Some Thoughts on Reuters, NY Times, and Yahoo
Benjamin Wittes flags that much of the Yahoo story is unclear, including legal arguments and the objective of the search, and further reporting from Motherboard and the Intercept further confuses the possible mechanism of the search, describing it not as a spam filter but like a "rootkit" and opening up a significant vulnerability.
But there is one factor that all stories agree on: this was a request which, for Yahoo, required a search on all messages in Yahoo's incoming mail feed, searching every incoming mail message on the behest of the government. Such searches may be reasonable but they need careful consideration. And the government's non-response is yet another in the long list of reasons why Silicon Valley doesn't trust the intelligence community. So although much remains unclear, there are now threads that seem worth commenting on.
It is already the case that email is automatically searched by the provider's robots for all sorts of things, including spam, malcode, and child exploitation materials. All these rely on the notion that having "a provider's robot search everything to find the one bad thing" is not a privacy nightmare. And while the search process may be incredibly broad, what the search looks for can be quite precise. Whether the government should be able to compel such a robot search is an open question that needs its own set of public debate, but the actual precision does matter for anyone who, like me, is already comfortable with the other robot searches which happen every day.
For example, the NSA reportedly considers known encryption keys as hard selectors, since they uniquely identify individual actors. PGP, the most popular tool for encrypting email, is particularly revealing: by default, every message includes, in the clear, the identity of all PGP keys capable of decrypting that message. It's one reason why I argue that far from being an obstacle, PGP is a gift to the NSA. So if the NSA or FBI knew that John Q Badguy used PGP key ID 0xBADD00D31337BAD1, that PGP key ID is a strong identifier that exclusively identifies John Q Badguy's communications. Any request for this key ID would require that Yahoo scan all incoming mail to discover emails to or from Mr. Badguy, but it would only reveal communications belonging to that one person.
So it may very well be that the items being searched are reasonable, even if the methods are questionable. Yet we simply don't know, because the FBI or NSA is leaving Yahoo defenseless by refusing to release a declassified order or opinion.
It is the abandoning of Yahoo that may be the biggest impact of this story. Any company not reliant on seeing customer data has an explicit incentive to "go dark" since the company itself, not the government, gets the blame when a story like this comes out. So better to simply protect customer privacy by ensuring that such orders are unenforceable. Yahoo tried very hard to fight PRISM but in the face of a $250k per day contempt order, gave up. Yet when the PRISM story emerged, Yahoo gained no credit as its fight still remained classified. I can't blame the company for not fighting again, especially if the order itself had some exigency which would prevent an appeal from delaying the need for Yahoo to follow the order.
There is a huge amount of bad blood between Silicon Valley and the intelligence community. If the IC ever wants cooperation from the big Silicon Valley firms, it should declassify as much as possible on this order immediately. A search for, say, John Q. Badguy's PGP key or some similarly strong selector may not placate the strongest of civil libertarians, but it would certainly improve the picture for everyone else. But left to fester, this simply adds one more item to the long list of "Why Silicon Valley thinks the IC should [REDACTED]"