Deep packet inspection could be outlawed in US | Hacker News Reader

Deep packet inspection could be outlawed in US

techworld.com

5 pointsvaksel9 comments

Totally misleading headline. DPI-based behavioral advertising and customer tracking might (might!) become unlawful. But even the people advocating for DPI regulation concede that it will be used for "security" and "network management", which means that ISPs will still be able to use it for:

* Blocking or throttling BitTorrent and P2P

* Satisfying requests from the RIAA and MPAA, and

* So-called "lawful intercept".

Note also that when the government "wiretapped" the Internet, they didn't do it by deploying Narus directly ISP networks. Instead, they offramped traffic to their own lab network. There's no reason you couldn't build a shrink-wrap commercial offering based on the same idea; ISPs already offramp internally to "scrubbing networks" when handling DDoS and botnet traffic.

What is the problem with DPI? It slows down torrent downloads, which are already slow, but speeds up more synchronous activity like web browsing and chatting. Don't we all want that?

DPI does not specifically slow down bittorrent traffic or speed up http traffic. DPI filters traffic based on the data in the packet, which affects all traffic. What happens to that traffic is up to the person implementing the DPI.

The problem with DPI is that its akin to the post office reading all of your mail.

The post office has a similar system, except they charge customers differently to get information to the recipient faster. Overnight costs more than first class for example.

If DPI were used to allow customers to pay based on priority, would that be acceptable? For example, if you only want to pay 1 cent per GB for a torrent download that gets packets to you at 1/3 the rate of HTTP packets, which cost 2 cents per GB, would something like that work?

I am not proposing it, just continuing the post office analogy.

I've never understood this analogy. We already pay more for more bandwidth which is supposed to allow us to receive packets faster. This always sounds like they're expecting me to pay twice for service.

Those protocol discrimination and QoS concerns are possible results of "deep packet inspection" (how deep?) but just because these results are good does not mean we should accept DPI (~converse fallacy).

The article outlines some of the problems with DPI but does not make a compelling case. See http://en.wikipedia.org/wiki/Deep_packet_inspection#Deep_Pac... for some uses of DPI at the ISP level. Each of these use cases has negative features. Those are "the problem with DPI".

In terms of a long-term solution to the DNS cache poisoning issue it seems deep packet inspection could be the only way to go. The work-around was to extend the encryption involved with DNS but that only makes it harder, but not impossible for DNS hacks to take place. Only if deep packet inspection happens would DNS hacks truly become a thing of the past, since if my ISP is inspecting all the packets, and it's stored IP for say cnn.com gets re-routed to a hacker's IP, an inspection of the packets coming across could flag a problem to the ISP. I am not an expert in this area but I just recall from reading a 'Wired' article on this issue that the only really good solution here is in fact deep packet inspection. So outlawing it would be a hacker's wet dream, and by hacker I mean "cracker" or "black hat" people obviously.

This is completely nonsense.

The current workaround to the DNS spoofing hacks has nothing to do with encryption.

The solution to DNS spoofing has nothing to do with packet inspection (Google for DNSSEC and DNSCurve for details).

(1) No part of the DNS is currently any hacker's wet dream. Why bother spoofing at all? You don't need to bust out the batmobile to get into most people's accounts.

(2) The technique that "extended" DNS' security is the same technique that protects plenty of other core Internet protocols, so if it doesn't work at least in the medium term, we're all doomed.

(3) The issue isn't whether DNS is breakable --- it always has been --- it' s how easy it is. The problem is about cost, not about raw capability. If you can't pull the attack off drive-by, it doesn't make a difference in 2009.

(4) Anybody who tries to sell you on a DPI solution to DNS security is scamming you. You'd need a globally deployed network of DPI boxes, all synchronized, to make a dent in the problem.

Reply on news.ycombinator.com