回复:RE: RE: [itu-sg16] Q22 Question - SATC
Paul, you are a smart hacker, DPI can not stop you,:-). DPI hardly can tackle encrypted communication now, but this do not prevent DPI from getting more and more market deployment. Maybe someday,most internet communications will be encrypted to circumvent such kind of inspection, then I think new techonolgies will be devised to face these embarrassment. Anyway, I think content protection like copyright protection is a interesting area that worth doing something. ****************************************************************************************** This email and its attachments contain confidential information from HUAWEI, which is intended only for the person or entity whose address is listed above. Any use of the information contained here in any way (including, but not limited to, total or partial disclosure, reproduction, or dissemination) by persons other than the intended recipient(s) is prohibited. If you receive this email in error, please notify the sender by phone or email immediately and delete it! ***************************************************************************************** ----- 原邮件 ----- 发件人: "Paul E. Jones" <paulej@packetizer.com> 日期: 星期五, 九月 26日, 2008 下午2:06 主题: RE: RE: [itu-sg16] Q22 Question - SATC 收件人: 'zourong 52447' <zou.rong@huawei.com> 抄送: ''Noah Luo ' ("罗忠")' <noah@huawei.com>
Rong,
You are right that current file sharing networks (mostly) do not use encryption. But, they would if necessary. In fact, I have a file sharingprogram on my machine that does offer encryption for file transfers. If such measures were employed to block traffic, I think all tools would simply encrypt contents by default. One could try to guess the contents of the encrypted file by looking at such things as the file length, etc., but nothing is perfect.
I can accept that no solution is perfect, but I honestly believe that this one just will not work in practice. I used to be one of the kids in high school that would remove copy protection from software products by manuallyaltering machine code right on the disk. I even wrote my own utilities to allow me to modify code on disk -- and I wrote that in assembly language. Lawful intercept works for a few reasons: 1) Most criminals are stupid and use public communication 2) Almost everybody connects to a carrier network to carry voice -- so carriers become a natural intercept point
Lawful intercept does not work for smart criminals. They use secure communication channels.
But, we can't really compare this to Lawful Intercept. Normal users just use the tools that are out on the Internet. If I were writing a file sharing program and encountered packet filtering issues, I would immediatelyemploy encryption. I could write the necessary code in an afternoon and effectively defeat systems employed to stop it. And, those systems are all very expensive hardware-based solutions that would be rendered useless. Honestly, I'm not trying to be an irritant :-) But, I'm quite doubtful that any DPI scheme can work to stop the proliferation of copyrighted material.It's simply too easy to devise a work-around in software and, while the DPI logic can get more complicated, the algorithms in a PC-based application can be revised and updated far more rapidly. Heck, I would even have the packet encoding algorithms written in such a way as to be downloadable when the program starts. That way, the newest encoding algorithms are always available every time the program starts.
Paul
-----Original Message----- From: zourong 52447 [mailto:zou.rong@huawei.com] Sent: Thursday, September 25, 2008 11:48 PM To: Paul E. Jones Cc: 'Noah Luo ' ("罗忠") Subject: 回复:RE: [itu-sg16] Q22 Question - SATC
Paul，
Sorry for taking so long to respond to your mail because I have taken a long vacation and then followed by a busy business trip.
I agree with you that the method proposed in my contribution AVD 3541> is not a perfect solution to the copy-right protection issue. It could not handle the case that the communication is encrypted ?C as you mentioned. In fact, I think it is one of the limitations of DPI technology. The signature-based DPI technology could not inspect the encrypted data flow. But I still believe that it could work and do something useful in a broad range of use cases, just like the lawful interception: it hardly could handle the encrypted communication but it still can work> well.
As to your suspicion whether it could work in a practical situation.> I hope it could work in some public file sharing scenarios especially the P2P file sharing. Why I put my effort in study the P2P file sharing> scenario is that P2P file sharing is the most popular file sharing tools nowadays and most of the files being shared are multimedia files> and many of them are pirated.
P2P file sharing has one important character that could be exploited:> the file information is almost public. It should let people know what the resource is, for example, the file name, file type, file size, file hash value etc., these file describe information will be transmitted in the file sharing control messages and could be intercepted by SATC.
Another character of P2P file sharing is that it seldom uses encrypted> communication because that will bring burden to its efficiency and affect its popularity.
Then to the question of “good database” and “suspicious database”. What I want to do is a multi-level inspection and filter. It maybe works as following:
Extracting_file_information(); Searching_good_database(); If(got matched item) { Compare_signature(); If(match) { It is copy-righted file and transmitted; } Else { Searching_suspecious_database(); If(match) { It is a known pirated version, stop it; } Else { It maybe a new pirated version, record it temporality and notify the content provider to make sure and to update the suspicious database ; Transmit_flow(); } } }
In this multi-level filtering, the first level ”good database” filtering can be performed with the help of a reasonably small- sized> database and achieve rather fast processing speed, the second level ”suspicious database” filtering may require a much bigger database. But only searching only a part of them each time also could yield reasonable processing speed.
So that is what I want to do, only those truly pirated file sharing will be stopped right now, other suspected file sharing will be recorded and let the technicians to do further judgment.
This email and its attachments contain confidential information from> HUAWEI, which is intended only for the person or entity whose address is listed above. Any use of the information contained here in any way (including, but not limited to, total or partial disclosure, reproduction, or dissemination) by persons other than the intended recipient(s) is prohibited. If you receive this email in error,
***********************************************************************> ******************* please> notify the sender by phone or email
immediately and delete it!
***********************************************************************> ******************
----- 原邮件 ----- 发件人: "Paul E. Jones" <paulej@packetizer.com> 日期: 星期五, 九月 12日, 2008 上午11:50 主题: RE: [itu-sg16] Q22 Question - SATC 收件人: 'Noah Luo(罗忠)' <noah@huawei.com> 抄送: zou.rong@huawei.com
Noah,
I have no objection to exploring the idea. But, I'm 100% certain it will not work :-) I am various serious about that. When I was a teenager, I spent a lot of time removing copy protection from software
(actually> > modifying the machine code by hand), hacking computer systems,> > etc. Even
when I was in college, I had a professor who called me an unscrupulous hacker. He was only half joking. He had a real admiration for my technicalability, but wasn't very happy with some of the hacking I used to do. Still, he recognized that I was a good kid who just wanting to explore and learn as much as possible.
I used to play with viruses. I would even write some that would not cause harm -- they would never replicate. Rather, they would just run in the background of machines that I placed them on, loading into the system at boot time. I still have a collection of probably 1,000 virus files, many with the assembly code.
The short of it is that I have a lot of experience at this sort of thing.While I don't do the hacking I did when I was a kid, I can guarantee you that there will be some just ready to tackle the challenge of breaking any kind of defense like what this standard intends to provide. And, as I said, the simplest thing to do is simply use TLS. In fact, I could exchange pirated music or video via e-mail between my server at home and anybody else who uses TLS -- my servers all use encryption. But, we can certainly employ simpler techniques. For example, I could do this:
K = 256_bit_random_key; while(not_end_of_file) { M = read_file(16 octets of the file); Y = M XOR K; transmit_message(Y); K = rotate_right(K); } transmit_message(K);
This is a very simple and extremely fast encryption algorithm that willactually transmit the session key at the end. So, there is no secrethidden, but it avoids the need of exchanging encryption keys some other way. This is not a flawless procedure, though. In fact, this is very easy to "crack", since we're just XORing the bits of the input stream with the key and then rotating the key 1 bit after each message. If an input stream had a number of 0s, then the key would appear as repeated data blocks that can be easily identified by a cryptography person. Heck, even I would recognizethe pattern. But, this might still be complex enough to avoid real-time detection of pirated content. But, I could just as easily use AES and *really* encrypt the media and place the 256-bit initialization vector and key at the end of the transmitted media stream :-)
Anyway, like I said-- I have no objections to studying this. I just am just a firm believer that it will only stop the casual pirates: those who want to use gmail.com or something to send an e-mail containing music. It would not stop software designed for secure file exchange, either person- to- person or across file sharing networks.
My primary concern now is with the wording in the report. Rong felt it was OK, but I still think it's wrong. It says: "... if SATC finds a data stream carrying content failing to match a known signature, it will block the stream or mark it as being suspicious.”> > Worded another way, this sentence says: "if the data stream contains an unknown data stream, block the data stream."
Now that would actually work to block pirated content, because you essentially only let known data through the network. But, I don't thinkthat is the intent, and it's definite not what the AVD document said.
Paul
-----Original Message----- From: Noah Luo(罗忠) [mailto:noah@huawei.com] Sent: Thursday, September 11, 2008 5:46 AM To: 'Paul E. Jones' Cc: zou.rong@huawei.com Subject: 答复: [itu-sg16] Q22 Question - SATC
Dear Paul Thank u very much for your good analysis and insightful comments on this specific aspect SATC.I will work together with Rong and come up with a solution as to how the Q22 meeting report will be appropriatedly reworded.Based on C3541 itself,actually, Q22 meeting report just faithfully reflects the basic idea of the scheme described.But if we also take into account the supplementary information Rong provided in his mail to you, then Q22 report may need incorporate some more information.
As for your questions regarding the feasibility of using signature> > > matching to block pirated contents transmission.I agree that we need more careful consideration.I think it will be ideal if more contributions will be submitted by Rong to the next SG16 meeting so that we can have more extensive discussion.My personal feeling is that it is a rather innovative idea to use this kind of techniques to block illegally distributed> > > contents, but there may still lack some feasible technical points to enable this idea to be transformed into a practical solution.
Paul, do you think this arrangement will be okay?
Best Regards
Noah
华为技术有限公司 huawei_logo
地址：深圳市龙岗区坂田华为基地 邮编：518129 http://www.huawei.com -------------------------------------------------------------
----- --------------------------------------------------------- 本邮件及其附件含有华为公司的保密信息，仅限于发送给上面地址中列出的个人 或群 组。禁 止任何其他人以任何形式使用（包括但不限于全部或部分地泄露、复制、或散 发）本 邮 件中 的信息。如果您错收了本邮件，请您立即电话或邮件通知发件人并删除本邮件！ This e-mail and its attachments contain confidential information from> HUAWEI, which is intended only for the person or entity whose address is
use of the information contained herein in any way (including, but not
total or partial disclosure, reproduction, or dissemination) by persons other
intended recipient(s) is prohibited. If you receive this e-mail in error,
phone or email immediately and delete it!
-----邮件原件----- 发件人: itu-sg16-bounces@lists.packetizer.com [mailto:itu-sg16-bounces@lists.packetizer.com] 代表 Paul E. Jones 发送时间: 2008年9月11日 10:48 收件人: 'zourong 52447' 抄送: itu-sg16@lists.packetizer.com 主题: Re: [itu-sg16] Q22 Question - SATC
Rong,
I think the meeting report is not exactly correct. The Q22 meeting> > > report said that SATC would block content that that failed to match a signature. What you said is that SATC would block media content that matches a "good" signature, but is apparently altered in some way. That is, it would block media that it can clearly identify as pirated content.
So, perhaps we should revise the wording that is presently in
meeting report related to AVD-3541. Can you consult Noah on what the correct statement should be?
As for whether this will work, I still have doubts. Even storing> > > signatures of all current media could be a massive database and, quite
that would need to be in memory, as you need real-time (and likely wire- speed) access to the signatures.
I have no objection to doing work on this: if you can meet a
match against signatures will not stop any serious pirate. It might stop the casual pirate, though.
A data stream of pirated content can be altered in so many ways
would be easy for somebody to create a program to stream pirated content that does not match a signature in the database. I could run
stream through a cipher algorithm, for example. And, I could even transmit the cipher key as the last payload of the message: after all, the entire file has been delivered! Secrecy wasn't the purpose of the encryption,> but merely to disguise what is being transmitted until we've successfully> bypassed the SATC "detectors".
Perhaps the simplest thing to do, actually, given the wide proliferation of tools like OpenSSL, is to just use TLS between nodes that want to transmit pirated content. That provide more-than-acceptable level of encryption.
In any case, I did not intend to bring the work to a halt, by any means. But, be cognizant of the fact that this is a very complex
seriously doubt you can devise a method that would prevent a good> > > hacker from getting around the system. At best, you will only stop the casual> "pirates". You would not stop serious pirates or popular "file> sharing" software. Such software would definitely employ techniques to get around any detection/blocking logic in the router/switch.
What's more important right now is understanding what should have been stated in the meeting report. I do not think this sentence is accurate:
“... if SATC finds a data stream carrying content failing to match a known signature, it will block the stream or mark it as being suspicious.”
Paul
-----Original Message----- From: zourong 52447 [mailto:zou.rong@huawei.com] Sent: Wednesday, September 10, 2008 4:05 AM To: paulej@packetizer.com Cc: itu-sg16@lists.packetizer.com Subject: [itu-sg16] Q22 Question - SATC
Paul, really thanks your's comments, here is the clarification. 1) As to the circumventing problem, I really could not say
a perfect way that could not be circumvented. But I don’t
easy to do so. This scenario is attempted to provide a tool for content provider to prevent their “specially processed” content, i.e. copy- righted content from being pirated. It is assumed the “specially> > > > processed” content have some characteristic information that would be destroyed by any piratical actio n such as using a different decoding method, and these characteristic> > information or “signature” would be provided to SATC system by content provider. As to the compress scheme to circumvent, I think the SATC> > could handle most of the common compress schemes such as winzar,> > winzip, tar etc., but I also think it is impossible the SATC could handle all compress schemes. 2) The method mentioned in AVD-3541 is not attempted to create a database of all pirated media, it’s a database that store the signature of good media especially those most popular video or audio> > recently, so the database will keep in a state of acceptable size. The periodical updating of the database is necessary just like the IDS do to update the anti-virus database. 3) The pirated media database could exist as an subsidiary database.> > This was not mentioned in the AVD-3541 because I
detail to describe it in a scenario. The file transmitted will be suspected as pirated file when satisfied following two points: (1) There is a item in the “good media database” that means the file is a target of protecting. (2)the signature of the file transmitted do not match with that stored in the data base. Then the suspected file could be blocked directly, or in a more prudent way, it could be just marked as suspicious and search in a subsidiary “pirated media database” to make sure it is a existing> > > > pirated version or it maybe a new pirated version and need more manual analytical works of content provider. 4) As to the performance problem, I think this kind of inspection may have some negative impact on the packet transmission but may not as terrible as thought, in fact, many DPI product in market can achieve> > 10G to 40G, even 80G throughput, their process capability is remarkable.
发件人： Paul E. Jones 发送时间： 2008-09-05 10:59:38 收件人： itu-sg16@lists.packetizer.com 抄送： 主题： [itu-sg16] Q22 Question - SATC
Q22 Experts,
While reviewing the Q22 meeting report, I was intrigued
by a
comment related to AVD-3541. This contribution proposes
can be used to block illegal content from being transmitted over the Internet by examining the media flow and comparing that media> > > flow (or some signature thereof) against a database. SATC would then block content as appropriate.
I would like to comment: 1) I could easily circumvent any such measure, so any such system would prove to work only temporarily; and 2) While a given copy of a digital copy of some content, such as music or a movie, will have a clearly recognizable signature, the same media can be “substantially” altered by merely using a different encoding method (e.g., a different video codec or compression scheme); and 3) Actually attempting to create a database of all known> > > > > pirated media, including all variations, would result in the creation of a massively huge database; and 4) Maintenance of such a database would be a never- ending> > > > > chore; and 5) Notwithstanding the foregoing, while such a database could> > > be constructed and flows could be examined, inspecting flows in real-time and trying to positively identify a given flow is a monumental task that will result in significantly negative impact> > > on packet transmission performance, even if a bank of specialized> > > network processors were employed per router or switch
In short: do we really want to attempt this? While it’s all technically possible (at the risk of turning a 1Gbps pipe into a 1Kbps pipe), I would argue that it’s not going to work. I would be happy to be first in line to write the software necessary to circumvent such a system. And I would do it just because I can ;-)
Having said that, the meeting report said that, the meeting report> > > said that “if SATC finds a data stream carrying content failing to match a known signature, it will block the stream or mark it as being suspicious.”
I underlined the text of significance. This suggests
------ listed> > > above. Any limited to, than the please> notify the sender by the Q22 likely, one particular> business requirement, great. But, I personally think that trying to that it the media problem and I their is think it is think it maybe too that SATC that the
database would only be populated with signatures of known good media flows and would only block unrecognized flows. This seems different than what I read and understood from AVD-3541. What the meeting report suggests would be a simpler problem with a substantially smaller database. So, what was actually discussed and decided for SATC? A system that blocks unrecognized flows or a system that blocks recognized flows?
Thanks, Paul
participants (1)
-
zourong 52447