Canonicalization Actions from Canonical XML Version 1.0 - Security

Following is a list of the changes that occur to an XML document when it goes through a Canonical XML version 1.0 Transform (you can find the XML Canonicalization specification at http://www.w3.org/TR/2001/REC-xml-c14n-20010315):

• The document is encoded in UTF-8.

• Line breaks are normalized to #xA on input, before parsing.

• Attribute values are normalized, as if by a validating processor.

• Character and parsed entity references are replaced.

• CDATA sections are replaced with their character content.

• The XML declaration and Document Type Definition (DTD) are removed.

• Empty elements are converted to start-end tag pairs.

• Whitespace outside the document element and within start and end tags is normalized.

• All whitespace in character content is retained (excluding characters removed during linefeed normalization).

• Attribute value delimiters are set to quotation marks (double quotes).

• Special characters in attribute values and character content are replaced by character references.

• Superfluous namespace declarations are removed from each element.

• Default attributes are added to each element.

• Lexicographic order is imposed on the namespace declarations and attributes of each element.

Although the concept of standardizing the XML before digesting and before verifying is reasonably simple, complex subtleties arise in some specific situations. One of the subtleties became a significant issue as practical implementations of XML Signing were in the early stages of creation. It came up when applying XML Signing to fragments of XML within a larger XML document—which is the predominant scenario in Web services using SOAP. The issue was primarily related to namespaces. The XML Canonicalization 1.0 specification says that namespaces are to be propagated down to all its descendents. This makes sense because, in the context of the document, these namespaces are implicit, so making them explicit seems like a reasonable strategy. However, when you remove an XML fragment from the context of a document, as in the case of SOAP when you have an XML payload within the overall document, this can cause significant problems.

This issue resulted in the creation of a new canonicalization algorithm called Exclusive Canonicalization (http://www.w3.org/TR/2002/REC-xml-exc-c14n-20020718/). This canonicalization method strives to "exclude ancestor context" as much as is practical. It primarily does this by not propagating the ancestor namespaces down to the children nodes. This canonicalization approach turns out to be the most practical in most circumstances and the one that you should use.

The extensibility of the canonicalization method is a tribute to the XML Signature working group and is what made the advent of the Exclusive Canonicalization method possible. Custom canonicalization methods are also possible; however, you should be extremely careful about using one because the canonicalization algorithm modifies the XML to be signed and verified in a way that is difficult to see. It would not be difficult for a canonicalization algorithm to do something such as modify the XML to have all signatures verified correctly.

To summarize, canonicalization is an important concept that is straightforward but can quickly become complex. (Just search canonicalization on the Net, and you will see what we mean.) Most of the time in Web Services Security, you can use Exclusive Canonicalization and you will be fine. We describe canonicalization further when we discuss the Transform element because, as we mentioned earlier, the CanonicalizationMethod designates which Canonicalization Transform to use over the SignedInfo element.