Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Someone told me that deep packet inspection works on the content of the video being said - aka if i am watching a video they can get the content/text and filter/censor it based on that.

They also told me that it doesn't happen in real time yet

So i was wondering can anyone confirm/deny the fact or provide with links on how it works.

Thank you

I don't really know if this question belongs to this site - but i thought this is a place where i can find my answer consider the type of people who view this :)

share|improve this question

3 Answers 3

up vote 2 down vote accepted

Yes, possibly. Problem is - IT NEEDS A LOT MORE PROCESSING POWER.

This is like saying "can SSL be cracked" (yes, in theory) and then saying "can a spy tool watch all SSL connections running out of a country in real time (no, unless the country spends thousands of billions on computers just for that).

So, deep packet inspection CAN work on video, but for a censoring tool it is simply not feasible at the moment. Too unreliable and will require way too much processing power.

share|improve this answer
    
I think the OP is looking to find out if the stream can be automatically assessed and blocked because they're watching, say, a subversive video and end up in jail for it. It sounds like this isn't anything involving deep packet inspection so much as just recording a packet stream and later analyzing it, which is childs play for sniffers. But automatically done by a filter using deep packet inspection? No. –  Bart Silverstrim Sep 25 '10 at 11:58
    
How you get the idea? I have strong hints of REAL TIME in the original question. –  TomTom Sep 25 '10 at 14:33
1  
i am looking to find out if censorship tools can analyze what is being said in a video - like someone saying "screw dictators" , will the censorship tool know that the person said this and block it ? –  mireille raad Sep 25 '10 at 17:07
1  
Are you joking? I mean, lagnauge recognition is hard without the backround noise and context analysis. Just impossible in a real movie with music and funny pronouciation. –  TomTom Sep 25 '10 at 18:54
1  
Thas is why i am asking - when they told me it is possible to do - i was like can't be ! so i wanted to double check the facts to be able to follow up –  mireille raad Sep 26 '10 at 14:52

Are you asking if someone knows you're watching a particular kind of video? Any network traffic that's not encrypted can be intercepted and monitored, and/or the source can be easily tracked (it's not hard to figure out what someone on the network is watching if there's a gig of data streaming from an IP belonging to hustler.com). But the firewall doesn't know what is in the video. If this level of sophistication were available then we'd have far better content filters in schools.

TL:DR-no, routers and firewalls aren't this sophisticated, or we'd have far better filters. But your activity can be monitored if someone were interested.

share|improve this answer
    
We were discussing censorship tools - usually those tools read the text- search for keywords and decide to block or not - I am wondering if the same can happen but not on text content but on the "what is being said in a video" –  mireille raad Sep 25 '10 at 11:20
    
It's not encoded as text in the stream, it's a graphic within the video. The packet inspection would need to be able to decode the image, read the font, take into account variances for the appearance of letters, etc. To put some perspective on this, if this type of thing were readily feasible, you wouldn't have spammers who farm out captchas to groups of people to decode. –  Bart Silverstrim Sep 25 '10 at 11:54
    
If you could readily automatically censor based on videos, you wouldn't have the failure rate (and amount of human intervention) that is necessary for content filters used in schools. It's just not there. But you can filter and get content from other things like origin of IP (if it comes from a particular site, chances are...) and amount of data (we know it's video, just not what it is). –  Bart Silverstrim Sep 25 '10 at 11:56

With enough computing power it is possible to do litteraly anything with deep packet inspection. For example see ever last image, video, or text sent across the wire aslong as it is "in the clear" meaning not SSL or other strong crypto. However doing this in near realtime requires an insain amount of computing power far more then a decent size server can deliver, something along the lines of a HPC GPU cluster. The cost of doing this would be in the mid 6 figures GBP.

share|improve this answer
    
You're thinking narrow, you're imagining this being done at a central location. Delegate the task to the client and processing power is a non-issue. Even from a centralized location, targeted analysis handles conservation of resources and again makes it feasible. Also, SSL isn't an issue either under the right circumstances. A custom CA that the client has been made to trust could easily make DPI on SSL packets possible. –  Technik Empire Oct 16 at 10:59

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.