SOA Security 101: Patching the Firewall Hole
Since SOAP messages containing XML payloads traverse across runtime service activities using the HTTP/HTTPS protocols, packet filtering firewalls and routers do not provide adequate security controls. Packet filtering devices will not be able to differentiate authorized well-formed XML messages with malicious messages. However, XML firewalls can bridge this gap.
First, XML firewalls provide packet level inspection of all inbound and outbound traffic to the back-end web application server. For inbound traffic, the device tears open packets to scan its contents against a catalog of threats before it forwards it on to the application for processing. For outbound traffic, the device may apply response filters to inspect packets for sensitive information inadvertently included in the response. Response filters can serve as a critical component to prevent confidential information from leaving an organization through Web services.
Next, the XML firewall must be strategically placed on the network. Depending on the application requirements, it is most typically placed in a DMZ security zone in front of the web application server, as illustrated in Figure 2. Coupled with the traditional packet filtering firewalls, placement in the DMZ offers the ideal position for the XML firewall to protect the business logic harvested in the application server. As the services become exposed to external networks beyond the host organization’s control, additional XML firewalls may be deployed to add more depth of security across the network.
Figure 2: A typical XML firewall architecture.
Furthermore, many of the major application servers, such as IBM WebSphere and BEA WebLogic, offer embedded packet inspection services. These value-add features may be appropriate depending on the amount of messages the server processes and the performance requirements of the application. Considering the processing overhead of packet inspection, the use of these embedded features may introduce significant overhead as the application scales. The industry best practice is to offload this security processing into a hardware-based XML firewall, such as IBM Data Power, Intel Sarvega, or Cisco Reactivity. Using a hardware device also adds an additional layer of protection into the overall security architecture.
In addition to packet inspection from an XML firewall, Web services may also be protected using message-level security. The use of standards offers a consistent framework for developers to implement security at the message level, especially since the information security community has helped mature Web services security standards over the past few years. Figure 3 describes the security mechanisms available in reference to layers of the Open Systems Interconnection (OSI) model.
Figure 3: Security protocols categorized by OSI layer.
A key feature of the WS-Security standard is that it makes use of existing technologies instead of inventing new ones. Specifically, WS-Security calls upon the following proven security techniques:
|•||XML Digital Signatures - Used to ensure that a message is not altered while in transit.|
|•||XML Encryption - Used to encrypt messages to make them unreadable except by those possessing the proper keys.|
|•||Authentication Tokens - A mechanism to restrict access to Web services to only those users that are authorized with supported authentication types: usernames and passwords, X.509 digital certificates, Security Assertion Markup Language (SAML).|
For instance, the WS-Policy specification offers security governance for encryption but it does not specify all of the key exchange properties necessary to make this work in the real world. The exchanging parties must still determine and agree upon on a key size, cipher, and algorithm. Also, Web security standards provide little support beyond SOAP and HTTP/HTTPS. Some solutions may still require proprietary specifications and technologies to fully protect their perimeters.
Next, if the technology infrastructure is not sized correctly, the advantages provided by message-level security may result in detrimental performance. Despite these inconsistencies, an architecture based upon standards is still far superior than relying on proprietary methods.
The proliferation of Web services has pushed the business case for SOA, for they have enabled unparalleled interoperability among a heterogeneous set of enterprise applications. Along with the increased integration, there has been a paradigm shift on how to secure an application given the new frontier of attacks on interdependent enterprise solutions. The new generation of attacks categorically include threats based upon XML denial of service, unauthorized access, data integrity, data confidentiality, and system compromise.
As the proverbial saying goes, “Security is only as good as its weakest link.” The external exposure of Web services has warranted inspection of the transaction payload and the use of message-level security. Thus conventional security controls must be amended to include Web service security mechanisms for service-oriented solution designs. The inclusion of XML firewalls for packet inspection and message-level security are therefore critical for the protection of SOA.
Industry standards are available and continue to evolve, but do not provide a comprehensive solution. The onus is on the security architect to use the standards as a baseline and devise appropriate security controls applicable to the specific implementation of the service-oriented solution.
(Note: Opinions expressed in this article and its replies are the opinions of their respective authors and not those of DZone, Inc.)