Yahoo Encrypting Data In Wake of NSA Revelations 137
Nerval's Lobster writes "Following reports that the NSA aggressively targets Google and Yahoo servers for surveillance, Yahoo is working to encrypt much of the data flowing through its datacenters. 'As you know, there have been a number of reports over the last six months about the U.S. government secretly accessing user data without the knowledge of tech companies, including Yahoo,' Yahoo CEO Marissa Mayer wrote in a Nov. 18 blog posting. 'I want to reiterate what we have said in the past: Yahoo has never given access to our data centers to the NSA or to any other government agency.' In order to make Yahoo's systems more secure, she added, the company is introducing SSL (Secure Sockets Layer) encryption to Yahoo Mail with a 2048-bit key. That security measure will supposedly be in place by January 8, 2014. Beyond that, Yahoo plans on encrypting all information that moves between its datacenters by the end of the first quarter of 2014. Around that same time, the company will give users the option to encrypt all data flowing to and from Yahoo; it will also 'work closely with our international Mail partners to ensure that Yahoo co-branded Mail accounts are https-enabled,' Mayer wrote. (While it's not a crushing expense for massive companies such as Yahoo, introducing this sort of security does add to infrastructure and engineering costs, and takes time to actually put in place.)"
Which Encryption Scheme is Safest? Can we tell? (Score:5, Interesting)
Not mentioned was which encryption schemes Yahoo is considering. Maybe it's simply HTTPS, but is that good enough? Are there other possibilities?
Since the NSA has backdoored encryption schemes in the past, how can Yahoo determine if the scheme they implement is actually going to prevent the NSA from decrypting it? It's a serious question, and you can patly answer "you can't", but if I were responsible for implementing this scheme, this is the question I would pose to the team and require some sincere digging because it would be an even bigger embarrassment to implement the encryption, and then read another Snowden-esque revelation showing it was for nothing, and I was made a fool of.
Re:Which Encryption Scheme is Safest? Can we tell? (Score:5, Funny)
Yeah. SSL/TLS. That's just as effective against a determined state actor, with access to the telecommunication infrastructure [wikipedia.org], as a Kleenex Condom.
"You don't need to see his papers. This is the certificate you are looking for..."
Re: (Score:2)
Encryption assumes the channel is compromised. Avoiding an intercepter being able to make use of the data is the point.
Re: (Score:2)
I can middle-man the entire credential exchange - pretending to be the secure channel requestor and resubmitting the altered request for response. I can then pretend to be the responder, etc. This is how SSL examination proxy works as an edge device.
Re: (Score:3)
Will a Kleenex Condoms protect me from a man in the middle?
Re: (Score:2)
That depends if you are topping him.
Re: (Score:2)
No. They'd only need to control DNS for completely bogus keys.
Now. How'd they pull that off?
Beside which, getting real private keys hasn't been hard for them, it seems. Got PEM?
Re: (Score:2)
And do you trust all the ssl certificate authorities? How many of those are based in the US and thus fall under the jurisdiction of the NSA? Come to think of it so does yahoo, who's to say the government wont simply demand that they hand over all their keys?
Re: (Score:2)
If the government demands Yahoo turn over all transmissions they fight it. If they lose (which they likely will) it becomes hard to keep that secret from lots of employees who might squeal.
Re: (Score:2)
... particularly if those employees are foreign nationals (from an American perspective) living abroad under their native jurisdiction.
Which is where the Assange question suddenly veers back into the spotlight.
Re: (Score:2)
Yep exactly Especially since from their perspective helping the NSA spy on the citizens of their country is a crime, preventing it their duty.
Re: (Score:2)
There was a story recently about how yahoo did fight such a demand, lost, and were forced to keep it a secret...
Re: (Score:2)
Unless it was very recently yahoo is publicly denying it. Companies don't generally lie that directly.
Re: (Score:3)
If the government has the keys it doesn't matter how many bits they use.
cop with a warrant (Score:3)
+1 Insightful on the "government has the keys" point...
here it is: law enforcement & NSA must have the ability to access anything, given proper rights & proceedures
no one can make successful counter-point...all arguments are arguments over ***under what conditions*** the LE/NSA can access the information
Yahoo is doing absolutely nothing other than PR 'damage control' by manipulating the facts with this news.
Yahoo will give up **anyone's** data as fast as humanly possible when asked by a legal author
Re: (Score:2)
LE has to get a subpoena and that's overseen by the courts.
NSA is semi bypassing the entire system and that's overseen by congress and intelligence courts for whatever that's worth.
What Yahoo is avoiding is being spied on not being subject to a court order.
howso? (Score:2)
howso?
FISA courts are there exactly for this right? Are you assuming the government is up to something or can you be specific?
I'm talking **current policy**
Re: (Score:2)
FISA courts aren't really courts in that the court doesn't have the same degree of authority nor the same checks as a normal court. That's why I consider it semi-bypassing. It is arguably either a court only in name or a weak court. On the other hand this structure is overseen by congress.
Comment removed (Score:5, Informative)
Re: (Score:2)
As with everything, there's a level of third party trust (the certificate authorities) or shoe-leather (exchanging keys in person) that's required regardless of the ciphers you end up using. That's a whole different discussion though.
You're of course right in pointing out the distinction between the transfer protocol and the encryption.
I don't believe Yahoo (or any other big player) facilitate the shoe-leather alternative though, it's third party certification or nothing.
Re:Which Encryption Scheme is Safest? Can we tell? (Score:5, Informative)
IPSec is no more an encryption scheme than HTTPS. Both are protocols that use authentication and encryption schemes, they just work at different layers of the stack.
Re: (Score:2)
Modern encryption works by making it take a very, very, very long time to brute-force the encrypted data. Part of that lengthy time is the hardware involved in the brute-force effort.
The NSA has resources well beyond what are available to the rest of us - the joke is the NSA measures its computing power in acres.
Add that to large budgets to develop specialized hardware, and nice standard encryption algorithms to target with that hardware, and it's not clear that the NSA can't read everything. Encrypted or
Re:Which Encryption Scheme is Safest? Can we tell? (Score:5, Insightful)
The issue is not whether they can brute force encryption.
We already assume they have the capability of brute forcing all encryption within a reasonable time frame. Something hilariously well protected? 3-6 months.
That being said, the NSA, still only has so many units of discrete work it can perform in a given period . Now, unless you are going to try to convince me that the NSA has computing power many orders beyond the total computing power of the entire planet, it means there is still safety in numbers.
Mass. Surveillance.
That's the real game. That's the real threat to privacy and freedom. If everyone makes sure that the NSA has to waste those work units decoding a pair of testicles you sent to your best friend, the NSA is still left with picking and choosing its battles .
I'm okay with that. If the NSA really can break all of my communication and files within a week or two, but can only do it for several dozen Americans at a time during that period, we are all still protected as a whole. The NSA can still do its job. Yes, there was an original job they ostensibly are supposed to perform in my best interests.
The sheer magnitude of what would need decryption for mass surveillance makes it illogical to worry about, IF WE ARE USING ENCRYPTION EVERYWHERE AND ZERO-KNOWLEDGE 3RD PARTY SERVICES. I can't stress that last part enough.
Re: (Score:2)
They can't brute force. Do the math. There are 2.8x10^147 primes which are good candidates for 1024-4096 bit keys. If you were to use every atom in the universe to compute 1 prime per nanosecond you still ain't close to brute forcing your way out.
Re: (Score:2)
We don't know what they have. Never underestimate the "geek factor". There may very well be a vulnerability that we have no idea of that reduces the complexity sufficiently that decryption is possible within a viable time frame.
I'm taking a conservative view point on what real risk the NSA could pose with ubiquitous encryption.
Re: (Score:2)
There are many algorithms for reduction of the decryption problem. Orders of magnitude get pulled off aver 5 years or so. But that ain't brute force.
I'd say right now it is likely the complex reduction algorithms still aren't good enough to overcome the sheet mathematical complexity given 256-bits. Certainly for something like 1024-4096 nowhere near.
Re:Which Encryption Scheme is Safest? Can we tell? (Score:5, Informative)
Yes, that is how encryption works. But if your key is large enough, the time & energy to brute force it will take much longer than your lifespan. As an example I just googled, brute-forcing AES-128 at 10 Petaflops would take 10 quintillion years (10^18). http://www.eetimes.com/document.asp?doc_id=1279619 [eetimes.com]
The _real_ concern is that the NSA knows of weaknesses in these encryption schemes, and doesn't have to brute force it.
Re: (Score:2)
As an example I just googled, brute-forcing AES-128 at 10 Petaflops would take 10 quintillion years (10^18).
Brute-forcing Yahoo's CIO at about 1 lash per second with a rubber hose won't even take 5 minutes. And a single National Security Letter will shut the whole thing down anyway.
Where's the beef?
Re: (Score:2)
The simplest concern is the way sites are authenticated by certificate authorities... Some of those certificate authorities are under US jurisdiction and thus beholden to the NSA, and others are under the jurisdiction of other governments who may well want to do the same thing.
Re:Which Encryption Scheme is Safest? Can we tell? (Score:4, Informative)
No one is ever going to brute force a 256-bit symmetric key. Even if you imagine a matrioshka brain (turn the entire energy output of a star into computation) it would take longer than the age of the universe. A 128-bit symmetric key is safe from brute force vs all realistic threats.
If the math is flawed, OTOH, or your "random" key wasn't so random, it's easy (there is deep suspicion about the RNG built into Intel procs these days).
Re: (Score:2)
The NSA has resources well beyond what are available to the rest of us -
Yawn. This myth has been debunked millions of times now.
Short version: Brute-force of 128-bit encryption cannot be done.
Longer version: The laws of thermodynamics are what prevent it, not the size of the NSA's budget.
Re: (Score:2)
To brute-force a 4096-bit key pair in a reasonable time would require not acres of computers,
but entire galaxies composed of nothing but computers.
And these computers would have to be able to communicate with each other instantly
and not be limited by the speed of light.
Re: (Score:2)
Maybe it's simply HTTPS, but is that good enough?
No, not really. In security, you don't have infallible trust in a system which isn't verifiable -- which means, just because Yahoo says they are doing transport encryption in their data center doesn't automatically make that system trustworthy. You need proof that they have done what they said. Otherwise, it's just their word.
Secondly, Yahoo's biggest problem is the data laying on the disks. They can encrypt the traffic all they want that doesn't do anything for the mail stores on disk. Ms. Mayer is basic
Re:Which Encryption Scheme is Safest? Can we tell? (Score:4, Interesting)
Most of the SANs I've seen support disk encryption and IPSec encryption between the SAN and the host or OS talking to it. If your OS writes encrypted data to storage (encrypted filesystem) as well, you have two layers of encryption on the platter and two layers of encryption in transit.
Of course that doesn't address weaknesses in ciphers or key exchange systems, but it seems like it would make it a lot harder to get at the data because the only place it is decrypted is during interprocess communication (decrypting from the filesystem and before re-encrypting it for final transit to client).
Not that this trivializes that risk, but it seems to make it a lot tougher.
Re: (Score:2)
I'd be surprised if any "big data" uses SAN. I don't know about Yahoo, but Google, MS, and I'm pretty sure Amazon all use simple direct-attach storage. It's a bit silly to be worried about anyone reading the data off of 10000 servers through some backchannel without being noticed. Encrypting the links between those servers would accomplish a lot, IMO.
Re: (Score:2)
And I would imagine that the distributed filesystems they use on those systems probably aren't even very coherent even if you could read a physical disk.
Re:Which Encryption Scheme is Safest? Can we tell? (Score:4, Informative)
It depends on where the "brains" are. Facebook (IIRC) has the redundancy on the backend app layer where coupled with NoSQL, if something drops... there is some redundancy built in somewhere to pick it off, or drop a couple tuples, but the tables still have their integrity. Whole servers can drop off the map, and Facebook will keep going. Isn't pretty, but their model really can handle stuff getting tossed here and there.
Apple, on the other hand, uses Teradata systems with NetApp appliances on the backend, so one large cloud provider does go with the more traditional storage stack model found in the enterprise. However, unlike losing a FB post or two, a user losing chunks of their data would not be a good thing, so Apple's model tends to be more rigidly ACID compliant.
Re: (Score:2)
Facebook puts all "content" data on MySQL (only Google uses NoSQL for content, everyone else uses sharding over SQL servers).
Apple uses Azure for their cloud stuff, you might be thinking of their internals (for billing and customer data and so on) with that storage stack.
Re: (Score:2)
Thanks for the correction. I got the Teradata/NetApp info from an article of when their data center originally opened, and it was claimed that was their mainstay for storage.
Re: (Score:2)
Only a small fraction of people understand the technology enough to know that proof is required, not only proof that encryption is being used but also that its implemented correctly and the keys are securely stored.
The vast majority of people will just read the marketing literature and assume that yahoo aren't trying to mislead them. They don't understand how these things work and don't care to, they simply put blind trust in what they're being told.
Re: (Score:2)
Re: (Score:3)
Encryption schemes are important, but whenever someone mentions "let's encrypt it", I cringe.
Encryption isn't some magic switch that you turn on and all your data is 100% secure from bad guys. What happens is that it makes a smaller chunk of data (i.e. the key or keys) the valuable part.
Key management isn't a cookie-cutter thing. Error on security, and your data can't be recovered. Error on accessibility, and the bad guys now have your keys and can get your data.
A small company can get by with burning an
Re: (Score:2)
For [1], printing the key out in a sequence of QR codes (or variant) would probably be a good compromise between plaintext paper and disc. QR codes have better built in error correction than teams of people typing in a long sequence of arbitrary characters...
Re: (Score:2)
I've tried that, with varying success. A lot of readers will not work with the larger QR codes (which would be needed for decoding a 2048-4096 bit key block), and other readers just give up with regards to alignment if the block is too big.
If there is very standard code system which would work for this (a QR variant, since it does have built in error correction), it is a good thing to have included with a human readable (and retypable) output.
Re: (Score:1)
Well,
My Guess is SSL/TLS is fine but the trust system is broken against a motivated state actor.
(forgive my terminology...it's been a while since I've delt with certificates)
1.) Steal the private key from Google, Yahoo, etc.
2.) Force a trusted certificating authority to issue a leaf/signing certificate. Then the government can issue SSL certs for Yahoo, Google, etc on behalf of that certificating authority who gave them a leaf certificate.
3.) Steal the signing certs of a root authority
4.) Setup a shell trus
Re: (Score:1)
...how can Yahoo determine if the scheme they implement is actually going to prevent the NSA from decrypting it? It's a serious question
Yes it is. And these SE-Corps (looking at you GOOGLE!) should be much more vocal and transparent in letting the public know the predicaments they as a Corporation are in, what factors are there to consider, and their respective weights, what the options are that they're contemplating, and what their decisisions, when made, are based on.
In short -- your users may be searchers, and as such they are learners They are not dumb and will not be kept in the dark!
To NOT mention, to AVOID SUBJECT, to be SILENT, ar
Re: (Score:3)
Since the NSA has backdoored encryption schemes in the past, how can Yahoo determine if the scheme they implement is actually going to prevent the NSA from decrypting it?
You have to understand that any key based encryption technique is breakable. It doesn't matter what key based technique you use, it can eventually be brute forced. All you can hope to do is make it take a very long time to decode, so long that the message becomes not worth the effort.
There are "unbreakable" techniques, but they all require a one use a random pad that both parties know, but never disclose or reuse. That's about the *only* way to make sure the NSA cannot decode your stuff. Good luck doing
Re: (Score:1)
"...any key based encryption technique is breakable... it can eventually be brute forced."
Sure. Anything can happen when "eventually" is "the age of the universe".
Re: (Score:2)
Email is insecure by design (Score:2)
Re: (Score:2)
Use multiple encryption schemes from different providers, and shuffle them around intermittently. So that one week the NSA can decrypt the data stream to reveal material that has been encrypted by a Russian system, but the next week they get something that has been encrypted by a Chinese system. And the same goes for the Russians at their t
Re: (Score:2)
I love how so many people here are so sure the NSA can do magic on encryption... it's like complexity doesn't exist. They can solve any problem expressible in a general formal system! Halting problem? Fuck the halting problem, that's like stealing candy from a child.
NSA - the Chuck Norris of the agencies.
Re: (Score:3)
It's not that we think the NSA can brute force SSL; we think the NSA has compromised the certificate authorities.
Re: (Score:3)
I just picked that comment because it said "thru [sic] ssl", and I interpreted as "breaking DH" or something. But I was referring to a somewhat spread sentiment that breaking encryption is just a matter of developing technique, which may not be the case (hence my sarcastic reference to the halting problem & Godel's incompleteness theorem).
Like this post above [slashdot.org]:
The issue is not whether they can brute force encryption.
We already assume they have the capability of brute forcing all encryption within a reasonable time frame. Something hilariously well protected? 3-6 months.
Re: (Score:2)
Re: (Score:1)
The DH exchange only works if you don't get Man-in-the-Middled. Thats the point. Once the certificate authority is compromised, they can create a cert that makes them appear as the server you think you're supposed to talk to, so you do the DH exchange with their server, so the DH exchange isn't a problem. Then they just make another connection on to the destination which does a whole new DH exchange for keys.
Re: (Score:2)
Weasel words (Score:1)
Yahoo CEO Marissa Mayer wrote in a Nov. 18 blog posting. 'I want to reiterate what we have said in the past: Yahoo has never given access to our data centers to the NSA or to any other government agency.
The operative phrase here is "our data centers". A little less than half the data centers that Yahoo have their servers in are not owned by Yahoo, they lease space there. So, Yahoo's data flows in and out of the cage(s) they have their servers in into the house network. You can work it out from there.
Took them way too long (Score:2)
Re: (Score:2)
Well, actually it's quite embarrassing that they're only doing this now...
I agree. It is amazing what little effort companies make to "protect" their data. They seem to think that having a password is all that is needed.
Open bag (Score:2)
Insert cat.
She's given me 1/2 of what I want to hear (Score:3)
Strongly worded without PR-crafted terminology. Now, have you given these entities private information without a warrant?
Does it even matter... (Score:4, Insightful)
...if they can be forced to turn over encryption keys at the whim of some NSA/government authourity?
Er...what about encryption at REST? (Score:2)
>> encrypt all data flowing to and from Yahoo
BFD - since all the data is still sitting on servers somewhere, why would this offer any protection at all?
>> introducing this sort of security does add to infrastructure and engineering costs
BFW - welcome to 2008, Yahoo.
Yahoo can't even keep spam out of my inbox (Score:4)
Re:Yahoo can't even keep spam out of my inbox (Score:4, Interesting)
While I would have agreed with you two weeks ago, bizarrely, I have recently started getting a ton of spam in my Gmail account - really obvious stuff that should have been filtered. And Yahoo has been almost perfect filtering the same crap. Several people I have talked to have noticed the same thing. It's almost like someone at Google accidentally turned off the spam filter...
Re: (Score:2)
Really. (Score:3)
Doesnt do any good, if the law enforcement organizations (etc), have a warrant they can record all traffic from your IP/Phone. Depends on the company, but at AT&T Wireless they could turn on full sniffing from a mobiles internet traffic and record all TCP/UDP and even overlay it with location based service (tower strength triangulation). My boss said they had a group to assist in warrants, but after I setup the servers and routers, I NEVER saw an email, name or department identified, and I worked there for years setting up hardware from old packet data to 3G routers before I left.
So anyways, they record the entire SSL handshake so they can decrypt the session. You too can even try it for yourself in wireshark.
And who knows what is going on at the AT&T datacenters in those secret rooms...
This will never work. (Score:2)
That security is going to last as long as it takes to find one exploit against an endpoint that can be used to pull the key out of memory one time.
Ya Who? (Score:1)
Useless (and an obvious deception) (Score:3)
But didn't NSA compromise SSL already? (Score:2)
Waste of time - all keys already held (Score:1)
I agree with someone who suggested one of the early pre-NSA encryption schemes.
You'd be better to roll your own, mind you. Remember, they already have your make files if you used Win 8 or Win 8.1, since it "indexes your local drive for fast search" which is a polite way of saying "spies on you".
Re: (Score:2)
Re: (Score:1)
Most people are not competent to write their own encryption system. It takes a lot of very advanced math and a very careful implementation to make a secure comm system.
Last time I checked there were many hundreds of millions of people on the Internet.
Are you saying none of them can roll a new encryption scheme?
We used to do stuff with only 50 people working worldwide on stuff.
I think you just want to believe it's "too hard" so you don't have to do any actual work.
But how is the NSA going to fix Flickr's bad UI (Score:2)
if you change the code and lock them out?
now, if they would fix the UI (Score:2)
I don't just blame the NSA (Score:2)
I don't just blame the NSA for this situation. The providers are at fault for assuming that leased lines can be run unencrypted between their data centers because they're "private". Any time data enters or leaves a data center, one should assume it is being monitored. Everyone knows that's the most basic tenet of security.
But all these lazy vendors from Google to Yahoo and Microsoft and hundreds of others have taken the easy, lazy way out for years.
We all owe Snowden a big "Thank you" for kicking th
Re: (Score:2)
Bad news (Score:2)
I already had trouble understanding Marissa Meyer.
Erm, remember Lavabit (Score:2)
Opportunistic encryption ... (Score:2)
I remember the paranoid rantings of those in the FreeS/WAN community back in the day (that's IPSec software for Linux fyi) about needing opportunistic encryption support and DNS based keys so any two hosts on the Internet could communicate securely and prevent Big Brother from listening. ... "Its
I also recall that I wished it would work, and set up my own hosts with it, but it never did work well and there just weren't enough participants to hit critical mass.
Thirdly I remember a quote from my old BBS days
Hey Crypto Nerds! (Score:2)
http://xkcd.com/538/ [xkcd.com]
This is NOT an issue with encryption.
This is an issue with a wrench. You can have it encrypted 5 different ways, but when the NSA comes a knocking, DEMANDING The data, and your alternative is to get shut down, go to jail, etc... guess what, they key's become suddenly available anyway.
Its another type of brute force encryption hacking that always succeeds. The RIAA and MPAA figured this out (mostly) long ago when they realized that from a technical standpoint it is a no win situation. At
NSA can crack SSL (Score:1)