From the course: AI Security Tools and Automation

Unlock this course with a free trial

Join today to access over 25,200 courses taught by industry experts.

Designing with security-by-design

Designing with security-by-design

Security can't be bolted on after the fact. If you build an AI tool first and then try to secure it, you're going to miss things. Security by design means thinking about threats from day one. What data are we handling? Who can access it? What happens if the API key leaks? And what if someone tries prompt injection? I'll show you how the security principles we built into our Privacy Gap Analysis tool work. First up is least privilege. Let's give users and systems only the access they need and nothing more. Then we've got defense in depth. This is multiple layers of security, not just one. We want to fail securely. When something goes wrong, we've got to fail in a secure state. We've got separation of duties, and this means don't let one component do everything. With secure defaults, the default configuration should be secure and not tamperable. We need to audit everything, and that means logging all security relevant events and anomalies. We've got trust, but verify. And what does that…

Contents