习进度

5个模块

深入了解一个主题并学习基础知识。

中级等级

推荐体验

1 周完成

在 10 小时一周

灵活的计划

自行安排学习进度

您将学到什么

Analyse adversarial attack vectors targeting ML systems including poisoning, model stealing, & backdoor exploits, and assess their operational impact
Design & implement layered technical defences using differential privacy, guardrail protection, & secure algorithm design to maintain model integrity
Plan and conduct AI security testing using red, purple, and blue teaming approaches to validate ML model robustness under adversarial conditions
Evaluate responsible AI governance frameworks and regulatory requirements to ensure AI systems are ethical, fair, and compliant

要了解的详细信息

可分享的证书

添加到您的领英档案

了解顶级公司的员工如何掌握热门技能

了解关于 Coursera for Business 的更多信息

Petrobras, TATA, Danone, Capgemini, P&G 和 L'Oreal 的徽标

积累特定领域的专业知识

本课程是 AI-Powered Cybersecurity 专项课程专项课程的一部分

在注册此课程时，您还会同时注册此专项课程。

向行业专家学习新概念
获得对主题或工具的基础理解
通过实践项目培养工作相关技能
获得可共享的职业证书

该课程共有5个模块

As AI becomes central to cybersecurity defence, attackers are increasingly targeting the AI systems themselves. Model poisoning, adversarial inputs, backdoor exploits, and model stealing are active threats — and most security teams are unprepared to detect or defend against them. This course gives you the knowledge and practical strategies to secure ML systems from the inside out.

You'll examine how ML systems are manipulated through adversarial inputs, poisoning attacks, and threat models across real-world use cases including malware detection and fraud analytics. You'll then explore advanced attack vectors: model poisoning, information leakage, model stealing, and backdoor exploits, and assess their impact on data privacy, intellectual property, and user safety. From attack to defence, you'll learn to apply secure algorithm design, differential privacy, and guardrail protection — and conduct AI security testing using red, purple, and blue teaming approaches. The course closes with AI governance: responsible AI principles, bias mitigation, transparency, data ethics, and the global regulatory frameworks governing AI in cybersecurity. Designed for security analysts, ML engineers, security architects, and risk and compliance professionals working with AI-powered security systems. Job skills taught: Adversarial AI Defence · AI Security Testing · ML Threat Modelling · Model Robustness · Differential Privacy · Red/Blue/Purple Teaming · AI Governance · Responsible AI · Regulatory Compliance for AI Features Coursera Coach, Dialogues and Role Plays - a smarter way to learn with interactive, real-time conversations that help you test your knowledge, challenge assumptions, and deepen your understanding as you progress through the course.

单元详情

As machine learning integrates into cyber defences, so do methods for breaking it. This module helps you understand how machine learning systems are manipulated and how to defend against it. You will examine adversarial machine learning through examples of threat models, adversarial inputs, and poisoning attacks. Learn how data can compromise models and how attackers exploit vulnerabilities. This module also covers defensive techniques to build resilient models and implement countermeasures. Safeguard your models in malware detection, intrusion systems, or fraud analytics against sophisticated attacks.

涵盖的内容

1个作业5个插件

As AI systems deploy, exposure to adversarial threats and misuse increases. This module explores how AI is attacked and exploited, a critical focus for cyber professionals. You will dive into AI-specific attack vectors: model poisoning, information leakage, model stealing, and backdoor exploits. These threats compromise AI performance and pose risks to data privacy, intellectual property, and user safety. Examine harmful AI outputs from biased data or manipulation. Learn how output alignment, ethical censorship, and AI-powered surveillance affect public trust and legal compliance. Analyze case studies to identify AI vulnerabilities and understand societal consequences of insecure deployments. Ensure AI shapes the world securely and responsibly.

涵盖的内容

2个作业6个插件

Defending AI systems against emerging threats is critical. This module explores technical controls and testing strategies to secure AI models. You will learn to apply AI-specific defences, from secure algorithm design to privacy-preserving techniques like differential privacy. Examine how to test and validate AI model robustness using red, purple, and blue teaming approaches. Focus on balancing security, utility, and performance to make informed trade-offs. Gain practical skills to implement trusted controls and rigorously test for resilience against real-world threats, whether building or auditing AI systems.

涵盖的内容

2个作业8个插件

2个作业总计30分钟

End of module practice quiz15分钟
End of module quiz15分钟

8个插件总计112分钟

Overview2分钟
Introduction10分钟
Defence techniques and strategies30分钟
Defences for attacks on GenAI models20分钟
Selecting appropriate controls15分钟
Guardrail protection versus guardrail failure15分钟
AI security testing and benchmarking15分钟
Summary5分钟

As AI systems grow, responsible design, deployment, and governance are imperative. This module introduces Responsible AI principles: fairness, bias mitigation, transparency, and ethical accountability. You will explore how AI decisions impact individuals and communities, navigating trade-offs between user privacy, model performance, and transparency. Unpack challenges like data sourcing, labelling, and ethical implications of large-scale models. Learn practical strategies for enhancing trust in AI systems. Dive into global frameworks, policies, and governance models supporting secure, ethical AI adoption. Ensure AI systems are functional, fair, transparent, and aligned with regulatory expectations.

涵盖的内容

2个作业6个插件

In this module, you will analyse a simulated adversarial attack on a deployed ML model, identify the attack type, and recommend a defence strategy. The project allows you to build a comprehensive portfolio artefacts demonstrating your end-to-end capabilities.