Software Snack Bites

#17 - Ian Swanson (CEO of Protect AI) & Emilio Escobar (CISO of Datadog) - Deep Dive into ML & AI Security, Emerging Attack Surfaces, MLSecOps, and LLMs Use in Adversarial and Defensive Security

Aug 16, 2023

Ian Swanson, CEO of Protect AI, and Emilio Escobar, CISO of Datadog, discuss the importance of ML & AI security, emerging attack surfaces, MLSecOps, and LLMs use in adversarial and defensive security. They dive into the role of Protect AI in securing ML systems, the concept of MLSecOps, the difference between AI security and ML security, and the importance of data understanding and certification in ML and AI. They also explore enforcing policies, testing models for data security, and the use of AI tools in attack and defense strategies.

59:06

Creator website

Episode guests

Emilio Escobar

Ian Swanson

AI Summary

AI Chapters

Episode notes

Podcast summary created with Snipd AI

Quick takeaways

MLSecOps is a crucial practice in securing AI and ML systems by addressing vulnerabilities in the ML supply chain.

The adoption of ML and AI among enterprises is growing, but there is still a need for infrastructure to support ML operations.

Protecting the ML supply chain and implementing data security measures are essential to address threats like data poisoning and prompt injection attacks.

Deep dives

ML SecOps and Protect AI

Protect AI is a company that specializes in building security solutions for AI applications and ML systems. Their focus is on the emerging field of ML SecOps, which refers to the inclusion of security practices within the ML operations. They aim to address the vulnerabilities in the ML supply chain, including the use of open source libraries, data selection, and production deployments. Protect AI's mission is to provide security tools and solutions to ensure the safe and secure implementation of AI and ML systems.

Introduction

2min

ML Sec Ops and Protect AI

22min

ML Engineers as Shadow IT

7min

Exploring the Difference Between AI Security and ML Security with a Focus on LLMs

2min

Attestation, Certification, and the Importance of Data Understanding in ML and AI

14min

Enforcing Policies and Testing Models for Data Security

3min

AI Tools in Attack and Defense Strategies

9min

Ian Swanson is the Co-Founder & CEO of Protect AI which secures AI and ML systems. He was previously the Co-Founder & CEO of Datascience.com which sold to Oracle, where he was VP AI & ML and then was Worldwide Leader of AI & ML at AWS before starting Protect AI. I’m also joined by Emilio Escobar, CISO of Datadog as a special co-host! In this episode, we dive into what is ML & AI Security, the different types of attacks that hackers are using, and how teams can best secure their systems. We also talk about MLSecOps and why an ML BOM (Bill of Materials) is needed. Finally, we cover emerging attack surfaces and how LLMs will change both adversarial and defensive security products.

Where to Find Ian:

* Twitter: https://twitter.com/ianrswanson

* LinkedIn: https://www.linkedin.com/in/ianswanson/

* Company Website: https://protectai.com/

* MLSecOps Community & Podcast: https://mlsecops.com/

Where to Find Emilio:

* Twitter: https://twitter.com/eaescob

* LinkedIn: https://www.linkedin.com/in/emilioesc/

* Emilio’s Newsletter: https://emilioe.substack.com

Where to Find Shomik:

* Twitter: https://twitter.com/shomikghosh21

* LinkedIn: https://www.linkedin.com/in/shomik-ghosh-a5a71319/