Data Security for Data Engineers
summary
Data engineers must treat security as shared responsibility and adopt concrete controls: integrate GitHub secret scanner into CI/CD to catch leaked credentials, use a secret store for API keys, and run a safety dependency checker against PyPI packages after noting “150+ vulnerabilities in ~40% of packages.” Threat models focus on real exploits — stolen creds, dependency typosquats, and malicious ML pickles that spawn reverse shells (example: reduce -> shell to port 444) — so avoid pickle models or validate and sandbox them.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:32Agenda
- 0:50Defining Data Security
- 2:13My Journey
- 5:06Vulnerabilities
- 5:45Cyber Security Stats
- 6:58Targeted Opportunistic Attacks
- 10:15Threat Modeling
- 10:51Modeling for Data Engineers
- 12:48Frameworks
- 13:46Scenarios
- 14:09Risk Matrix
- 14:54What could go wrong
- 16:16What to do
- 17:16Dependencies
- 18:38Machine Learning
- 19:39Wrap Up
related talks
22mSecurity for Data Scientists
27mIncluding security in your engineering culture
37mSecurity Doesn’t Have To Be a Nightmare
28mFlawed ML Security: Mitigating Security Vulnerabilities in Data & Machine...
21mSecurity That Enables: Breaking Down Security Silos in the DevOps Ecosystem
33mUnderstanding the Threats and Attacks on Data Science Applications and Models
45mKeeping Your AI/ML Data Secure
24mBeing secure by design: Engineer-led security