Security considerations for Web services
In WebSphere® Application Server Version 6, when you enable integrity, confidentiality, and the associated tokens within a Simple Object Access Protocol (SOAP) message, security is not guaranteed. When you configure Web services security, make every effort to verify that the result is not vulnerable to a wide range of attack mechanisms. This article provides some information about the possible security concerns that arise when you are securing Web services. This list of security concerns is not complete. You must conduct your own security analysis for your environment.
- Ensuring the message freshness
Message freshness involves protecting resources from a replay attack in which a message is captured and resent. Digital signatures, by themselves, cannot prevent a replay attack because a signed message can be captured and resent. It is recommended that you allow message recipients to detect message replay attacks when messages are exchanged through an open network. You can use the following elements, which are described in the Web services security specifications, for this purpose:
- Timestamp
- You can use the timestamp element to keep track of messages and to detect replays of previous messages. The WS-Security 2004 specification recommends that you cache time stamps for a given period of time. As a guideline, you can use five minutes as a minimum period of time to detect replays. Messages that contain an expired timestamp are rejected.
- Nonce
- A nonce is a child element of UsernameToken in the UsernameToken profile. Because each Nonce element has a unique value, recipients can detect replay attacks with relative ease.
Important: Both the time stamp and nonce element must be signed. Otherwise, these elements can be altered easily and therefore cannot prevent replay attacks.
- Using XML digital signature and XML encryption properly to avoid a potential security hole
The WS-Security 2004 specification defines how to use XML digital signature and XML encryption in Simple Object Access Protocol (SOAP) headers. Therefore, users must understand XML digital signature and XML encryption in the context of other security mechanisms and their possible threats to an entity. For XML digital signature, be aware of all of the security implications resulting from the use of digital signatures in general and XML digital signature in particular. When you build trust into an application based on a digital signature, incorporate other technologies such as certification trust validation based upon the Public Key Infrastructure (PKI). For XML encryption, the combination of digital signing and encryption over a common data item might introduce some cryptographic vulnerabilities. For example, when you encrypt digitally signed data, you might leave the digital signature in plain text and leave your message vulnerable to plain text guessing attacks. As a general practice, when data is encrypted, encrypt any digest or signature over the data. For more information, see http://www.w3.org/TR/xmlenc-core/#sec-Sign-with-Encrypt.
- Protecting the integrity of security tokens
The possibility of a token substitution attack exists. In this scenario, a digital signature is verified with a key that is often derived from a security token and is included in a message. If the token is substituted, a recipient might accept the message based on the substituted key, which might not be what you expect. One possible solution to this problem is to sign the security token (or the unique identifying data from which the signing key is derived) together with the signed data. In some situations, the token that is issued by a trusted authority is signed. In this case, there might not be an integrity issue. However, because application semantics and the environment might change over time, the best practice is to prevent this attack. You must assess the risk assessment based upon the deployed environment.
- Verifying the certificate to leverage the certificate path verification and the certificate revocation list
It is recommended that you verify that the authenticity or validity of the token identity that is used for digital signature is properly trusted. Especially for an X.509 token, this issue involves verifying the certificate path and using a certificate revocation list (CRL). In the Web services security implementation in WAS Version 6, the certificate is verified by the TokenConsumer element. WAS provides a default implementation for the X.509 certificate that uses the Java™ CertPath library to verify and validate the certificate. In the implementation, there is no explicit concept of a CRL. Rather, proper root certificates and intermediate certificates are prepared in files only. For a sophisticated solution, you might develop your own TokenConsumer implementation that performs certificate and CRL verification using the online CRL database or the Online Certificate Status Protocol (OCSP).
- Protecting the username token with a password
It is recommended that you do not send a password in a UsernameToken to a downstream server without protection. You can use transport-level security such as SSL (for example, HTTPS) or use XML encryption within Web services security to protect the password. The preferred method of protection depends upon your environment. However, you might be able to send a password to a downstream server as plain text in some special environments where you are positive that you are not vulnerable to an attack.
Securing Web services involves more work than just enabling XML digital signature and XML encryption. To properly secure a Web service, have knowledge about the Public Key Infrastructure (PKI). The amount of security that you need depends upon the deployed environment and the usage patterns. However, there are some basic rules and best practices for securing Web services. It is recommended that you read some books on PKI and read information on the Web Services Interoperability Organization (WS-I) Basic Security Profile (BSP).