Applied Security at Scale #
How scale has changed security:
- Types of scale: people -> multiple devices per person -> multiple (hundreds!) services connected per device
- Surface area: hardware -> software
- Attackers: individuals -> nation states
- Consequecnes: websites down/data stolen -> countries down
- Defense: identity as an afterthought -> identity as frontline
Establishing Security at Scale #
Guest lecture: Eric Grosse, former Google VP of Security and Privacy
- Originally: Bell Labs, where telco industry did not care much about privacy + was resistant to new security technologies e.g. transport encryption
- Joined Google in 2007 after top researchers left Bell Labs; at this point Google no longer was a startup, but not a large multilateral corporation
- However, security was still in the style of a startup
- As a joining head of security, first priority is to assess what needs to be protected
- Initial report: “if an attacker wanted in, they’re in”
- Company decided that users’ data is what needed to be protected the most
- Lots to be done initially; needed to make sure team was not demoralized
- First order of business: detection capability, measures for if the company was being attacked
- Dec. 2009: detection mechanisms raised an alarm: Chinese PLA had broken into Google’s networks; it was clear that the hackers were very experienced (Operation Aurora)
- Launched major effort to find evidence and coverage of break-in
- Detection was through large data exfiltration that made no sense; locked down compromised machines in response
- Lateral movement was discovered via DNS logs
- Difficulty: how to coordinate a response while the attackers were still in the system?
- Created new, independent accounts on Chromebooks, which were very simplistic at the time, that they were reasonably confident attackers would not have accessed
- Not the most sophisticated or devastating at the time; was the one that was best-known and most publicized
- Founders were outraged and would not hush it up; devoted resources to publicizing what happened
- This lead to a new approach: make the company care about security: what can we do to make sure that users’ data is secure?
- Run read teams internally, tell people the real things that are happening
- Founders and top execs realized that Google was actually a target of espionage
- As a result: much more of a reason to spend a lot of money and head count on security
- Increased readiness to make security changes, even if they increased risk (e.g. changes that might temporarily take a service down)
- Takeaway: demonstrated threat has a large effect on an organization and its motivations
- Operation Aurora originated from a vuln in Internet Explorer, not a Google product -> jumpstarted Google Project Zero to inform vendors of their vulnerabilities
- Starter project: get the web to switch to HTTPS/TLS everywhere
- However, this was very complicated especially due to ads ecosystem; needed to convince publishers to switch
- Another difficulty: insider threat, protecting against rogue engineers and moles
- Rapid changes to security, but also many many issues, even those that shouldn’t be issues anymore