CVE-2016-8743: Input Validation

Apache HTTP Server, prior to release 2.4.25, accepted a broad pattern of unusual whitespace patterns from the user-agent, including bare CR, FF, VTAB in parsing the request line and request header lines, as well as HTAB in parsing the request line. Any bare CR present in request lines was treated as whitespace and remained in the request field member "therequest", while a bare CR in the request header field name would be honored as whitespace, and a bare CR in the request header field value was retained the input headers array. Implied additional whitespace was accepted in the request line and prior to the ':' delimiter of any request header lines.

These defects represent a security concern when httpd is participating in any chain of proxies or interacting with back-end application servers, either through modproxy or using conventional CGI mechanisms. In each case where one agent accepts such CTL characters and does not treat them as whitespace, there is the possiblity in a proxy chain of generating two responses from a server behind the uncautious proxy agent. In a sequence of two requests, this results in request A to the first proxy being interpreted as requests A + A' by the backend server, and if requests A and B were submitted to the first proxy in a keepalive connection, the proxy may interpret response A' as the response to request B, polluting the cache or potentially serving the A' content to a different downstream user-agent.

Affects versions since 2.2.0 up to 2.4.23

External Reference:

https://httpd.apache.org/security/vulnerabilities24.html#2.4.25

— Red Hat

apache. Multiple issues existed in Apache. These were addressed by updating Apache to version 2.4.25.

apache. Multiple issues were addressed by updating to version 2.4.27.

It was discovered that the HTTP parser in httpd incorrectly allowed certain characters not permitted by the HTTP protocol specification to appear unencoded in HTTP request headers. If httpd was used in conjunction with a proxy or backend server that interpreted those characters differently, a remote attacker could possibly use this flaw to inject data into HTTP responses, resulting in proxy cache poisoning.