Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?
Security Encryption

Hiding Packets in VoIP Chat 90

holy_calamity writes "Two Polish researchers say they have developed a system to hide secret steganographic messages in the packets of a VOIP connection. It exploits the fact that VoIP uses UDP, not TCP; it is designed to tolerate some packets going missing -- so hijacking a few to transmit a hidden message is not a problem." You may also be interested in reading the original paper.
This discussion has been archived. No new comments can be posted.

Hiding Packets in VoIP Chat

Comments Filter:
  • Too late (Score:5, Informative)

    by oodaloop ( 1229816 ) on Monday June 02, 2008 @12:42PM (#23628349)
    Didn't /. just post an article a few months ago about how the NSA figured out a way to block steganographic messages in VOIP?
    • Re:Too late (Score:5, Informative)

      by Zymergy ( 803632 ) * on Monday June 02, 2008 @12:46PM (#23628389)
      Sort of... "Blocking Steganosonic Data In Phone Calls" []
      There is this too... []
      • Wouldn't the same technique work for VOIP as well? And since when is referring to previous relevant /. articles being a troll?
        • Re:Too late (Score:5, Informative)

          by bickerdyke ( 670000 ) on Monday June 02, 2008 @01:19PM (#23628755)
          Only as long as you'd try to hide your secret data in the Audio stream. If you inject your secret data directly into the network "connection" (read: the sequence of UDP Packets sent) it bypasses manipulated background noise.
        • Re: (Score:3, Interesting)

          by redxxx ( 1194349 )
          The first link would not work because they can't just add noise. They would have to inspect and remove packets from the data stream. It works totally differently and would not be applicable.

          The second is just looking for out of band communication in data streams. It could be configured to look for it in Voip traffic, but most of it is encrypted. It wouldn't be easy, particularly doing it in something like real time, but not impossible.
        • Re:Too late (Score:5, Funny)

          by GuldKalle ( 1065310 ) on Monday June 02, 2008 @01:35PM (#23628941)

          And since when is referring to previous relevant /. articles being a troll?
          Probably just someone trying to post a steganographic message using the /. mod-system.
          • by StreetStealth ( 980200 ) on Monday June 02, 2008 @03:28PM (#23630199) Journal
            It does get one thinking, though... So many things on the internet appear to be governed purely by entropy; how many of them could conceivably be used for steganographic purposes?

            Imagine a series of /. accounts set up for bots to automatically comment on stories, with an algorithm somewhere to scrape and concatenate certain characters based on a key consisting of times and offsets...

            Come to think of it, there's no reason why this necessarily couldn't be the case with some of the vast volumes of blog comment spam out there. Spread out wide enough and with a resilient enough algorithm, there could be more than enough signal to cover for the noise of spam-killed comments...
  • Pay for 388 words? (Score:5, Insightful)

    by CogDissident ( 951207 ) on Monday June 02, 2008 @12:44PM (#23628363)
    To continue reading this article, subscribe to New Scientist. Get 4 issues of New Scientist magazine and instant access to all online content for only USD $5.95

    Thanks Slashdot, because I really want to go to Slashdot to get links to a story that I have to pay to read.
  • Complete article (Score:5, Informative)

    by TripMaster Monkey ( 862126 ) on Monday June 02, 2008 @12:45PM (#23628379)
    The complete article, accessible without NewScientist subscription, may be found here [].
  • Well... (Score:4, Insightful)

    by Vectronic ( 1221470 ) on Monday June 02, 2008 @12:45PM (#23628385)
    It's not a sectret anymore now is it?
    • Re:Well... (Score:5, Funny)

      by Vectronic ( 1221470 ) on Monday June 02, 2008 @12:46PM (#23628393)
      Nor a secret for that matter.
      • Re:Well... (Score:5, Funny)

        by fracai ( 796392 ) on Monday June 02, 2008 @01:03PM (#23628563)
        I assumed the misspelling was one part of a larger steganographic message. Let it be known that I am now browsing over your comment history looking for further "mistakes".

        I'm on to you.
        • Re: (Score:3, Insightful)

          by Vectronic ( 1221470 )
          You are abolutely right, however, you forgot that I may have multiple accounts, and may be sending messages across more than just Slashdot.

          You would have to know all my accounts, on all forums, plus know the method to decipher the data.

        • Re: (Score:1, Offtopic)

          Missing a comma after 'anymore'.
        • Re:Well... (Score:5, Interesting)

          by h4ck7h3p14n37 ( 926070 ) on Monday June 02, 2008 @03:13PM (#23630063) Homepage

          That reminds me of a neat story.

          A few years ago at a tech conference I met someone who worked for the data storage division at Dell. Some of the technical manuals that the engineer needed for their work were classified as secret (product hadn't gone to market yet) and the engineer had to sign various NDAs with the company to get access to the documents.

          Said engineer compared their copy of a manual with another engineer's copy and discovered that each manual had a different set of spelling errors. Apparently Dell was generating documents with unique sets of typos in order to be able to track down the identify of the person who leaked a document.

          • If I had mod points. (it wouldnt do any good since envoled in the conversation) +1 Interesting, anyways.

            Although Dell wasnt the first to do so, it is still generally a good idea, because serial numbers, and other tags, can easily be swapped/removed.

            But it's not perfect, given that if someone manually typed out the document, and removed all spelling mistakes, or even created new ones, the system fails, likewise, if someone was "in the know" about the scheme, they could essentially impersonate another (rival)
          • Re: (Score:2, Funny)

            by AioKits ( 1235070 )

            Apparently Dell was generating documents with unique sets of typos in order to be able to track down the identify of the person who leaked a document.
            Either that, or Dell has taken the 'million monkeys at a million keyboards' approach to producing technical manuals!
          • Dell learned something from having an ex-NSA Director and Deputy CIA Director on their Board of Directors.
          • Re: (Score:1, Insightful)

            by Anonymous Coward

            Said engineer compared their copy of a manual with another engineer's copy and discovered that each manual had a different set of spelling errors. Apparently Dell was generating documents with unique sets of typos in order to be able to track down the identify of the person who leaked a document.

            That's crude. There are other schemes that encode the identity of a document in the microspacing between the letters.

    • it Never was a secret. yOu could alSo hide messages In ceReal boxes and floral arrangements.
  • No way (Score:4, Funny)

    by William Robinson ( 875390 ) on Monday June 02, 2008 @12:47PM (#23628401)

    secret steganographic messages in the packets of a VOIP connection

    Stop this research. No way I am going to say GoodBye to my Secretary. She knows a lot more than just stenography;)

  • UDP Only... (Score:5, Interesting)

    by mchawi ( 468120 ) on Monday June 02, 2008 @12:52PM (#23628445)
    Based on the RFCs for VOIP they are supposed to support UDP and TCP per the new specs. Most companies are moving to support both so you can choose, but some of the large companies are going to TCP because this is what all of the 'Unified Communications' packages go with (such as Microsoft Office/Live/Communicator, etc).

    One of the reasons they are leaning this way is security. Go figure.

    Besides that, I don't really see the point. What does this solve that just encrypting sensitive data wouldn't?
    • Re:UDP Only... (Score:4, Insightful)

      by k_187 ( 61692 ) on Monday June 02, 2008 @12:58PM (#23628517) Journal
      If somebody's looking for something encrypted data is something. With this method, there isn't anything to find, unless I'm totally misunderstanding it.
      • Re: (Score:3, Insightful)

        by Vectronic ( 1221470 )
        Yeah thats what I got from it aswell.

        If someone is using an encrypted connection/transfer, then its obvious they are doing something, and also trying to keep it hidden, whereas, if they were to carry out a normal transmition, but have the "secret" part of it hidden in this, someone looking, would see a normal interaction and possibly skip over the noise.

        You could also have an encrypted message, that also requires data from the steganographic 'noise' and vice versa to become usable data, that way if one is "
      • Re: (Score:2, Informative)

        by papna ( 1242200 )

        With this method, there isn't anything to find, unless I'm totally misunderstanding it.
        Or rather, there's nothing to notice.

        Plain cryptography is something like having a locked safe sitting in a room. It might not be easy to get into, but you know it when you see it. This is like having a safe behind a painting. You don't notice that there is anything being kept away from you.
      • by Zadaz ( 950521 )
        And yet everyone says security by obscurity isn't security at all.
        • "And yet everyone says security by obscurity isn't security at all."

          Obscurity is not security, but can be complementary. In the VoIP example the security would be the encryption of your signal, the obscurity would be the addition of meaningful UDP packets.

          Obscurity is helpful when dealing with cursory inspections, but doesn't actually increase security because being secure requires more than being non-obvious.

          Think of contraband transportation. Driving around with illegal contraband in plain sight - say s
        • by Sancho ( 17056 ) *
          The term Security Through Obscurity is overused and poorly understood. The key is that most Security Through Obscurity has cryptography in plain sight with an "obscure" encryption mechanism. It's the "we created our own cryptography implementation, but we can't tell you what it is because it would compromise the security of the algorithm" that causes the problem. It's usually quite possible to reverse-engineer such algorithms, so if the system relies on secrecy which can be discovered (as opposed to the
    • Re:UDP Only... (Score:5, Interesting)

      by zappepcs ( 820751 ) on Monday June 02, 2008 @01:04PM (#23628573) Journal
      Well, it might ensure that the NSA et al are not infecting your VoIP equipment with tracing software while you are talking, and those pesky terrorists might not be able to send text data about the next planes to hijack while having a bad conversation quality exchange about prayer times and how to find Mecca while in Chicago.

      When a security hole is found, it needs to be plugged because the threats it poses are not always explicitly understood at first glance.

      In fact, in computing in general, there are multiple ways to sneak a couple of packets through here and there if you're willing to be patient. I'd mention a few of them, but that would probably get me on a fucked up watch list. The fact remains that this is but one way to do so. Monitoring the network packet for packet won't uncover them all either, nor will it out any terrorists who don't want anyone watching their communications. Why, even my music on hold can contain data for transmission to the right person with the right audio equipment. Never mind a blog post, or email. In fact... woooootttt! I could use the NSA's website as the key for an encryption routine that they would never decode in several decades of trying. sigh, but that won't stop them from telling us that it's all for our protection.

      Just encrypting it would not stop the possibility of rogue data if your application can withstand a few missing packets. VoIP is not the only protocol which is susceptible.
    • Re:UDP Only... (Score:5, Informative)

      by PhuCknuT ( 1703 ) on Monday June 02, 2008 @01:06PM (#23628599) Homepage
      The idea behind steganography is not just to encrypt the data, but to hide the fact that you're sending it in the first place.
      • by Anonymous Coward
        ...if you're using steganography in your VoIP data stream to imbed pr0n images, then you've invented a clever new form of digital phone sex, right?
      • by mpe ( 36238 )
        The idea behind steganography is not just to encrypt the data, but to hide the fact that you're sending it in the first place.

        Even though specific applications may use steganography in conjunction with encryption it does not imply that encryption is involved.
    • Re:UDP Only... (Score:5, Informative)

      by Kr3m3Puff ( 413047 ) <me.kitsonkelly@com> on Monday June 02, 2008 @01:09PM (#23628633) Homepage Journal
      First, Stenographic or Stenophonetic solutions are supposed to disguise that you are actually communicating encrypted information, which is 1/2 the battle. If you know two parties are transmitting encrypted information that is sometimes enough (especially in this day and age) to either attack via brute force, or even worse, make them legally hand over their decryption keys, where then you need plausible denability. When the third party doesn't even know you are transmitting information, you are in a much better situation.

      First, wide adoption of RTP transmission via TCP is highly unlikely, due to the nature of streaming media in general which UDP is designed for and TCP is not. Fixed datagrams and packet ordering protocol are a major pain in the a$$ for streaming media.

      Where as the call control protocol (SIP, H.323, MGCP, etc) via TCP is probablly more likely and most standards support transmission under either, though the vast majority is still UDP based.

      You are right from a security perspective with TCP you know if information is gone missing, where as UDP you never really know.
    • Re: (Score:1, Informative)

      by Anonymous Coward
      Besides that, I don't really see the point. What does this solve that just encrypting sensitive data wouldn't?
      A lot. Remember when W. told OBL that we were listening in on their sat phones? Well, between that incident and the time that reagan gave up info about the KAL incident told a lot about our intel world (the 2 should have been swung, or gone on a hunting trip with cheney, for those actions of being traitors; it took several years for pilots to talk again and a number of interesting channels were s
    • Geekier.
  • Make noises (Score:5, Funny)

    by tristian_was_here ( 865394 ) on Monday June 02, 2008 @01:05PM (#23628587)
    If you want to hide packets over VoIP I suggest making "beeping" noises.
  • authors (Score:1, Funny)

    by Anonymous Coward
    Article by Wojciech Mazurczyk and Krzysztof Szczypiorski... wow ... Did they encrypt and hide their original names ?
    • Re: (Score:3, Funny)

      by Anonymous Coward

      A Polish guy goes in for his yearly eye examination.
      The eye doctor says, "OK, read the smallest line down on the chart that you can."
      The guy reads out, "W... Z... P... X... Y... I... Z... Y... K...".
      The doctor says, "Wow, that's great, you can read the bottom line?"
      The Polish guy says, "Read it? Hell, I know the man!"
    • ATTACK!!! (Score:3, Funny)

      by sznupi ( 719324 )
      Say this:

      W Szczebrzeszynie chrzszcz brzmi w trzcinie.

      (note: your head may explode)

      (PS. and don't look at my nickname ;P )
      • /. comment system cut out one letter with, I'll just use closest thing from the roman alphabet:


        There, should be much easier to you ;)
  • by Kyont ( 145761 ) on Monday June 02, 2008 @01:13PM (#23628691)

    You may also be interested in reading the original paper.
    CmdrTaco, you must be new here.
  • as more and more companies move their voice system over to VOIP, this creates an interesting dilemma: how do you prevent information leaks from secure sites when your telephone system can act as the carrier? Which probably means that we'll have more company snooping around and more "by using this system you agree that your privacy will be raped daily" forms we all have to sign when we get hired.
    • Telephone systems have been possible carriers for far longer than digital telephony has been around. While analog phones do not operate well below 100 cycles, they carried enough information to incorporate inaudible data at well below 20 cycles and imperciptible to your casual listener. The quality isn't good, but it doesn't have to be to bury a trigger message.
  • I didn't even know we knew what a Stegosaurus sounded like, and these guys hid its messages in VoIP traffic!
  • While VoIP certainly can use UDP, it's also quite possible (and even common) for VoIP calls to use TCP as the transport. Hell, the original paper even mentions steganography over TCP.

    Saying "VOIP uses UDP, not TCP" is overly simplistic. RTP can run over either UDP or TCP, while SRTP runs over TLS-over-TCP.
    • Re: (Score:3, Interesting)

      by profplump ( 309017 )
      I know people are still confused by the magic of IPSec, but seriously, UDP over IPSec is a vastly superior way to secure RTP in any situation where packets might be dropped or re-ordered. SSL+TCP+RTP might work on a LAN with lots of bandwidth to spare, but it just doesn't work across the Internet.

      I used to have an IPSec bridge to the office, with RTP running over UDP on that bridge. Everything worked great. Now my company has turned off end-user IPSec, and requires use of the Cisco SSL/TCP-based VPN client.
  • by Alarash ( 746254 ) on Monday June 02, 2008 @02:29PM (#23629539)

    VoIP doesn't "use UDP instead of TCP". VoIP (which is usually SIP+RTP, but there are other protocols out there used to carry voice over IP networks) can use UDP over TCP, and that configuration is the most common one. But not the only one possible as the article suggests.

    Also, the article in the /. article kind of suggests that VoIP (which is a concept, not a protocol) can use only UDP, which is not true. It's like saying Internet is used only for HTTP.

  • Isn't VOIP illegal in most of the countries where data hiding needed to protect yourself from the political police?
    Telephone service is usually a government monopoly in the developing world. VOIP bypasses the government telecommunications monopoly. And since that monopoly is so profitable, the government authorities in these places violently suppress anyone that they catch using VOIP.

    What kind of information would be hidden in VOIP transmissions? General political tracts and religious boo
    • "We received your PayPal transaction, Thank you very much. Eight grams of dynamite skunk weed for you is located in a crushed Mountain Dew can in the gutter exactly sixteen feet south east of the bus stop sign at the corner of First and Main. We will pick it up if you don't do so by 3:30pm Tuesday"? I've always wondered why simple dope dealers don't use Internet technology for anonymous untraceable transactions? Could it be because most dope dealers are stupid, or just old-fashioned?

      PayPal is anonymous??

    • Re: (Score:3, Interesting)

      You can purchase a Vonage/etc. adapter in the USA, and then plug it in anywhere in the world. This works in a lot of places that VOIP is officially "not available" - exactly where depends on the settings of that country's firewall.
  • by suitepotato ( 863945 ) on Monday June 02, 2008 @03:47PM (#23630451)
    I think the future will see the use of trojan/virus techniques to send data. It's already been fairly well proven that stopping botnets is next to impossible given current technologies, attitudes and ideas on the part of administrators and engineers, and most importantly that AI bears not a candle compared to Natural Stupidity.

    Forget just VoIP. In the future we'll hide communications networks under multiple layers of encryption inside trojan'd everything that is awfully hard to tell innocent user data from something else. We'll probably also host websites and files that way in a coalescence and then expansion of BT/P2P and anonymous remailer methods but not so much with identifiable clients but instead viral ware that people choose to allow on their machines so as to prevent privacy invasion by government and business.
  • It must be cuz it sure as shit don't work right. I could reach more people sticking my head out the window and yelling than with AT&T.
  • by karl.auerbach ( 157250 ) on Monday June 02, 2008 @04:59PM (#23631419) Homepage
    There are sometimes other places to hide data:

    I can't remember whether it was FTP Software of NetManage, but one of those used to hide the serial number of the software in the bits between the end of broadcast ARP requests and the end of the Ethernet frame.

    That way they could check for duplicate license keys on the same net without bothering anybody. Only worked across the broadcast domain, but that was adequate for that purpose.

    There's lots of other places too.

    RTP packets have optional extension headers that can be used, DNS can hold extra information in parts of the query and response packets - I once encountered someone tunneling music feed via buggered DNS packets. (It became very visible when it caused a Cisco firewall to go haywire.)
  • Voice is one place for stego, but Video over IP can use a lot more bandwidth, and gives you more places to hide info--you can do more with the codecs, and can "hide" information in the picture itself (hey, the bad guys could use sign language.) :)

    One interesting thing about the paper is that it implies that some types of DRM mimic stego. Is this a reason to outlaw DRM?
  • Just read the paper. While their research is entirely sound (no pun intended), the value of their research is pretty limited.

    In circumstances like Skype (not RTP), it is possible to talk and text chat at the same time. All of it is encrypted.

    The application of this type of stegonographic message is for stored data. But for that, the data would have to be stored. There's just not point in storing a voice conversation as RTP packets on the users' system. In fact, it would be almost ridiculous to store audio i

Make it myself? But I'm a physical organic chemist!