8 ms·

How to Detect Sneaky N.S.A. ‘Quantum Insert’ Attacks

123 points by jakobdabo 12y ago

ilaksh 12y ago
Why do we permit these groups to act as if they were in a James Bond movie? Only incredible ignorance and lack of perspective could permit this. Its just criminal activity excused by puerile worldviews shaped by propaganda and elitist 18th century philosophies propagated by powerful thugs.
- anonymfus 12y ago
  May be people assume that things already work as shown in such movies, a la http://en.wikipedia.org/wiki/CSI_effect http://en.wikipedia.org/wiki/CSI_effect
- higherpurpose 12y ago
  Most people who support mass surveillance when asked in surveys only do it because they don't fully understand what's happening. Let's go back to John Oliver's recent show episode where people seem very concerned that the government would have their naked pics - yet they were completely unaware that the government actually does that (as in collecting everything, including their naked pics). So most people just don't fully (or even partially) grasp the government's surveillance power right now. The media doesn't help here. The media is supposed to inform people about this, but instead either it doesn't talk too much about it, or if it does, the mainstream media is usually pro-mass surveillance (because the powerful friends of the networks are).
  - pdkl95 12y ago
    While the Snowden leaks were being released, I made a habit of doing a few very general searches to get a sampling of how the events were being reported. I really wish I had taken it a step further and saved regular tables of those observations, because I suspect it could have been made into a very interesting paper. Most of the time the lazy reprints of an AP/Reuters story.would dominate, as expected. Once and a while - usually after a really import release such as COTRAVELER[1] - the story would happen... and then suddenly everybody is talking about Prism and explaining how it is "just metadata". Any momentum the new story had ended was redirected into topics that were already known. There would be small-time media doing proper reporting, of course, but it was clear who the "big media" worked for. The media following the orders of those that sign their paychecks isn't really news, but it was very interesting to watch it happen in realtime. [1] I still believe that COTRAVELER is one of the most important thing we have learned from Snowden, as it builds relationship maps and it doesn't rely on the target performing some specific action such as an HTTP requests that QUANTUM can race. Bonus: it only relies on "metadata".
    - pakled_engineer 12y ago
      The discussion about COTRAVELER was usually totally poisonous too https://firstlook.org/theintercept/2014/02/24/jtrig-manipulation/ https://firstlook.org/theintercept/2014/02/24/jtrig-manipula...
  - LLWM 12y ago
    That's a bit patronizing. Of course it is true that most people who support surveillance don't understand what's happening, but only because most people in general don't understand what's happening. Plenty of us understand what's happening and also support it - anecdotally, at least. Fully understanding is asking for too much, though. I doubt anyone outside of a select few at the NSA fully understand what's happening, and it seems reasonable to guess that they support it.
- rhino369 12y ago
  Are you asking why America (along with most countries) spy? Or why they try to be good at spying? America spies because it wants information it can't otherwise have. What are Russia's plans in Ukraine? Is Pakistan playing both sides re: Taliban? Can China shoot down a B2 bomber? We want to know and nobody and asking won't help. Why we allow them to be better at their art is obvious.
  - smokeyj 12y ago
    This is like breaking into your neighbors house to see if they have any plans to break into your house. How tragically ironic.
dtech 12y ago
TLDR: Send spoofed TCP package with HTTP redirect to malicious site right after victim does a GET request. This attack, while impressive and nifty, would seem to be easily subverted by SSL right?* Yet another example that shows that non-SSL HTTP should be phased out. * In the ideal situation where users check URL's and the CA system is good enough to keep the NSA out.
- ilaksh 12y ago
  Why shouldn't we assume the NSA has subverted the CA system? Why would we all rely on central authority for security if not to support centralized surveillance?
  - Dylan16807 12y ago
    Because that subversion can be relatively easily detected, and then shouted to the world.
    - motoboi 12y ago
      Given the episode where stolen Realtek certificates been used for years by a government-made malware, I think we should not be so sure that they can't do that. A single request going through a bogus certificate emitted by NSA "acquired" CA would be hardly detected. Or, given the fact that they could indeed broke keys, they could simply do a man-in-the-middle, instead of a man-in-the-side.
      - Dylan16807 12y ago
        A stolen key is a server-being-hacked problem, not a CA problem. If the server is hacked, game over. There are browser extensions that detect bogus certificates; there's a good risk of catching them.
        SG- 12y ago
        Should we pretend the NSA and various organizations don't have the means to infiltrate a CA?
tux 12y ago
I think this type of attack will be useless if you use DNSCrypt + VPN (encrypted OpenVPN/SSH traffic) not to mention TOR. Also, there is a new type of VPN @ https://www.goldenfrog.com/vyprvpn/chameleon https://www.goldenfrog.com/vyprvpn/chameleon
- SG- 12y ago
  Tunneling is good if they are targeting you locally, but if they see tunnel data going to a server which is easy to see they'll simply target you at the other end.
  - tux 12y ago
    Good luck with that when there is thousands of people using the same VPN service, also randomization of IP addresses. However this is only if Evercookie are dealt with. @ http://en.wikipedia.org/wiki/Evercookie http://en.wikipedia.org/wiki/Evercookie Blocking Evercookie blocking is relatively easy now, if you disable JavaScript and use things like NoScript and Random Agent Spoofer. But then also don't forget about the recent VPN's ip leak through WebRTC which can be disabled with setting media.peerconnection.enabled to false If you're paranoid you can also use Qubes or Tails OS.
    - SCHiM 12y ago
      Hmm I'm not sure if a vpn will protect you from this attack. Your traffic will need to cross out of the vpn somewhere/sometime. Unless you're connecting to a hidden service or something. It actually depends more on whether you use https on your first connection or not. Because even if the server is set up to only accept https and it tries to redirect you at the first opportunity you'll be vulnerable. Generally a server configured to use https only will still listen on port 80, and send a 302 to redirect you to it's https version. That redirection can be hijacked by QI.
      - tux 12y ago
        Well if that true then everyone has much bigger problem. Because then any VPN services becomes useless. It would be nice to see if Fox-IT could do similar tests by using VPN/TOR. But I think that VyperVPN still is the most secure VPN if it really does have this; "Chameleon scrambles OpenVPN packet metadata to ensure it’s not recognizable via deep packet inspection (DPI)"
yAnonymous 12y ago
Aren't there firewall rules to prevent IP spoofing in most civilized countries?
- uulbiy 12y ago
  IP spoofing can be prevented by ISPs[1] but I don't thing there is an incentive for them to do it. That could not stop the NSA anyway. [1]: http://security.stackexchange.com/questions/1062/why-dont-isps-filter-on-source-address-to-prevent-spoofing http://security.stackexchange.com/questions/1062/why-dont-is...
  - SCHiM 12y ago
    Also some (many?) ISPs have asymmetric routing, preventing an easy implementation of anti-spoofing mechanisms.
- zurn 12y ago
  "Great Firewall of Belgium only allows ingress traffic from non-Belgian source addresses" wouldn't prevent IP spoofing inside or outside Belgium. Nor would doing it the other way around. If such national firewalls existed in all countries and forced all IP traffic to flow through them, it might help some.
- tptacek 12y ago
  No.
lawnchair_larry 12y ago
I'm not sure why the comments here are focused on how the attack, as that is not the news and was well known for quite some time. This article is about how to thwart the attack. tl;dr Watch for the site to appear to respond twice with the same sequence number. One of those is the NSA trying to get in ahead of the real site.
- pekk 12y ago
  If the attack exists in the wild, its use isn't in any way limited to the NSA.
  - tptacek 12y ago
    This attack has existed in the wild since the mid-1990s; it is literally the reason SSH exists. We didn't call our TCP hijacking tools "quantum inserts" back then, but that's exactly what they were. The only reason we didn't redirect HTTP connections to malicious sites is that in the 1990s, that would have been a waste of time; the r-commands were much lower hanging fruit.
    - tedunangst 12y ago
      TIL the NSA invented rlogin.
    - schoen 12y ago
      An important example from that timeframe publicly documenting the risk you're referring to: https://www.usenix.org/legacy/publications/library/proceedings/security95/full_papers/joncheray.pdf https://www.usenix.org/legacy/publications/library/proceedin...
      - tptacek 12y ago
        In fact, Joncheray "invented" this attack.
  - csears 12y ago
    True, but state actors are much more likely to have ISP/backbone fiber taps and DPI equipment in place to make the attack really effective at scale.
4096 12y ago
The blog post with all the technical details: http://blog.fox-it.com/2015/04/20/deep-dive-into-quantum-insert/ http://blog.fox-it.com/2015/04/20/deep-dive-into-quantum-ins... And the Github account with the tools to do this yourself, PCAP and rules for snort, bro and suricata: https://github.com/fox-it/quantuminsert https://github.com/fox-it/quantuminsert
- smellf 12y ago
  Thanks for the technical link. You gotta love Wired, paraphrasing down to get gems like: “The first TCP packet will be the ‘inserted’ one while the other is from the real server, but will be ignored by the [browser]”. Oh my browser deals with TCP packets??
nayuki 12y ago
I found the title to be misleading. I thought it referred to the use of quantum computing to break public key cryptography and inject data. But instead, it's just some dumb TCP man-in-the-middle spoofing, only a bit more sophisticated than Comcast's RST attacks.
- schoen 12y ago
  QUANTUM is NSA's own name for its man-on-the-side attacks, so the infelicitous naming started elsewhere, not with the author of this article.
eyeareque 12y ago
Are there any details on when Quantum Insert started to be used by the gov? I wonder if the government built it off of the idea shared here: http://www.willhackforsushi.com/presentations/PenTest_PerfectStorm_Part_1.pdf http://www.willhackforsushi.com/presentations/PenTest_Perfec... (See slide 11-14) These slides describe the wifi tool called AirCSRF, which does basically the same attack, but over open wifi.
terminado 12y ago
Hmmm, what a short article. All it says is: It's actually impossible to detect these kinds of attacks. Please go about your business.
- un1xl0ser 12y ago
  I completely agree. While theoretically possible to do given full packet captures, it does not scale. IDS/IPS systems are not in a place to operate at this level and keep this much state. It's more likely that it would be detected based on the nature of the injected malicious data.

123 points by jakobdabo 12y ago

Why do we permit these groups to act as if they were in a James Bond movie? Only incredible ignorance and lack of perspective could permit this. Its just criminal activity excused by puerile worldviews shaped by propaganda and elitist 18th century philosophies propagated by powerful thugs.

anonymfus 12y ago
May be people assume that things already work as shown in such movies, a la http://en.wikipedia.org/wiki/CSI_effect http://en.wikipedia.org/wiki/CSI_effect
higherpurpose 12y ago
Most people who support mass surveillance when asked in surveys only do it because they don't fully understand what's happening. Let's go back to John Oliver's recent show episode where people seem very concerned that the government would have their naked pics - yet they were completely unaware that the government actually does that (as in collecting everything, including their naked pics). So most people just don't fully (or even partially) grasp the government's surveillance power right now. The media doesn't help here. The media is supposed to inform people about this, but instead either it doesn't talk too much about it, or if it does, the mainstream media is usually pro-mass surveillance (because the powerful friends of the networks are).
- pdkl95 12y ago
  While the Snowden leaks were being released, I made a habit of doing a few very general searches to get a sampling of how the events were being reported. I really wish I had taken it a step further and saved regular tables of those observations, because I suspect it could have been made into a very interesting paper. Most of the time the lazy reprints of an AP/Reuters story.would dominate, as expected. Once and a while - usually after a really import release such as COTRAVELER[1] - the story would happen... and then suddenly everybody is talking about Prism and explaining how it is "just metadata". Any momentum the new story had ended was redirected into topics that were already known. There would be small-time media doing proper reporting, of course, but it was clear who the "big media" worked for. The media following the orders of those that sign their paychecks isn't really news, but it was very interesting to watch it happen in realtime. [1] I still believe that COTRAVELER is one of the most important thing we have learned from Snowden, as it builds relationship maps and it doesn't rely on the target performing some specific action such as an HTTP requests that QUANTUM can race. Bonus: it only relies on "metadata".
  - pakled_engineer 12y ago
    The discussion about COTRAVELER was usually totally poisonous too https://firstlook.org/theintercept/2014/02/24/jtrig-manipulation/ https://firstlook.org/theintercept/2014/02/24/jtrig-manipula...
- LLWM 12y ago
  That's a bit patronizing. Of course it is true that most people who support surveillance don't understand what's happening, but only because most people in general don't understand what's happening. Plenty of us understand what's happening and also support it - anecdotally, at least. Fully understanding is asking for too much, though. I doubt anyone outside of a select few at the NSA fully understand what's happening, and it seems reasonable to guess that they support it.
rhino369 12y ago
Are you asking why America (along with most countries) spy? Or why they try to be good at spying? America spies because it wants information it can't otherwise have. What are Russia's plans in Ukraine? Is Pakistan playing both sides re: Taliban? Can China shoot down a B2 bomber? We want to know and nobody and asking won't help. Why we allow them to be better at their art is obvious.
- smokeyj 12y ago
  This is like breaking into your neighbors house to see if they have any plans to break into your house. How tragically ironic.

dtech 12y ago

TLDR: Send spoofed TCP package with HTTP redirect to malicious site right after victim does a GET request. This attack, while impressive and nifty, would seem to be easily subverted by SSL right?* Yet another example that shows that non-SSL HTTP should be phased out. * In the ideal situation where users check URL's and the CA system is good enough to keep the NSA out.

ilaksh 12y ago
Why shouldn't we assume the NSA has subverted the CA system? Why would we all rely on central authority for security if not to support centralized surveillance?
- Dylan16807 12y ago
  Because that subversion can be relatively easily detected, and then shouted to the world.
  - motoboi 12y ago
    Given the episode where stolen Realtek certificates been used for years by a government-made malware, I think we should not be so sure that they can't do that. A single request going through a bogus certificate emitted by NSA "acquired" CA would be hardly detected. Or, given the fact that they could indeed broke keys, they could simply do a man-in-the-middle, instead of a man-in-the-side.
    - Dylan16807 12y ago
      A stolen key is a server-being-hacked problem, not a CA problem. If the server is hacked, game over. There are browser extensions that detect bogus certificates; there's a good risk of catching them.
      - SG- 12y ago
        Should we pretend the NSA and various organizations don't have the means to infiltrate a CA?

tux 12y ago

I think this type of attack will be useless if you use DNSCrypt + VPN (encrypted OpenVPN/SSH traffic) not to mention TOR. Also, there is a new type of VPN @ https://www.goldenfrog.com/vyprvpn/chameleon https://www.goldenfrog.com/vyprvpn/chameleon

SG- 12y ago
Tunneling is good if they are targeting you locally, but if they see tunnel data going to a server which is easy to see they'll simply target you at the other end.
- tux 12y ago
  Good luck with that when there is thousands of people using the same VPN service, also randomization of IP addresses. However this is only if Evercookie are dealt with. @ http://en.wikipedia.org/wiki/Evercookie http://en.wikipedia.org/wiki/Evercookie Blocking Evercookie blocking is relatively easy now, if you disable JavaScript and use things like NoScript and Random Agent Spoofer. But then also don't forget about the recent VPN's ip leak through WebRTC which can be disabled with setting media.peerconnection.enabled to false If you're paranoid you can also use Qubes or Tails OS.
  - SCHiM 12y ago
    Hmm I'm not sure if a vpn will protect you from this attack. Your traffic will need to cross out of the vpn somewhere/sometime. Unless you're connecting to a hidden service or something. It actually depends more on whether you use https on your first connection or not. Because even if the server is set up to only accept https and it tries to redirect you at the first opportunity you'll be vulnerable. Generally a server configured to use https only will still listen on port 80, and send a 302 to redirect you to it's https version. That redirection can be hijacked by QI.
    - tux 12y ago
      Well if that true then everyone has much bigger problem. Because then any VPN services becomes useless. It would be nice to see if Fox-IT could do similar tests by using VPN/TOR. But I think that VyperVPN still is the most secure VPN if it really does have this; "Chameleon scrambles OpenVPN packet metadata to ensure it’s not recognizable via deep packet inspection (DPI)"

yAnonymous 12y ago

Aren't there firewall rules to prevent IP spoofing in most civilized countries?

uulbiy 12y ago
IP spoofing can be prevented by ISPs[1] but I don't thing there is an incentive for them to do it. That could not stop the NSA anyway. [1]: http://security.stackexchange.com/questions/1062/why-dont-isps-filter-on-source-address-to-prevent-spoofing http://security.stackexchange.com/questions/1062/why-dont-is...
- SCHiM 12y ago
  Also some (many?) ISPs have asymmetric routing, preventing an easy implementation of anti-spoofing mechanisms.
zurn 12y ago
"Great Firewall of Belgium only allows ingress traffic from non-Belgian source addresses" wouldn't prevent IP spoofing inside or outside Belgium. Nor would doing it the other way around. If such national firewalls existed in all countries and forced all IP traffic to flow through them, it might help some.
tptacek 12y ago
No.

lawnchair_larry 12y ago

I'm not sure why the comments here are focused on how the attack, as that is not the news and was well known for quite some time. This article is about how to thwart the attack. tl;dr Watch for the site to appear to respond twice with the same sequence number. One of those is the NSA trying to get in ahead of the real site.

pekk 12y ago
If the attack exists in the wild, its use isn't in any way limited to the NSA.
- tptacek 12y ago
  This attack has existed in the wild since the mid-1990s; it is literally the reason SSH exists. We didn't call our TCP hijacking tools "quantum inserts" back then, but that's exactly what they were. The only reason we didn't redirect HTTP connections to malicious sites is that in the 1990s, that would have been a waste of time; the r-commands were much lower hanging fruit.
  - tedunangst 12y ago
    TIL the NSA invented rlogin.
  - schoen 12y ago
    An important example from that timeframe publicly documenting the risk you're referring to: https://www.usenix.org/legacy/publications/library/proceedings/security95/full_papers/joncheray.pdf https://www.usenix.org/legacy/publications/library/proceedin...
    - tptacek 12y ago
      In fact, Joncheray "invented" this attack.
- csears 12y ago
  True, but state actors are much more likely to have ISP/backbone fiber taps and DPI equipment in place to make the attack really effective at scale.

4096 12y ago

The blog post with all the technical details: http://blog.fox-it.com/2015/04/20/deep-dive-into-quantum-insert/ http://blog.fox-it.com/2015/04/20/deep-dive-into-quantum-ins... And the Github account with the tools to do this yourself, PCAP and rules for snort, bro and suricata: https://github.com/fox-it/quantuminsert https://github.com/fox-it/quantuminsert

smellf 12y ago
Thanks for the technical link. You gotta love Wired, paraphrasing down to get gems like: “The first TCP packet will be the ‘inserted’ one while the other is from the real server, but will be ignored by the [browser]”. Oh my browser deals with TCP packets??

nayuki 12y ago

I found the title to be misleading. I thought it referred to the use of quantum computing to break public key cryptography and inject data. But instead, it's just some dumb TCP man-in-the-middle spoofing, only a bit more sophisticated than Comcast's RST attacks.

schoen 12y ago
QUANTUM is NSA's own name for its man-on-the-side attacks, so the infelicitous naming started elsewhere, not with the author of this article.

eyeareque 12y ago

Are there any details on when Quantum Insert started to be used by the gov? I wonder if the government built it off of the idea shared here: http://www.willhackforsushi.com/presentations/PenTest_PerfectStorm_Part_1.pdf http://www.willhackforsushi.com/presentations/PenTest_Perfec... (See slide 11-14) These slides describe the wifi tool called AirCSRF, which does basically the same attack, but over open wifi.

terminado 12y ago

Hmmm, what a short article. All it says is: It's actually impossible to detect these kinds of attacks. Please go about your business.

un1xl0ser 12y ago
I completely agree. While theoretically possible to do given full packet captures, it does not scale. IDS/IPS systems are not in a place to operate at this level and keep this much state. It's more likely that it would be detected based on the nature of the injected malicious data.