Lecture

Machine Learning Security

Graph Chat

Chat with Graph Search

Ask any question about EPFL courses, lectures, exercises, research, news, etc. or try the example questions below.

DISCLAIMER: The Graph Chatbot is not programmed to provide explicit or categorical answers to your questions. Rather, it transforms your questions into API requests that are distributed across the various IT services officially administered by EPFL. Its purpose is solely to collect and recommend relevant references to content that you can explore to help you answer your questions.

Description

This lecture covers the basics of machine learning security, including stealing models, altering model outputs, issues with federated learning, and pitfalls in machine learning taxonomy. It also delves into adversarial conditions, model stealing, machine learning as a service, and privacy challenges in machine learning. The instructor discusses the implications of attacks against neural networks, defenses against model stealing, and the challenges of preventing adversarial examples. The lecture concludes with insights on biases, fallacies, and the importance of addressing biases in machine learning models.

Official source

Related concepts (32)

Privacy

Privacy (UK, US) is the ability of an individual or group to seclude themselves or information about themselves, and thereby express themselves selectively. The domain of privacy partially overlaps with security, which can include the concepts of appropriate use and protection of information. Privacy may also take the form of bodily integrity. There have been many different conceptions of privacy throughout history. Most cultures recognize the right of an individual to withhold aspects of their personal lives from public record.

Machine learning

Machine learning (ML) is an umbrella term for solving problems for which development of algorithms by human programmers would be cost-prohibitive, and instead the problems are solved by helping machines 'discover' their 'own' algorithms, without needing to be explicitly told what to do by any human-developed algorithms. Recently, generative artificial neural networks have been able to surpass results of many previous approaches.

Privacy law

Privacy law is the body of law that deals with the regulating, storing, and using of personally identifiable information, personal healthcare information, and financial information of individuals, which can be collected by governments, public or private organisations, or other individuals. It also applies in the commercial sector to things like trade secrets and the liability that directors, officers, and employees have when handing sensitive information.

Right to privacy

The right to privacy is an element of various legal traditions that intends to restrain governmental and private actions that threaten the privacy of individuals. Over 150 national constitutions mention the right to privacy. On 10 December 1948, the United Nations General Assembly adopted the Universal Declaration of Human Rights (UDHR), originally written to guarantee individual rights of everyone everywhere; while right to privacy does not appear in the document, many interpret this through Article 12, which states: "No one shall be subjected to arbitrary interference with his privacy, family, home or correspondence, nor to attacks upon his honour and reputation.

Privacy by design

Privacy by design is an approach to systems engineering initially developed by Ann Cavoukian and formalized in a joint report on privacy-enhancing technologies by a joint team of the Information and Privacy Commissioner of Ontario (Canada), the Dutch Data Protection Authority, and the Netherlands Organisation for Applied Scientific Research in 1995. The privacy by design framework was published in 2009 and adopted by the International Assembly of Privacy Commissioners and Data Protection Authorities in 2010.

Related lectures (60)

Privacy Technologies and Data Protection

Explores privacy technologies, data protection, surveillance risks, and Privacy Enhancing Technologies for social and institutional privacy.

Machine Learning and Privacy

Explores the intersection of machine learning and privacy, discussing confidentiality, attacks, differential privacy, and trade-offs in federated learning.

Désanonymisation: Risks and Techniques

Explores risks and techniques in de-anonymizing data, including flaws in methods and real-world examples of failed attempts.

Edward Snowden: Whistleblower and Data Privacy Advocate

Explores Edward Snowden's whistleblowing on NSA's data collection and the challenges of data security.

Privacy technologies and data protection