The Foundations of Distributed Message-Level Security
Security is one of the most vital topics in Web services development today and will be for the foreseeable future. The lack of maturity of standards and tools in this area is the reason most often cited for large organizations delaying their commitment to Web services. The most important security standards are ready now, though, and the tools are coming online. Importantly, these Web Services Security standards are really not groundbreaking; they are, in turn, just extensions of very well-established information security standards.
In Chapter 2, "The Foundations of Web Services," we introduced Web services as a new form of middleware for building and integrating distributed applications by sending XML messages between computing nodes. Making Web services secure means making those messages secure and keeping them secure wherever they go. This chapter builds on the preceding chapter by adding a solid foundation in the principles of distributed message-level security. Those principles depend on solid knowledge of shared key cryptography and public key cryptography. We begin by setting the information security context for Web services.
The Challenges of Information Security for Web Services
Securing distributed systems is hard. Securing exchanged information between those systems is harder. Securing Web services with their distributed, shared, and exposed information (XML messages) is much harder still. In the following sections, we address each of these challenges in order.
Security of Distributed Systems Is Hard
In distributed systems, you are not securing just one system, but many. In addition, you are securing the interconnections between the distributed systems. The goal of distributed systems is to make a collection of independent computers appear to their users as a single integrated system. When you can successfully integrate separate and distinct networked computers together, the integrated system can handle higher workloads, aggregate more functionality, and share data that previously was locked up and inaccessible. But the more access points there are, the more places an attacker has to attack. In other words, a distributed system has all the security issues of one system multiplied many times over.
These distributed system security issues include access control, identity management, authentication, password management, authorization, encryption of confidential information on each node, integrity of information passing between nodes, and more. This is the reason the CORBA security specification is 430 pages long.
Because Web services use the Web and enable and even encourage integration between systems across organizational boundaries using public networks, they are, in fact, much more distributed than CORBA or any previous form of middleware could contemplate. Web services, when used in remote procedure call mode, involve some of the same issues as CORBA object invocation security.
A service-oriented architecturesuch as what can be achieved when Web services become pervasiveprojects a vision of geographically distributed services, many of which are publicly available shared services incorporated into numerous applications. Knowing who is using these services, what these users are authorized to do, and how to protect all information at these nodes is a difficult security task.
Security of Exchanged Information (Messages) Is Harder
When security administrators think about securing their systems today, they typically think about securing their organization's "four walls." They think about perimeter security such as firewalls, intrusion detection, honeypots, and DMZs. None of these elements deal with the security of messages in transit. Security administrators focus most on access control security believing that, if they know for sure who has access to their network, they implicitly have secured their critical assets: their information. In fact, perimeter security schemes are really a proxy for information security, because all along, what people are really trying to protect is the information created, processed, or stored on the machines inside their perimeters.
Distributed applications send back and forth messages that are the command, control, and coordination that make the distributed applications function. This means you have the problems of computer and network security you always had, plus the problems of information security for critical command and control messages flowing through the distributed system. Transport-based security like SSL provides one type of protection for messages in transitjust server-to-server non-persistent confidentialitybut offers little or no control at the application level, which is the place where the messages have meaning.
Messages must remain intact from sender to receiver regardless of how many hops occur in between. Shared services create a many-hops scenario in which messages go through many service endpoints on the way to their destination. The receiver must know for sure who the sender was to establish trust, which is critical in all aspects of business. The sender must be able to control who is allowed to see her messages because vital (and confidential) information is contained in them. In the case of transmitted legal documents, being able to prove what transpired and repudiate any denials is critical. Messages need to remain persistently secure not just while in transit, but also when residing in permanent storage because the message might contain highly confidential information such as a credit card number, Social Security number, or business deal terms.
Security of Web Services Is Hardest
Web services add a new dimension to the challenges already faced with distributed systems. Critical security questions must be reliably answered. Who is the end user? How do you maintain security when routing messages through multiple servers? How do you maintain security when using shared services?
Web services are fundamentally about how to share the burden ofand derive the benefits fromcomputing across a distributed network of computers connected with Web infrastructure and standards. These messagesbeing XMLare text-based, readable, and self-describing. Security of Web services is about knowing whether the Web services are talking to the correct endpoints; the communications between all the disparate elements are kept confidential; the messages transported maintain their integrity at all times; and the entities for which these Web services are being invoked are known, trusted, and authorized to use the services and can be clearly identified to all other authorized services.
The good news is that Web Services Security builds on existing security standards; very little that is new is being proposed. In general, Web Services Security provides an XML-based abstraction layer for established security technologies that delivers confidentiality, integrity, non-repudiation, authentication, and authorization.
Why All the FUD about Web Services Security?
Some Web Services Security vendors, the analysts that cover these companies, and the media are spreading fear, uncertainty, and doubt (FUD) about how an organization must fully implement Web Services Security before deploying the first Web service. Because early Web service deployments tend to be simple, straightforward, internal, behind the firewall, point-to-point integrations, sophisticated Web Services Security technologies are not always necessary. In fact, the same security used by these companies for securing their Web sitesnamely SSLis usually sufficient.
For the vendors who are selling either a product or service that implements Web Services Security in some form, however, it is in their best interest to have the early Web services adopters firmly believe they cannot be successful without these vendors' products or services. This thinking is a huge stretch, to put it mildly.
When designing an information security strategy for a Web services deployment, you must think like an information security professional. Think in terms of confidentiality, integrity, and the other building block security principles. If your messages need legal non-repudiation, perhaps you need to employ more sophisticated technologies. But if all you need to do is keep messages confidential between point A and point B, maybe what you already know how to do so well (use SSL) is perfectly sufficient.
To build on your understanding of these information security building blocks, you need to understand how the discipline of cryptography is applied to distributed messages. The following sections cover the shared key technologies and then cover public key technologies.