05-SEP-2013: PKI and Tcl r13 (See the current copy)

It's secure. Trust me.

The purpose of Public Key Infrastructure (PKI) is to provide a mechanism to establish the authenticity of an unknown party. Most people assume PKI is a complex system with a lot of moving parts. However, this is not the case. A simple PKI implementation is fairly straight-forward.

PKI establishes authenticity by using certificates, which certify that another entity (whom you already trust) has verified that this entity is who it claims to be. This certificate is presented by the unknown party as a means of identifying itself to you.

So what is to stop someone from providing false information in a certificate? That's where the public key aspect of PKI comes into play. Certificates are digitally signed by the issuer, also called a Certificate Authority, using public key cryptography.

To explain how digital signatures work, we must first explain public key cryptography. Public key cryptography uses mathematical operations to perform operations on numerical values that can only be reversed or verified with the opposite key. That is, if something is encrypted with the private key it can only be decrypted or verified with the public key. Conversely, if something is encrypted with the public key it can only be decrypted or verified with the private key.

The most common public key cryptography algorithm in use today is RSA [citation needed]. It is also one of the easiest to demonstrate. A simple definition of applying RSA is modular exponentiation. Modular arithmetic (as in the modulus operator) is applied to values that have had the the exponentiation operator applied to them. For example:

    26729 (plain-text) ^ 65537 (public-exponent) mod 37837 (public-modulus) = 36784 (cipher-text)

is modular exponentiation of 26729 raised the power of 65537 all modulo 37837, which results in 36784.

Looking closer at this example of modular exponentiation we can see some of the public key cryptography properties that we need starting to emerge. Specifically we cannot take the cipher-text and convert it back to the plain-text with just the information that we know. We must know the private-exponent (or be able to derive it by factoring the public-modulus, but more on that later).

Thus the above example is also an example of encrypting the value 26729 (which is 0x6869, or "hi" in ASCII) with an RSA key whose public-exponent (often simply referred to as the exponent) is 65537 and whose public-modulus (often simply referred to as the modulus) is 37837 resulting in the encrypted value of 36784.

However, if we know the private key we can reverse the operation:

    36784 (cipher-text) ^ 40193 (private-exponent) mod 37837 (public-modulus) = 26729 (plain-text)

This is not very useful for digital signatures because it requires the private-exponent to do anything meaningful. With a digital signature, we want to perform an operation on some plain-text using our private data (for RSA private-exponent) that can be verified using only public information (such as public-exponent and public-modulus). Fortunately RSA allows us to do that by simply swapping the values around. That is:

    12345 (plain-text) ^ 40193 (private-exponent) mod 37837 (public-modulus) = 3293 (cipher-text)

which CAN be reversed using only public information, as in:

    3293 (cipher-text) ^ 65537 (public-exponent) mod 37837 (public-modulus) = 12345 (plain-text)

RSA guarantees that the chance of there being another value of private-exponent which generates the same cipher-text for a given plain-text is extremely low relative to the size of the key. Therefore we can assume that if a given cipher-text can be decrypted to a plain-text using a given public-exponent and public-modulus then it must have been encrypted with the secretly held private-exponent and thus as long as private-exponent is protected as a secret, only the holder could have generated this message.

So to generate a digital signature, one just encrypts the data with the "private key" (private-exponent in RSA) and the message is valid when decrypted with the "public key" (public-exponent in RSA) then it must have been written by the holder of the private key.

This is a workable solution for short messages. However, RSA cannot encrypt messages larger than the size of the key. That is, if the key is 16 bits then the plain-text message can be no larger than 16 bits. Instead a cryptographically secure message digest algorithm such as MD5 (defined in RFC ???), SHA1 (defined in RFC ???), or SHA256 (defined in RFC ???) is used to compute a digest of the message which is typically much smaller than the message itself. This digest is then encrypted with the private key and verified with the public key. In this way, arbitrarily long messages can be digitally signed using digital signatures, ensuring that if the certificate is altered or forged after being signed it will be detected.

So what makes up a certificate ? Certificates are specified by the ITU-T standard X.509 [1] and contain the following information:

  1. X.509 Standard version number (optional) which identifies the revision of X.509 that this certificate is written in
  2. Issuer, which is the "distinguished name" of the entity who issued (that is, signed) the certificate
  3. Serial Number, which is a unique number per issuer to uniquely identify this certificate from the issuer
  4. Subject, which is the "distinguished name" of the entity who is being certified (and also who holds the private key)
  5. Issue date and Expiration date, which define the time frame in which the certificate is valid
  6. The public key, including the algorithm and algorithm-specific public key data -- for RSA this is the public-modulus and public-exponent
  7. If this is X.509 version number 3 then X.509 extensions may be specified which restrict the uses of this certificate
  8. The digital signature of all of the previous data, which for RSA is the cryptographic message digest of the previous data (encoded in ASN.1 Distinguished Encoding Rules (DER) as specified by ITU-T standard X.690 [2]) and then encrypted with the private key of the Issuer

So now we have a system where an unknown and untrusted party can assert to you a fully-qualified (also known as "distinguished") name and also provide a reference to who is certifying that this assertion is true within the parameters specified in the certificate.

What is to stop someone from taking a certificate from an existing entity, which must be publicly accessible otherwise there would be no way to identify the entity, and using it? Once again, public key cryptography is the answer here. A certificate only proves that the Issuer signed a request for the Subject -- it does not prove that you you are talking to is legitimately the subject specified in the certificate. But it does provide you a mechanism to do that. The public key being certified is in the certificate so all that is needed is for you to issue some sort of challenge for the party presenting the certificate to prove that it has the private key that corresponds with the certified public key. How this challenge is done depends on the protocol and is outside the scope of this document.