4 ms·

this isn't 'everything' - the article 'only' mentions phone calls and emails. that's the easy stuff NSA can access from ISPs like AT&T, Verizon, etc. that's i

this isn't 'everything' - the article 'only' mentions phone calls and emails. that's the easy stuff NSA can access from ISPs like AT&T, Verizon, etc. that's important because it doesn't contradict respected people at Google, or require magic crypto cracking. i'm repeating myself here, so sorry, but there are two separate issues: 1 - nsa is getting a lot of data from the people routing traffic. likely all metadata. and perhaps some (significant) subset of unencrypted traffic. 2 - despite the misleading prism slide, they don't seem to be getting complete, direct access to the web companies' data. the uproar about the slides put the initial emphasis on the web companies and has confused discussion since. but the consistent story that seems to be emerging is that it's the ISPs, who have much more of a tradition of collaborating with NSA, who are providing a lot more data than people expected. and that data is being stored without legal approval, relying on the idea that "no-one looks at it" without a warrant. at least, that's the most consistent view i can find. what worries me most at the moment is how much hardware would be needed to store and process all that data, and whether that is feasible.

marshray 13y ago
http://www.wired.com/threatlevel/2012/03/ff_nsadatacenter/all/ http://www.wired.com/threatlevel/2012/03/ff_nsadatacenter/al... As a result of this “expanding array of theater airborne and other sensor networks,” as a 2007 Department of Defense report puts it, the Pentagon is attempting to expand its worldwide communications network, known as the Global Information Grid, to handle yottabytes (1024 bytes) of data. (A yottabyte is a septillion bytes—so large that no one has yet coined a term for the next higher magnitude.) It needs that capacity because, according to a recent report by Cisco, global Internet traffic will quadruple from 2010 to 2015, reaching 966 exabytes per year. (A million exabytes equal a yottabyte.
- whiddershins 13y ago
  so, according to this article, their intended capacity is adequate to store all internet traffic. All da bytes.
- mikegioia 13y ago
  This is insane. Apparently Titan can do 10-20 petaflops per second. I didn't think before, but they may be decoding AES-128.
  - betterunix 13y ago
    I doubt that they are decoding AES, except perhaps for the highest-value targets. I would guess that that computing power is a shared resource, not something that can be dedicated to a single task like that. They are probably using it for numerous data mining applications, probably a lot of traffic analysis, some signal processing, etc. Even if the NSA has the resources needed to crack AES128, those resources would almost certainly be better used on other tasks.
    - marshray 13y ago
      I agree. There is no known attack on AES-128 which comes anywhere close to the capability of any known supercomputer within any reasonable time period. Except maaaybe the aggregate computing power of the entire PlayStation Network for a long long time. I haven't run the numbers for AES. EDIT: Yeah, no way without an algorithmic breakthrough.
      - jacquesm 13y ago
        There is no publicly known attack, it is possible that there are undisclosed vulnerabilities in AES-128 that the NSA is aware of today. Is there any precedent for a weakness like that where an agency like the NSA was already aware of a weakness before it became public?
        marshray 13y ago
        A couple of cases come to mind where the NSA is seemed to demonstrate superior awareness of cryptanalytic attacks. But note that in neither case did the NSA recommend the use of a cipher that was weaker than the number of bits on the label. 1. Differential Cryptanalysis https://en.wikipedia.org/wiki/Data_Encryption_Standard#NSA.27s_involvement_in_the_design https://en.wikipedia.org/wiki/Data_Encryption_Standard#NSA.2... A general technique, but the DES standard turned out to have been developed to be resistant to it before it was known about in public circles. 2. Impossible differential cryptanalysis https://en.wikipedia.org/wiki/Impossible_differential_cryptanalysis https://en.wikipedia.org/wiki/Impossible_differential_crypta... This general technique was invented and quickly broke 31 out of 32 rounds of Skipjack. Schneier wrote in 1998 "I can't imagine that the NSA would field an algorithm with 32 rounds when they could break 31, so I believe that the NSA did not know about this attack. And the cryptanalysts are still working on improving their attack, so it is possible that they will get to 32 rounds." But no one yet has.
leoc 13y ago
The telcos are a bigger problem if you are living in the USA. If you are a non-US-citizen outside the USA, the web companies are a bigger problem, because a) you're probably well out of the reach of Verizon (though also within reach of your domestic phone snooping) and b) unlike an American or US resident, you have no legal protection if the NSA decides it would like to look at your web-company data for any reason, though it still has to issue an individualised order to make you one of the few thousand people or groups it's following on Facebook (as 'twere).

marshray 13y ago

http://www.wired.com/threatlevel/2012/03/ff_nsadatacenter/all/ http://www.wired.com/threatlevel/2012/03/ff_nsadatacenter/al... As a result of this “expanding array of theater airborne and other sensor networks,” as a 2007 Department of Defense report puts it, the Pentagon is attempting to expand its worldwide communications network, known as the Global Information Grid, to handle yottabytes (1024 bytes) of data. (A yottabyte is a septillion bytes—so large that no one has yet coined a term for the next higher magnitude.) It needs that capacity because, according to a recent report by Cisco, global Internet traffic will quadruple from 2010 to 2015, reaching 966 exabytes per year. (A million exabytes equal a yottabyte.

whiddershins 13y ago
so, according to this article, their intended capacity is adequate to store all internet traffic. All da bytes.
mikegioia 13y ago
This is insane. Apparently Titan can do 10-20 petaflops per second. I didn't think before, but they may be decoding AES-128.
- betterunix 13y ago
  I doubt that they are decoding AES, except perhaps for the highest-value targets. I would guess that that computing power is a shared resource, not something that can be dedicated to a single task like that. They are probably using it for numerous data mining applications, probably a lot of traffic analysis, some signal processing, etc. Even if the NSA has the resources needed to crack AES128, those resources would almost certainly be better used on other tasks.
  - marshray 13y ago
    I agree. There is no known attack on AES-128 which comes anywhere close to the capability of any known supercomputer within any reasonable time period. Except maaaybe the aggregate computing power of the entire PlayStation Network for a long long time. I haven't run the numbers for AES. EDIT: Yeah, no way without an algorithmic breakthrough.
    - jacquesm 13y ago
      There is no publicly known attack, it is possible that there are undisclosed vulnerabilities in AES-128 that the NSA is aware of today. Is there any precedent for a weakness like that where an agency like the NSA was already aware of a weakness before it became public?
      - marshray 13y ago
        A couple of cases come to mind where the NSA is seemed to demonstrate superior awareness of cryptanalytic attacks. But note that in neither case did the NSA recommend the use of a cipher that was weaker than the number of bits on the label. 1. Differential Cryptanalysis https://en.wikipedia.org/wiki/Data_Encryption_Standard#NSA.27s_involvement_in_the_design https://en.wikipedia.org/wiki/Data_Encryption_Standard#NSA.2... A general technique, but the DES standard turned out to have been developed to be resistant to it before it was known about in public circles. 2. Impossible differential cryptanalysis https://en.wikipedia.org/wiki/Impossible_differential_cryptanalysis https://en.wikipedia.org/wiki/Impossible_differential_crypta... This general technique was invented and quickly broke 31 out of 32 rounds of Skipjack. Schneier wrote in 1998 "I can't imagine that the NSA would field an algorithm with 32 rounds when they could break 31, so I believe that the NSA did not know about this attack. And the cryptanalysts are still working on improving their attack, so it is possible that they will get to 32 rounds." But no one yet has.

leoc 13y ago

The telcos are a bigger problem if you are living in the USA. If you are a non-US-citizen outside the USA, the web companies are a bigger problem, because a) you're probably well out of the reach of Verizon (though also within reach of your domestic phone snooping) and b) unlike an American or US resident, you have no legal protection if the NSA decides it would like to look at your web-company data for any reason, though it still has to issue an individualised order to make you one of the few thousand people or groups it's following on Facebook (as 'twere).