Loading ...
Sorry, an error occurred while loading the content.
 

Re: [soapbuilders] Re: Super-Encryption AND Digital Signatures

Expand Messages
  • mlong@bridgetonconsulting.com
    ... There is nothing wrong with the dsig, but it s overhead (and plenty of it). IMHO, the necessity for large and scaleable secure applications it great. I
    Message 1 of 22 , Dec 3, 2003
      Quoting Rich Salz <rsalz@...>:

      > I'm curious why a standard dsig/enc combination isn't appropriate?

      There is nothing wrong with the dsig, but it's overhead (and plenty of it).

      IMHO, the necessity for large and scaleable secure applications it great. I
      don't believe that xml-enc + xml-dsig on a one to many basis will ever be
      functionally scaleable (except in name only).

      If you take my premise based on an intermediary, then the requirements for
      secured/encrypted data are only based on a single and relatively simple
      concept. Not to mention the reduction in overhead in the fact the sender and
      receiver need not exchange keys. Only the intermediary-2-sender and
      intermediary-2-receiver need to exchange public keys, i.e., the intermediary
      decrypts the sender's message; then encrypts the message for the receiver.
      That scenario to me dramatically increases the scalability of
      secured/encrypted interop over an unsecured channel.

      Thoughts?

      -Matt
    • Rich Salz
      In talking it over with a colleague here, we did find one weakness in your scheme. Once the recipient has unwrapped the outer part, and then unwrapped the
      Message 2 of 22 , Dec 3, 2003
        In talking it over with a colleague here, we did find one weakness in
        your scheme. Once the recipient has unwrapped the outer part, and then
        unwrapped the internal key, they can forge any message and make it
        appear as if it came from the sender. I don't know if you're worried
        about that or not. "But you promised me $10,000. not $1,000. See, I have
        your original message."

        One way to fix this might be to include a signed hash of the original
        document.

        You should look at PKCS#7 (sorry I wasn't clear, when I said what's
        wrong with a standard I meant any standard, not just XML DSIG).

        As for your intermediary approach.. you know about Kerberos, right?
        /r$
        --
        Rich Salz, Chief Security Architect
        DataPower Technology http://www.datapower.com
        XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
        XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
      • mlong@bridgetonconsulting.com
        ... You are correct, but sender-2-recipient is secured AFAIK, e.g., using SSL to send credit card info to a processor doesn t guarantee the processor isn t
        Message 3 of 22 , Dec 3, 2003
          Quoting Rich Salz <rsalz@...>:

          > In talking it over with a colleague here, we did find one weakness in
          > your scheme. Once the recipient has unwrapped the outer part, and then
          > unwrapped the internal key, they can forge any message and make it
          > appear as if it came from the sender. I don't know if you're worried
          > about that or not. "But you promised me $10,000. not $1,000. See, I have
          > your original message."

          You are correct, but sender-2-recipient is secured AFAIK, e.g., using SSL to
          send credit card info to a processor doesn't guarantee the processor isn't
          publishing the information to a chat room, but you inherently trust that VISA
          isn't doing that. Only the sender and the intended recipient can see/decrypt
          the information. Right!?!
          >
          > One way to fix this might be to include a signed hash of the original
          > document.
          >
          > You should look at PKCS#7 (sorry I wasn't clear, when I said what's
          > wrong with a standard I meant any standard, not just XML DSIG).
          >
          > As for your intermediary approach.. you know about Kerberos, right?

          I have doubts about Kerberos in the short-term, because of the overhead of
          ticket exchange and the decentralization of KDCs. One can certainly utilize
          Kerberos intra-enterprise effectively as generally you are working with a
          single KDC, but once you start scaling with reckless abandon outside the
          enterprise engineering issues compond rather quickly.

          This is a good discussion.

          -Matt
        • mlong@bridgetonconsulting.com
          ... Isn t the security equally as strong as sending CC info over SSL, or not? -Matt
          Message 4 of 22 , Dec 3, 2003
            Quoting Rich Salz <rsalz@...>:
            > Anyhow, when I posted the weakness I said it's something you *might*
            > care about. As long as the sender is is not concerned about the
            > possibility of receiver fraud (even if the receive has been compromised
            > by an adversary), than it doesn't matter. Whether or not that is an
            > issue depends on the application. But it certainly makes the technique
            > inapplicable for widepsread general use.

            Isn't the security equally as strong as sending CC info over SSL, or not?

            -Matt
          • Rich Salz
            ... Since VISA is liable for any fraud if they publish your ccard number, there is strong incentive for them to not do that kind of thing. Similarly, there are
            Message 5 of 22 , Dec 3, 2003
              > You are correct, but sender-2-recipient is secured AFAIK, e.g., using SSL to
              > send credit card info to a processor doesn't guarantee the processor isn't
              > publishing the information to a chat room, but you inherently trust that VISA
              > isn't doing that. Only the sender and the intended recipient can see/decrypt
              > the information. Right!?!

              Since VISA is liable for any fraud if they publish your ccard number,
              there is strong incentive for them to not do that kind of thing.
              Similarly, there are strict rules and penalties for disclosure of
              personal information. Now, SET was worried about this, and they made
              sure that things were encrypted every-which-way, so that, e.g., the
              Merchant never saw your CCard number, but passed it along to the
              CreditCard clearinghouse, who *could* see the number, but couldn't see
              some merchant info, etc. etc. It all turned out to be too heavyweight,
              and the existing legal framework (plus incentives from the CCard
              companies to merchants) was enough to support "just use SSL."

              Anyhow, when I posted the weakness I said it's something you *might*
              care about. As long as the sender is is not concerned about the
              possibility of receiver fraud (even if the receive has been compromised
              by an adversary), than it doesn't matter. Whether or not that is an
              issue depends on the application. But it certainly makes the technique
              inapplicable for widepsread general use.

              > This is a good discussion.

              Yup. Hope we're not boring everyone else. :)
              /r$

              --
              Rich Salz, Chief Security Architect
              DataPower Technology http://www.datapower.com
              XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
              XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
            • Rich Salz
              ... Yes; the weaknesses are pretty much the same: with a symmetric key either side can forge content. But for Ccards over the web, the out-of-band framework
              Message 6 of 22 , Dec 3, 2003
                > Isn't the security equally as strong as sending CC info over SSL, or not?

                Yes; the weaknesses are pretty much the same: with a symmetric key
                either side can forge content.

                But for Ccards over the web, the out-of-band framework makes it
                reasonable to do accept this risk. Neither the merchant nor the payment
                house want to expose the information they receive, nor do any other kind
                of "defraud the sender" thing because there are financial and legal
                incentives to not do that. SSL gives hop-by-hop privacy, with some
                information about the entity at the other end. It does *not* give
                signature-style content integrity, and it's not a end-to-end
                authenticated signed content. It's transport-level security (that's why
                the IETF renamed it to TLS). For retail CCard transactions over the Web,
                TLS is okay. For many other things -- where there are outside
                agreements, risk management, or other factors influencing things -- it
                is not okay.

                For example, a doctor should not be able to use either SSL or your
                technique to do online perscriptions. A corrupt pharmacist (perhaps
                he's a friend of Rush Limbaugh :), could turn the scrip from 100 pills
                of Vicodin to 500 pills of vicodin, and *the doctor can't prove
                otherwise.* That's why I said what you want to do might be okay for
                your application, but not for general purpose.

                /r$

                --
                Rich Salz, Chief Security Architect
                DataPower Technology http://www.datapower.com
                XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
                XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
              • mlong@bridgetonconsulting.com
                ... Hmmm...under my scenario the content can be duplicated, but not altered. Because the receiver does not have the private key of the sender, i.e., the
                Message 7 of 22 , Dec 4, 2003
                  Quoting Rich Salz <rsalz@...>:

                  > > Isn't the security equally as strong as sending CC info over SSL, or not?
                  >
                  > Yes; the weaknesses are pretty much the same: with a symmetric key
                  > either side can forge content.
                  >
                  > For example, a doctor should not be able to use either SSL or your
                  > technique to do online perscriptions. A corrupt pharmacist (perhaps
                  > he's a friend of Rush Limbaugh :), could turn the scrip from 100 pills
                  > of Vicodin to 500 pills of vicodin, and *the doctor can't prove
                  > otherwise.* That's why I said what you want to do might be okay for
                  > your application, but not for general purpose.


                  Hmmm...under my scenario the content can be duplicated, but not altered.
                  Because the receiver does not have the private key of the sender, i.e., the
                  receiver cannot re-encrypt the inner cipher value identically (to spoof the
                  original sender) due to the fact that the original sender's private key is not
                  known to the receiver.

                  However, the receiver could create a new outer cipher value (assuming the
                  receiver is the intended recipient) and forward the message. Then the
                  recipient of the forwarded message would then receive an inner cipher value !=
                  to the authenticated sender. Which tells the sender something is wrong
                  assuming a point-2-point processing model.

                  -Matt
                • Rich Salz
                  ... Perhaps I don t understand. I was talking about re-using the cipher to create a modified message. The adversary (compromised recipient) doesn t need to
                  Message 8 of 22 , Dec 4, 2003
                    > Hmmm...under my scenario the content can be duplicated, but not altered.
                    > Because the receiver does not have the private key of the sender, i.e., the
                    > receiver cannot re-encrypt the inner cipher value identically (to spoof the
                    > original sender) due to the fact that the original sender's private key is not
                    > known to the receiver.

                    Perhaps I don't understand. I was talking about re-using the cipher to
                    create a modified message. The adversary (compromised recipient)
                    doesn't need to generate a new inner key, he just re-uses it generate a
                    new messsage. Now the original sender can't prove the receiver wrong.
                    /r$

                    --
                    Rich Salz, Chief Security Architect
                    DataPower Technology http://www.datapower.com
                    XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
                    XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
                  • mlong@bridgetonconsulting.com
                    Allow me to try this will a little pseudo-math. Encryption for Sender: SymmetricKey1 = TripleDES(IV1,Key1) CipherValue1 = Encrypt[TripleDES(MyMessage)]
                    Message 9 of 22 , Dec 4, 2003
                      Allow me to try this will a little pseudo-math.

                      Encryption for Sender:

                      SymmetricKey1 = TripleDES(IV1,Key1)
                      CipherValue1 = Encrypt[TripleDES(MyMessage)]
                      InnerEncryptedKey = RSA_Encrypt(Key1) [using sender's RSA private key]

                      SymmetricKey2 = TripleDES(IV2,Key2)
                      CipherValue2 = Encrypt[TripleDES(CipherValue1)]
                      OuterEncryptedKey = RSA_Encrypt(Key2) [using receiver's RSA public key]


                      Decryption for Receiver:

                      Key2 = RSA_Decrypt[OuterEncryptedKey] [using receiver's RSA private key]
                      CipherValue1 = Decrypt[TripleDES(CipherValue2)]
                      Key1 = RSA_Decrypt[InnerEncryptedKey] [using sender's RSA public key]
                      MyMessage = Decrypt[TripleDES(CipherValue1)]

                      Naturally, IV1 and IV2 are the first 8 octets of their respective cipher
                      values. (omitted for brevity).

                      (1) You could decrypt MyMessage and send it two someone else.
                      (a) But you couldn't impersonate the sender (the senders private key is
                      not available)
                      (2) You could decrypt CipherValue2 and re-encrypt CipherValue1 which something
                      akin to CipherValue2' and forward the message.
                      (a) But the decryption of CipherValue1 now yields a sender not from the
                      source of the message sent, i.e. public key of forwarding sender cannot be
                      used to decrypt the Key1 of CipherValue1.

                      More thoughts and comments, please.

                      Thx,

                      -Matt









                      Quoting Rich Salz <rsalz@...>:

                      > > Hmmm...under my scenario the content can be duplicated, but not altered.
                      > > Because the receiver does not have the private key of the sender, i.e., the
                      >
                      > > receiver cannot re-encrypt the inner cipher value identically (to spoof the
                      >
                      > > original sender) due to the fact that the original sender's private key is
                      > not
                      > > known to the receiver.
                      >
                      > Perhaps I don't understand. I was talking about re-using the cipher to
                      > create a modified message. The adversary (compromised recipient)
                      > doesn't need to generate a new inner key, he just re-uses it generate a
                      > new messsage. Now the original sender can't prove the receiver wrong.
                      > /r$
                      >
                      > --
                      > Rich Salz, Chief Security Architect
                      > DataPower Technology http://www.datapower.com
                      > XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
                      > XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
                      >
                      >
                      >
                    • Rich Salz
                      Good -- a common notation. :) Now, a compromised receiver does the following. First, get the keys: Key2 = RSA_Decrypt[OuterEncryptedKey] [using receiver s RSA
                      Message 10 of 22 , Dec 4, 2003
                        Good -- a common notation. :)

                        Now, a compromised receiver does the following. First, get the keys:
                        Key2 = RSA_Decrypt[OuterEncryptedKey] [using receiver's RSA private key]
                        CipherValue1 = Decrypt[TripleDES(CipherValue2)]
                        Key1 = RSA_Decrypt[InnerEncryptedKey] [using sender's RSA public key]
                        Next, use those keys to create a bad message:
                        BadCipher1 = Encrypt[TripleDES(*BAD MESSAGE*, Key1)]
                        BadCipher2 = Encrypt[TripleDes(BadCipher1, Key2)]

                        The adversary can now present present the two encrypted keys and
                        BadCipher2, and nobody can prove they didn't come from the original
                        sender. ("But you told me to give 1,000 pills -- here's the message")

                        /r$
                        --
                        Rich Salz, Chief Security Architect
                        DataPower Technology http://www.datapower.com
                        XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
                        XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
                      • mlong@bridgetonconsulting.com
                        ... But BadCipher1 now has RSA_Encrypt(Key1) from the receiver and not the original sender, because the receiver never had access to the sender s RSA private
                        Message 11 of 22 , Dec 4, 2003
                          Quoting Rich Salz <rsalz@...>:

                          > Good -- a common notation. :)
                          >
                          > Now, a compromised receiver does the following. First, get the keys:
                          > Key2 = RSA_Decrypt[OuterEncryptedKey] [using receiver's RSA private key]
                          > CipherValue1 = Decrypt[TripleDES(CipherValue2)]
                          > Key1 = RSA_Decrypt[InnerEncryptedKey] [using sender's RSA public key]
                          > Next, use those keys to create a bad message:
                          > BadCipher1 = Encrypt[TripleDES(*BAD MESSAGE*, Key1)]
                          > BadCipher2 = Encrypt[TripleDes(BadCipher1, Key2)]
                          >
                          > The adversary can now present present the two encrypted keys and
                          > BadCipher2, and nobody can prove they didn't come from the original
                          > sender. ("But you told me to give 1,000 pills -- here's the message")

                          But BadCipher1 now has RSA_Encrypt(Key1) from the receiver and not the
                          original sender, because the receiver never had access to the sender's RSA
                          private key. Therefore, the proof of who sent the message resides with the
                          decryption of Key1 (since it is always encrypted with the sender's private
                          key).

                          Where is this breaking down?
                        • Rich Salz
                          ... No, use the original encrypted Key1. ... Right, but there s no proof of what content the sender provided. More precisely, the proof is who generated
                          Message 12 of 22 , Dec 4, 2003
                            >>Now, a compromised receiver does the following. First, get the keys:
                            >>Key2 = RSA_Decrypt[OuterEncryptedKey] [using receiver's RSA private key]
                            >>CipherValue1 = Decrypt[TripleDES(CipherValue2)]
                            >>Key1 = RSA_Decrypt[InnerEncryptedKey] [using sender's RSA public key]
                            >>Next, use those keys to create a bad message:
                            >>BadCipher1 = Encrypt[TripleDES(*BAD MESSAGE*, Key1)]
                            >>BadCipher2 = Encrypt[TripleDes(BadCipher1, Key2)]

                            > But BadCipher1 now has RSA_Encrypt(Key1) from the receiver and not the
                            > original sender

                            No, use the original encrypted Key1.

                            > Therefore, the proof of who sent the message resides with the
                            > decryption of Key1

                            Right, but there's no proof of what content the sender provided. More
                            precisely, the proof is "who generated Key1", which is even further from
                            protecting the content.

                            > Where is this breaking down?

                            Our terminology, I think.

                            Both sender and receiver need Key1 and Key2. Since both 3DES is
                            symmetric, Key1 and Key2 can both encrypt and decrypt. Once the
                            adversary (compromised receiver) has both keys, he can use them to
                            encrypt anthing he wants. Now take that bogus message -- and the keys
                            *in their original wrappers* -- and prove the sender didn't generate
                            that content.

                            Does this help?
                            /r$
                            --
                            Rich Salz, Chief Security Architect
                            DataPower Technology http://www.datapower.com
                            XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
                            XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
                          • mlong@bridgetonconsulting.com
                            ... Thanks, I m clear now. Receiver decrypts both wrappers then reuses RSA_Encrypted(Key) for the outbound traffic. Now, my trusted intermediary/trust-broker
                            Message 13 of 22 , Dec 4, 2003
                              Quoting Rich Salz <rsalz@...>:
                              >
                              > No, use the original encrypted Key1.

                              Thanks, I'm clear now. Receiver decrypts both wrappers then reuses
                              RSA_Encrypted(Key) for the outbound traffic.

                              Now, my trusted intermediary/trust-broker scenario to combat this.
                              Setup:
                              (a) Subdivide users into two categories, i.e., consumers/senders and
                              providers/receivers.
                              (b) Consumers create endpoints within the intermediary to connect to
                              providers.
                              (c) Both consumers and providers share Public Key with intermediary, but not
                              with each other (not that it matters).
                              (d) Providers share authentication information for their respective services
                              with the intermediary, but not the consumers. Likewise for consumers.

                              The Message:
                              (1) Encrypted message sent by consumer "C" to intermediary "I" targeted at
                              provider "P".
                              (2) "I" authenticates "C" and uses the auth info to determine Public Key match
                              for encrypted key.
                              (3) "I" fully decrypts the message, then re-encrypts the message based "I"
                              Private Key, "P" symmetric algorithm, and "P's" Public Key.
                              (3) "I" forwards message to "P" utilizing orthogonal authentication info
                              provided by "P" and original sender identification (which doesn't need
                              encryption AFAIK)
                              (4) "P" authenticates "I" and such decrypts message using "I's" PK.

                              Therefore, as long as the message authenticates for "I", then it must come
                              from "I" where the original sender has been authenticated and message properly
                              decrypted and re-encrypted. Naturally, this relies on the authentication
                              information being well protected from both perspective of "C","P", and "I".

                              Tell me where this break, please.

                              Thx,

                              -Matt
                            • Rich Salz
                              A simpler fix is for the sender to do SHA1(message), and then encrypt (key1+digest) with their private key. That s simpler because it s a classic digital
                              Message 14 of 22 , Dec 4, 2003
                                A simpler fix is for the sender to do SHA1(message), and then
                                encrypt (key1+digest) with their private key. That's simpler
                                because it's a classic digital signature, and its properties are
                                well understood.

                                The two biggest problems with your current idea are that
                                1. "I" must be online and completely trusted for every single
                                message exchange. This gives up all the benefits of public-
                                key crypto.
                                2. There's no end-to-end security link. What prevents P from
                                using his own keypair to forge a message that looks like
                                I-on-behalf-of-C?

                                A simpler fix for your first scheme might be for the sender to include
                                RSA_Encrypt(SHA1(message)) alongside the encrypted key1. Then perhaps
                                you include a timestamp, so adversaries can't capture and reply old
                                messages.

                                I know you think that the standard mechanisms are expensive and full
                                of overhead. There's a reason: without them, you leave yourself
                                open to various attacks.
                                /r$

                                --
                                Rich Salz Chief Security Architect
                                DataPower Technology http://www.datapower.com
                                XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
                                XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
                              • mlong@bridgetonconsulting.com
                                Rich, RSA_Encrypt(SHA1(message) + key1) this looks promising. Also, doesn t eliminate the need for a xml digital signature. You see what I m after, i.e., high
                                Message 15 of 22 , Dec 9, 2003
                                  Rich,

                                  RSA_Encrypt(SHA1(message) + key1) this looks promising. Also, doesn't
                                  eliminate the need for a xml digital signature.

                                  You see what I'm after, i.e., high security + scalable implementable features
                                  + compact wire format.

                                  Thoughts!?!

                                  -Thx,

                                  -Matt



                                  Quoting Rich Salz <rsalz@...>:

                                  > A simpler fix is for the sender to do SHA1(message), and then
                                  > encrypt (key1+digest) with their private key. That's simpler
                                  > because it's a classic digital signature, and its properties are
                                  > well understood.
                                  >
                                  > The two biggest problems with your current idea are that
                                  > 1. "I" must be online and completely trusted for every single
                                  > message exchange. This gives up all the benefits of public-
                                  > key crypto.
                                  > 2. There's no end-to-end security link. What prevents P from
                                  > using his own keypair to forge a message that looks like
                                  > I-on-behalf-of-C?
                                  >
                                  > A simpler fix for your first scheme might be for the sender to include
                                  > RSA_Encrypt(SHA1(message)) alongside the encrypted key1. Then perhaps
                                  > you include a timestamp, so adversaries can't capture and reply old
                                  > messages.
                                  >
                                  > I know you think that the standard mechanisms are expensive and full
                                  > of overhead. There's a reason: without them, you leave yourself
                                  > open to various attacks.
                                  > /r$
                                  >
                                  > --
                                  > Rich Salz Chief Security Architect
                                  > DataPower Technology http://www.datapower.com
                                  > XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
                                  > XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
                                  >
                                  >
                                  >
                                • Rich Salz
                                  ... Yeah, there s no such thing as a free lunch. :) Folks often complain about how big SSL is, or how complicated XML DSIG is, etc. Unfortunately, they are
                                  Message 16 of 22 , Dec 9, 2003
                                    > You see what I'm after, i.e., high security + scalable implementable features
                                    > + compact wire format.
                                    >
                                    > Thoughts!?!

                                    Yeah, there's no such thing as a free lunch. :)

                                    Folks often complain about how "big" SSL is, or how complicated
                                    XML DSIG is, etc. Unfortunately, they are that way because they need
                                    to be in order to be resistant to various threats. And then you have
                                    to fight the deployment barriers: if SSL, PKCS#7 and/or XML DSIG are
                                    already everywhere, what's the incentive to try something that hasn't
                                    had the same level of analysis? Unless you're Ron Rivest (the R of RSA)
                                    designing a new micro-payment protocol (www.peppercoin.com), you're
                                    generally better off accepting the trade-offs of commodity security
                                    mechanisms.

                                    Now, RSA_PublicKey_Encrypt(SHA1(message) + key1) seems reasonable
                                    to me. But it's quite possible that there's some obscure corner of
                                    crypto that makes this a bad idea. I still think it's worth
                                    posting it to the cryptography mailing list.

                                    /r$

                                    --
                                    Rich Salz Chief Security Architect
                                    DataPower Technology http://www.datapower.com
                                    XS40 XML Security Gateway http://www.datapower.com/products/xs40.html
                                    XML Security Overview http://www.datapower.com/xmldev/xmlsecurity.html
                                  Your message has been successfully submitted and would be delivered to recipients shortly.