Plain-language glossary

• Aegis: A platform that lets teams train ML models privately across multiple locations.
• Privacy budget (epsilon): A number that controls privacy strength; lower = stronger privacy.
• Differential Privacy (DP): A technique that hides the impact of any single person’s data.
• Federated Learning (FL): Training across sites without centralizing raw data.
• Aggregator: How updates from participants are combined (e.g., Krum, Trimmed Mean).
• RBAC: Role-based access control (Admin, Operator, Viewer) for safe operations.
• mTLS: Mutual TLS—both client and server prove who they are before talking.
• Audit log: A list of who did what and when, in a tamper‑evident way.