Advanced Topic of the Week: Request Header Tagging
Request Header Tagging
Wouldn't it be cool if your WAF could share its data with the application it is protecting? This concept is similar to anti-SPAM SMTP apps that will add additional mime headers to emails providing the SPAM detection analysis information. The CRS is attempting to mimic this concept at the HTTP layer by adding additional request headers that provide insight into any ModSecurity events that may have triggered during processing. The advantage of this approach is that it allows a WAF to be in a detection-only mode while still providing attack data to the destination application server. The recieving app server may then inspect the WAF request headers and make a determination whether or not to process the transaction. This concept is valuable in distributed web environments and hosting architectures where a determination to block may only be appropriate at the destination app server.
This concept has actually been discussed as part of the OWASP AppSensor Project and we have added a new Detection Point for it entitled - RP2: External User Behavior
Suspicious External User Behavior
External (to the application) devices and systems (e.g. host and network IDS, file integrity monitoring, disk usage monitoring, anti-malware service, IPS, network firewall, web application firewall, web server logging, XML gateway, database firewall, SIEM) detect anomalous behavior by the user (e.g. session and/or IP address).
This information can be used by the application to contribute to its knowleage about a potential attacker. In some cases, the information could be detected by the application itself (e.g. XSS pattern black listing), but may be more effectively identified by the external device, or is not known to the application normally (e.g. requests for missing resources that the web server sees, but does not pass onto the application).
The greater the knowledge a device or system has about the application, the greater confidence can be given to evidence of suspicious behaviour. Therefore, for example, attempted SQL injection detexcted by a web application firewall (WAF) might be given greater weight than information from a network firewall about the IP address.
The power of AppSensor is its accuracy and low false positive rate, and the usage of external data should be carefully assessed to ensure it does not contribute to a higher false positive rate.
Example 1: An IDS has detected suspicious activity by a particular IP address, and this is used to temporarily tighten the attack detection thresholds for requests from all users in the same IP address range.
Example 2: An application is using the ModSecurity web application firewall with the Core Rule Set, and utilises the anomaly score data passed forward in the X-WAF-Events and X-WAF-Score HTTP headers (optional rules in modsecurity_crs_49_header_tagging.conf) to adjust the level of application logging for each user.
Example 3: Information from an instance of PHPIDS suggests request data may be malicious.
This rule set file will take all of the TX attack variable data and populate Apache ENV variables that Apache can then use to add X-WAF-Event request header data to the request.
Example showing the consolidated X-WAF-Events and X-WAF-Score data -