This two part series discusses the W3C Recommendations Canonical XML and Exclusive XML Canonicalization. In this first part I describe the process of XML canonicalization, that is, of finding the simplified form of an XML document, as defined by the Canonical XML specification. We'll start by illustrating when and why we would need to canonicalize an XML document.
XML defines a format for structuring data so that information can be meaningfully interchanged between communicating parties. The rules for XML authoring are flexible in the sense that the same document structure and the same piece of information can be represented by different XML documents. Consider Listings 1 and 2, which are logically equivalent, i.e. they follow the same document structure (the same XML Schema) and are meant to convey the same information. In spite of being logically equivalent, the XML files of Listings 1 and 2 do not contain the same sequence of characters (or sequence of bytes or octets).
In this case the character and octet sequences of the two XML files
differ due to the order of attributes appearing in the
element. There can be other reasons for having different octet streams for
logically equivalent XML documents. The purpose of finding the canonical
(or simplified) form of an XML document is to determine logical
equivalence between XML documents. W3C has defined canonicalization rules
such that the canonical form of two XML documents will be the same if they
are logically equivalent.
Whenever we are required to determine whether two XML documents are logically equivalent, we will canonicalize each of them and compare the canonical forms octet-by-octet. If the two canonical forms contain the same sequence of octets, we will conclude that the two XML files are logically equivalent.
Before we start exploring the technical details of the canonicalization process, let's see when and why you would need to test logical equivalence between XML documents.
The Need to Find Logical Equivalence between XML Documents
XML Digital Signatures is a W3C recommendation defining an XML format for signing XML (or non-XML) documents. Signing XML documents can be considered analogous to signing paper documents in ink. If you ever need to sign paper documents (e.g. bank checks) while doing business in the conventional way, you'll likely need to sign digital documents while doing business electronically over the Internet.
Let's consider the process of signing a bank check. A bank check is part of a system that is designed to handle the following two critical requirements:
- Integrity: A check, once signed and issued cannot be edited by an unauthorized party. If someone tries to modify the amount written on a check, he will damage the background imprint on the check. The background imprint on the check, therefore, helps in maintaining confidence in the integrity of bank checks.
- Non-repudiation: The signing authority cannot deny the act of issuing the check. The signatures on the check are verifiable against sample signatures lying with the bank.
The two requirements are exactly the same in XML Digital Signatures. The background imprint on a bank check is analogous to a Message Digest algorithm, while the sample signatures lying with the bank are analogous to a private-public key pair in a PKI trust service.
Message Digests and PKI are very common terms used in literature covering security over the Internet. However, for those XML programmers who are new to security issues, the following is a brief description of these terms. For more details, I've included a reference at the end of the article.
Digest algorithms act upon and digest (consume) message octets to calculate a digest value. Digest values depend upon the message digest algorithm and the message itself. If an intruder modifies the message, its digest value can be used to detect the modification. Therefore message digests are the electronic way of enforcing integrity.
Public Key Infrastructure (PKI) Trust Services
Trust services produce and maintain private-public key pairs. Trust services are the custodians of private-public key pairs, just like your bank is the custodian of your money as well as your account details.
Private keys are kept confidential i.e. only the trust service and the user (signing authority) know the value of the private key. On the other hand public keys are open to everybody. A signer will use his private key to sign a message electronically along with the message digest value. Anyone can use the corresponding public key to verify the signature and the message digest. Public keys can only be used to verify signatures; they cannot be used to produce signatures. Therefore, as long as your private key is confidential, unauthorized signatures cannot be produced.
Signing Digital Data
When a user wants to sign an XML document, there is a two step procedure. First, the user digests the XML file to be signed and produces a digest value. Secondly, the user signs a reference to the XML message and the digest with the user's private key.
An application receiving the signed message will validate the signature using signer's public key and will verify the integrity of the message by checking the message digest. The process of digesting an XML message relies on the sequence of bytes that represent an XML message. It is actually the sequence of bytes that produces a message digest value and the signatures. This means that Listings 1 and 2, although representing the same information, will not produce the same digest value. If you sign Listing 1 and the receiving party tries to verify Listing 2 against your signature (which makes perfect sense, since the two documents are carrying the same information), the verification process will fail.
Canonicalizing Before Signing and Verifying
Canonical XML finds its application in this scenario. The signing authority will digest and sign the canonical form of the document instead of the original form. Similarly, at the time of verification, the canonical form will be verified instead of the original XML. Thus all logically equivalent versions of the signed XML document will result in successful verification (validation) of XML digital signatures.
Canonicalizing an XML document
The canonical XML specification has defined an algorithm to author the canonical form of XML documents. You will need to perform the following steps in order to canonicalize an XML document:
1. Encoding Scheme
All XML documents are composed of human readable text, which is a sequence of characters. Encoding schemes are meant to represent characters by octets. Therefore, the same XML file can be represented by an entirely different octet stream just by changing the character encoding of the XML file.
The canonical XML specification dictates that the canonical form of XML documents should be encoded in UTF-8 encoding. Therefore, if the XML file to be canonicalized has any other encoding, it should be changed to UTF-8.
2. Line Breaks
Line breaks in text files are normally represented either by hexadecimal A (decimal 10) or hexadecimal D (decimal 13) or a combination of these two octets. XML files are all simple text files, therefore #xA and #xD are used as line breaks in all XML files.
The canonical form of XML requires that all line breaks (#xD or a combination of #xA and #xD) be replaced with #xA. This should be done before starting to process the XML file.
3. Attribute values are normalized
All attributes are required to be normalized in canonical form, as if by a validating XML parser. The process of attribute value normalization is stated in the XML 1.0 recommendation by W3C (see Resources).
A simple example of attribute value normalization is demonstrated by Listings 3 and 4. Listing 3 is the original XML file before attribute value normalization, while Listing 4 shows all attribute values in normalized form.
All attributes in Listing
3 are of string type without any entity and character references. In
this case, attribute value normalization simply means the normalization of
white space (all types of white space i.e. tabs, line breaks and normal
non-breaking space should be converted to #x20, which is the octet that
represents non-breaking white space). All
id attributes in Listing 3 have two tabs in
them. Each occurrence of tab in the
id attribute in Listing 3 value has been
changed to a space in Listing
Pages: 1, 2