Keeping Your AI/ML Data Secure
summary
Claim: 80% of security breaches involve privileged credentials and roughly half of databases lack encryption; about 25% of organisations see cryptojacking. Concrete tools and failures: Azure Form Recognizer uploads files and JSON outputs to an Azure Storage account that teams often forget, leading to incidents like the public S3 backup leak. Recommended fixes actually used: Azure Key Vault + Azure RBAC, short‑lived SAS tokens (aim for 30–45 minute lifetimes), App Configuration, secret scanning (Spectral Ops for large apps, all secrets for quick MVPs), Defender for Cloud and Advisor.
AI-written summary · about
chapters (14)
- 0:00AI ML data security intro
- 1:15Why securing AI ML data matters
- 5:27Security breaches and risks
- 10:20Azure Form Recognizer pitfalls
- 13:10Cloud security shared responsibility
- 15:15AI service status and risks
- 18:10Managing secrets securely
- 24:32RBAC for AI systems
- 29:53Azure Key Vault integration
- 32:20Secure storage with SAS tokens
- 36:00Securing Azure SQL databases
- 40:11API security and defense
- 41:58Key takeaways
- 43:21Q&A
related talks
39mTop 13 best security practices for Azure
44mDeveloper Productivity Shouldn’t Breach Security
49mTop 5 ways to secure Azure resources
52mUncovering Azure's Silent Threats: A Journey Into Cloud Vulnerabilities
57mCloud’s Dirty Little Secret: It Was Misconfigs All Along
19mEnhancing Security in Azure Cosmos DB: Best Practices
46mSecrets Management for the Modern Tech. Pro
55mAzure Attacks Unpacked: Real-World Threats & Lessons