Threats for Machine Learning
The presentation illustrates where machine learning applications can be attacked, the means for carrying out the attack and some mitigations that can be employed. The elements in building and deploying a machine learning application are reviewed, considering both data and processes. The impact of attacks on each element is considered in turn. Special attention is given to transfer learning, a popular way to construct quickly a machine learning application. Mitigations to these attacks are discussed with the engineering tradeoffs between security and accuracy. Finally, the methods by which an attacker could get access to the machine learning system are reviewed.
What attendees will learn:
- What are the new attack surfaces exposed by machine learning application
- What is the tradeoff between security and accuracy in a machine learning application
- How might machine learning applications be attacked
Who should attend?
- Cyber security analyst
- Machine learning application developer
- Manager of data science or machine learning team
Speaker and Presenter Information
Dr. Mark Sherman is the Technical Director of the Cybersecurity Foundations group at CERT within CMU’s Software Engineering Institute. His team focuses on foundational research on the life cycle for building secure software and on data analytics in cyber security. Before coming to CERT, Dr. Sherman was at IBM and various startups, working on mobile systems, integrated hardware-software appliances, transaction processing, languages and compilers, virtualization, network protocols and databases. Dr. Sherman received his undergraduate degrees from MIT and his PhD in Computer Science from CMU.
Relevant Government Agencies
Air Force, Army, Navy & Marine Corps, Intelligence Agencies, DOD & Military, Dept of Education, Dept of Labor, Dept of State, Dept of Treasury, Dept of Veterans Affairs, GSA, NASA, Judicial Branch Agencies, County Government, FEMA, Coast Guard, National Guard Association, Federal Government, State & Local Government