Securing customer data

Customer data needs protection both at rest—while stored—and in transit—while moving between systems. This episode describes Google's 2020 approach: separate customers' data, encrypt it at multiple layers, and restrict communication between services.

Key Points Covered

• Shared hardware does not mean shared data access: The episode says Google logically separates customers' data even on the same physical server and restricts employee access to a small group. It also states provider policies against scanning customer data for advertising or selling it; these are the video's 2020 assurances 01:03–02:08(opens in a new tab).
• Encrypt before writing to physical storage: The episode describes storage services using centrally managed keys for application-layer encryption. That layer protects against threats below it, such as malicious disk firmware, while hardware encryption adds another layer 02:08–03:11(opens in a new tab).
• Network design can reduce exposure: Google's network and front-end servers are presented as limiting public-internet hops and filtering malicious traffic 02:08–03:11(opens in a new tab).
• Protect service calls even if the network is compromised: Google's remote procedure call (RPC) mechanisms use cryptography to keep data private, detect changes, and verify the communicating parties. The episode explains that encrypted inter-service traffic need not rely on the network path alone remaining secure 03:11–04:17(opens in a new tab).
• Restrict which servers can communicate: Google Front Ends can contact only an allowed set of internal servers. Firewalls and access-control lists add network separation around sensitive systems 03:11–04:17(opens in a new tab).

Full video: https://www.youtube.com/watch?v=qR-HMOcJC8c(opens in a new tab)