4 ms·

Once upon a time movies were too big to download, and now look.

by Canada 9y ago

aesclepius 9y ago
When was that, 1999? That's nearly 20 years ago.
cm2187 9y ago
Bandwidth has been increasing very slowly. And I think we will be stuck at 1gbit for a long time (and in many areas: if we ever reach it), just because there are no consumer use for higher bandwidth. Already 4k video resolution is a stretch, most people wouldn’t notice the difference to 1080p on a TV from they couch. So there will be a need to upgrade wholesale bandwidth just because the internet keeps growing, people are already talking about next gen game consoles to be cloud based (ie the rendering to be made remotely) which will add more traffic. But not 5 orders of magnitude. The amount of data stored by the gmails and facebooks of this world are mind blowing.
- Canada 9y ago
  Who says attackers must use home or small office connections? Why can't the data be exfiltrated to a top tier data center with excellent peering?
  - cm2187 9y ago
    But consumer/corporate demand for bandwidth is what should ultimately drive increase in data center bandwidth. Another point is defeating monitoring. I am sure the NSA (or Google/Facebook) could not notice 1GB of upload, but I like to think that uploading 1PB of data would make all sorts of red lights flash in they network security control room.
    - Canada 9y ago
      I'm with you that copying all of Google's data is unlikely. It's a serious project for Google itself to significantly move around its own data internally. My point is that very extensive, damaging information could amount to a mere 100TB subset of it and it's not implausible that could be copied in a day at 10 gig/s. To obvious? How about 100 hosts each pulling 100 meg/s? That's feasible right now. When you really get down to it the datasets I fear being leaked the most are a lot smaller than that. Most bandwidth is used sending many copies of the same content. Attackers aren't going to be interested in downloading the popular video 100 million times, they're just going to grab the logs which are nowhere near that size, and although large it's not implausible that even the best security teams wouldn't notice until it's too late. There is no hard rule that the leaks need to come from the same central database either. That is unlikely considering the fact that large scale services are already, and necessarily, distributed. Imagine thousands of attacker hosts receiving from thousands of compromised hosts.

Once upon a time movies were too big to download, and now look.

by Canada 9y ago

Once upon a time movies were too big to download, and now look.

aesclepius 9y ago

When was that, 1999? That's nearly 20 years ago.

cm2187 9y ago

Bandwidth has been increasing very slowly. And I think we will be stuck at 1gbit for a long time (and in many areas: if we ever reach it), just because there are no consumer use for higher bandwidth. Already 4k video resolution is a stretch, most people wouldn’t notice the difference to 1080p on a TV from they couch. So there will be a need to upgrade wholesale bandwidth just because the internet keeps growing, people are already talking about next gen game consoles to be cloud based (ie the rendering to be made remotely) which will add more traffic. But not 5 orders of magnitude. The amount of data stored by the gmails and facebooks of this world are mind blowing.

Canada 9y ago
Who says attackers must use home or small office connections? Why can't the data be exfiltrated to a top tier data center with excellent peering?
- cm2187 9y ago
  But consumer/corporate demand for bandwidth is what should ultimately drive increase in data center bandwidth. Another point is defeating monitoring. I am sure the NSA (or Google/Facebook) could not notice 1GB of upload, but I like to think that uploading 1PB of data would make all sorts of red lights flash in they network security control room.
  - Canada 9y ago
    I'm with you that copying all of Google's data is unlikely. It's a serious project for Google itself to significantly move around its own data internally. My point is that very extensive, damaging information could amount to a mere 100TB subset of it and it's not implausible that could be copied in a day at 10 gig/s. To obvious? How about 100 hosts each pulling 100 meg/s? That's feasible right now. When you really get down to it the datasets I fear being leaked the most are a lot smaller than that. Most bandwidth is used sending many copies of the same content. Attackers aren't going to be interested in downloading the popular video 100 million times, they're just going to grab the logs which are nowhere near that size, and although large it's not implausible that even the best security teams wouldn't notice until it's too late. There is no hard rule that the leaks need to come from the same central database either. That is unlikely considering the fact that large scale services are already, and necessarily, distributed. Imagine thousands of attacker hosts receiving from thousands of compromised hosts.