4 ms·
> Yes, it was stupid for them to have everybody's PII attached to that one webserver. A single point of failure should never result in compromising the whole sy
by mmagin 9y ago
> Yes, it was stupid for them to have everybody's PII attached to that one webserver. A single point of failure should never result in compromising the whole system. But think about how that architecture would work in practice. A customer service rep still needs to get at most of your data. It's a credit bureau. Where would the data be stored in a way that a remote code exec wouldn't be able to snag it?
A particular customer service rep needs to be able to get at most of one person's data at a given point in time. In fact, they should probably be somewhat rate-limited as to how many people's data they can look at in a given period of time. The front-end server that runs the app that serves up this data probably should not even be allowed access to the full set of data at any given point in time. Also, presumably this has to happen from a desktop computer at a call center that's at a company that contracts with Equifax, so it would be reasonable to assume those computers are VPNing to some Equifax network, not that access to all customer data from any internet address is a necessity.
- viraptor 9y agoThis is a the big issue. Too many people complain about that one patching issue. Yes, we know everything is broken in some way, but that doesn't stop anyone from constructing layers of defence where even with a full compromise of one system you get very limited access internally. That should've been a huge point of the published document, not just "we didn't patch, so it's game-over".