• Home
  • Resources
    • Find Resources by Topic Tags
    • Cybersecurity Policy Chart
    • CSIAC Reports
    • Webinars
    • Podcasts
    • Cybersecurity Digest
    • Standards & Reference Docs
    • Journals
    • Certifications
    • Acronym DB
    • Cybersecurity Related Websites
  • Services
    • Free Technical Inquiry
    • Core Analysis Task (CAT) Program
    • Subject Matter Expert (SME) Network
    • Training
    • Contact Us
  • Community
    • Upcoming Events
    • Cybersecurity
    • Modeling & Simulation
    • Knowledge Management
    • Software Engineering
  • About
    • About the CSIAC
    • The CSIAC Team
    • Subject Matter Expert (SME) Support
    • DTIC’s IAC Program
    • DTIC’s R&E Gateway
    • DTIC STI Program
    • FAQs
  • Skip to primary navigation
  • Skip to main content
  • Skip to primary sidebar
  • Skip to footer
Login / Register

CSIAC

Cyber Security and Information Systems Information Analysis Center

  • Resources
    • Find Resources by Topic Tags
    • Cybersecurity Policy Chart
    • CSIAC Reports
    • Webinars
    • Podcasts
    • Cybersecurity Digest
    • Standards & Reference Docs
    • Journals
    • Certifications
    • Acronym DB
    • Cybersecurity Websites
  • Services
    • Free Technical Inquiry
    • Core Analysis Task (CAT) Program
    • Subject Matter Expert (SME) Network
    • Training
    • Contact
  • Community
    • Upcoming Events
    • Cybersecurity
    • Modeling & Simulation
    • Knowledge Management
    • Software Engineering
  • About
    • About the CSIAC
    • The CSIAC Team
    • Subject Matter Expert (SME) Support
    • DTIC’s IAC Program
    • DTIC’s R&E Gateway
    • DTIC STI Program
    • FAQs
  • Cybersecurity
  • Modeling & Simulation
  • Knowledge Management
  • Software Engineering
/ All Podcast Series / CSIAC Webinars / A Fistful of Data, or the Good, Bad and Ugly of Adversarial Machine Learning

CSIAC Webinars - A Fistful of Data, or the Good, Bad and Ugly of Adversarial Machine Learning

Posted: 07/29/2020 | Presenter: Michael Weir | Leave a Comment

Notice: This podcast video may contain personal or third-party views and opinions not associated with the government.
Please see our terms of use located here: https://www.csiac.org/csiac-terms-of-use/
Recorded: 2020-08-12 | Series: CSIAC Webinars

This webinar provides an overview of Adversarial Machine Learning (AML), its relationship to Generative (Deep) Learning, and ways to view AML as a potential enabler for deploying more comprehensive system-level Machine Learning capabilities. The basic ideas driving AML and the system-level architecture needs of an effective integrated ML capability are compared to find areas of commonality and future utility beyond single-shot, algorithm-by-algorithm approaches to AML and remediation techniques.

The presenter of this webinar, Michael Weir, provided these notes, below, to go along with this webinar as well as reference materials:

Why do we “do” machine learning (ML) anyway? Because we can’t find another, better way to solve a problem. At the outset, we accept that we will get a probabilistic answer, and that’s good enough (or we find out how close to “good enough” we get, and use it or try again).

Our choice of “good enough” is usually associated with some performance parameter that is postulated, and then implemented in a mathematical process that calculates a loss function and an optimization function. The performance is then “proven” post-experiment as we check the results after tuning and multiple runs. That is, we are not directly measuring a given performance parameter, but measuring a mathematically constructed procedure that tells us whether or not the loss/optimization we selected (which resulted in an acceptable error rate in training) works in the real world – it “generalizes” to unseen examples.

The idea of “optimizing” the performance of an ML algorithm is not really tuning the performance parameter(s) that will solve your problem, it is tuning the optimization algorithm you built (or inferred by some mathematical construct) into the ML model. An example is the typical gradient descent 3D picture of “sliding down the slope” of a convex function. When you get to the bottom, the slope is zero and you’ve found the minimum of the two parameters you selected to represent your problem. To be clear, you are not measuring something like the specific spring characteristic in a weight/spring physics problem, but a mathematical representation that reflects some characteristic of a set of samples that might be close to a function that would operate similar to your real-world problem. You are trying to discover a function H(x) that operates well enough to approximate G(x), the problem space in the real world.

The reason we are talking about optimization is that it is the power and weakness of ML. Because we can’t know the G(x) [If we did, we wouldn’t need ML], we have to use what we know [the past examples we know] to build a “predictor” [what will happen with future unseen samples].

Reference materials associated with ideas in the Webinar:

1. References from the Webinar slides:
Leslie Valiant, A Theory of the Learnable, here: http://web.mit.edu/6.435/www/Valiant84.pdf
GPT-3 Model sizes: https://lambdalabs.com/blog/demystifying-gpt-3/
UncannyValley https://web.ics.purdue.edu/~drkelly/MoriTheUncannyValley1970.pdf
Generative Deep Learning: https://developers.google.com/machine-learning/gan/gan_structure
Stop Sign reference: https://arxiv.org/abs/1412.6572.pdf
Panda reference: https://arxiv.org/abs/1707.08945.pdf
Facebook reference: https://ai.facebook.com/blog/deepfake-detection-challenge
Microsoft reference: https://docs.microsoft.com/en-us/security/engineering/failure-modes-in-machine-learning

2. Other references that might be helpful in thinking about the applications of AML in new ways:
McAfee Face Fooling experiment: https://factschronicle.com/the-hack-that-can-fool-facial-recognition-algorithms-at-security-checkpoints-23740.html
Adversarial Robustness as a Prior: https://arxiv.org/abs/1906.00945
From Imagenet to Image Classification: https://arxiv.org/abs/2005.11295

Download Associated Files:
You must be logged in to download files associated with this video podcast. Click here to login.

Presenter

Michael Weir
Michael Weir
Michael Weir is currently working with Quanterion Solutions, Inc as a Senior Technical Advisor/Subject Matter Expert for the Cybersecurity and Information Systems Information Analysis Center (CSIAC), and with the Griffiss Institute as the developer/facilitator for the AFRL-sponsored Machine Learning Bootcamp, a multi-month immersion program for AFRL engineers. Mr. Weir was previously the Director of the CSIAC, and before that the Chief of Communications and Information Systems at the Eastern Air Defense Sector (EADS), Rome, New York. EADS is one of two NORAD/NORTHCOM air defense sectors in the Continental United States. He was responsible for setting up and maintaining the Sector's cyber posture during and after 9/11 and evolving the data/communication/sensor integration through the following decade. Mr. Weir has Bachelor's degrees in Music Performance and in Electrical and Computer Engineering, and a Master's Degree in Information Systems, along with certifications in the cybersecurity domain.

Tags: Adversarial Machine Learning (AML), Artificial Intelligence (AI), Machine Learning (ML)

Previous in this Series:
« Cyberphysical Infrastructure Testbeds
Next in this Series:
Big Data and Big Implications for Bio-cybersecurity »

Reader Interactions

Leave a Comment Cancel

You must be logged in to post a comment.

sidebar

Blog Sidebar

Featured Content

Data Privacy Day - Jan 28

Data Privacy Day is January 28th

You can help create a global community that respects privacy, safeguards data, and enables trust. You can help teach others about privacy at home, at work, and in your community.

Learn How

Featured Subject Matter Expert (SME): Daksha Bhasker

A dynamic CSIAC SME, Senior Principal Cybersecurity Architect, Daksha Bhasker has 20 years of experience in the telecommunications services provider industry. She has worked in systems security design and architecture in production environments of carriers, often leading multidisciplinary teams for cybersecurity integration, from conception to delivery of complex technical solutions. As a CSIAC SME, Daksha's contributions include several published CSIAC Journal articles and a webinar presentation on the sophiscated architectures that phone carriers use to stop robocalls.

View SME's Contributed Content

The DoD Cybersecurity Policy Chart

The DoD Cybersecurity Policy Chart

This chart captures the tremendous breadth of applicable policies, some of which many cybersecurity professionals may not even be aware, in a helpful organizational scheme.

View the Policy Chart

CSIAC Report - Smart Cities, Smart Bases and Secure Cloud Architecture for Resiliency by Design

Integration of Smart City Technologies to create Smart Bases for DoD will require due diligence with respect to the security of the data produced by Internet of Things (IOT) and Industrial Internet of Things (IIOT). This will increase more so with the rollout of 5G and increased automation "at the edge". Commercially, data will be moving to the cloud first, and then stored for process improvement analysis by end-users. As such, implementation of Secure Cloud Architectures is a must. This report provides some use cases and a description of a risk based approach to cloud data security. Clear understanding, adaptation, and implementation of a secure cloud framework will provide the military the means to make progress in becoming a smart military.

Read the Report

CSIAC Journal - Data-Centric Environment: Rise of Internet-Based Modern Warfare “iWar”

CSIAC Journal Cover Volume 7 Number 4

This journal addresses a collection of modern security concerns that range from social media attacks and internet-connected devices to a hypothetical defense strategy for private sector entities.

Read the Journal

CSIAC Journal M&S Special Edition - M&S Applied Across Broad Spectrum Defense and Federal Endeavors

CSIAC Journal Cover Volume 7 Number 3

This Special Edition of the CSIAC Journal highlights a broad array of modeling and simulation contributions – whether in training, testing, experimentation, research, engineering, or other endeavors.

Read the Journal

CSIAC Journal - Resilient Industrial Control Systems (ICS) & Cyber Physical Systems (CPS)

CSIAC Journal Cover Volume 7 Number 2

This edition of the CSIAC Journal focuses on the topic of cybersecurity of Cyber-Physical Systems (CPS), particularly those that make up Critical Infrastructure (CI).

Read the Journal

Recent Video Podcasts

  • Privacy Impact Assessment: The Foundation for Managing Privacy Risk Series: The CSIAC Podcast
  • Agile Condor: Supercomputing at the Edge for Intelligent Analytics Series: CSIAC Webinars
  • Securing the Supply Chain: A Hybrid Approach to Effective SCRM Policies and Procedures Series: The CSIAC Podcast
  • DoD Vulnerability Disclosure Program (VDP) Series: CSIAC Webinars
  • 5 Best Practices for a Secure Infrastructure Series: The CSIAC Podcast
View all Podcasts

Upcoming Events

Thu 21

SANS Cyber Security Central: Jan 2021

January 18 - January 23
Organizer: SANS Institute
Thu 21

AI Champions, Online – Supply Chain

January 19 @ 14:00 - January 21 @ 15:30 EST
Thu 21

SANS Cyber Threat Intelligence Summit 2021

January 21 - January 22
Organizer: SANS Institute
Fri 22

SANS Cyber Threat Intelligence Solutions Track 2021

January 22 @ 09:00 - 17:00 EST
Organizer: SANS Institute
Wed 27

Enterprise Data Governance Online 2021

January 27 @ 08:00 - 13:30 EST
Organizer: DATAVERSITY
View all Events

Footer

CSIAC Products & Services

  • Free Technical Inquiry
  • Core Analysis Tasks (CATs)
  • Resources
  • Events Calendar
  • Frequently Asked Questions
  • Product Feedback Form

About CSIAC

The CSIAC is a DoD-sponsored Center of Excellence in the fields of Cybersecurity, Software Engineering, Modeling & Simulation, and Knowledge Management & Information Sharing.Learn More

Contact Us

Phone:800-214-7921
Email:info@csiac.org
Address:   266 Genesee St.
Utica, NY 13502
Send us a Message
US Department of Defense Logo USD(R&E) Logo DTIC Logo DoD IACs Logo

Copyright 2012-2021, Quanterion Solutions Incorporated

Sitemap | Privacy Policy | Terms of Use | Accessibility Information
Accessibility / Section 508 | FOIA | Link Disclaimer | No Fear Act | Policy Memoranda | Privacy, Security & Copyright | Recovery Act | USA.Gov

This website uses cookies to provide our services and to improve your experience. By using this site, you consent to the use of our cookies. To read more about the use of our site, please click "Read More". Otherwise, click "Dismiss" to hide this notice. Dismiss Read More
Privacy & Cookies Policy

Privacy Overview

This website uses cookies to improve your experience while you navigate through the website. Out of these, the cookies that are categorized as necessary are stored on your browser as they are essential for the working of basic functionalities of the website. We also use third-party cookies that help us analyze and understand how you use this website. These cookies will be stored in your browser only with your consent. You also have the option to opt-out of these cookies. But opting out of some of these cookies may affect your browsing experience.
Necessary
Always Enabled

Necessary cookies are absolutely essential for the website to function properly. This category only includes cookies that ensures basic functionalities and security features of the website. These cookies do not store any personal information.

Non-necessary

Any cookies that may not be particularly necessary for the website to function and is used specifically to collect user personal data via analytics, ads, other embedded contents are termed as non-necessary cookies. It is mandatory to procure user consent prior to running these cookies on your website.