feat(policy/expressions): add userAgent.isBrowserLike method

Signed-off-by: Xe Iaso <me@xeiaso.net>

data/botPolicies.yaml

@@ -44,6 +44,10 @@ bots:
 #     algorithm: slow # intentionally waste CPU cycles and time
 
 # Challenge clients with "Mozilla" or "Opera" in their user-agent string
-- import: (data)/common/legacy-challenge-everything.yaml
+#- import: (data)/common/legacy-challenge-everything.yaml
+
+- name: reject-browsers
+  action: DENY
+  expression: userAgent.isBrowser()
 
 dnsbl: false

data/common/challenge-browser-like.yaml

@@ -5,12 +5,4 @@
       all:
       - '"X-Http-Version" in headers'
       - headers["X-Http-Version"] == "HTTP/1.1"
-      - >-
-        (  userAgent.contains("Mozilla")
-        || userAgent.contains("Opera")
-        || userAgent.contains("Safari")
-        || userAgent.contains("Edge")
-        || userAgent.contains("Gecko")
-        || userAgent.contains("Windows")
-        || userAgent.contains("Linux")
-        )
+      - userAgent.isBrowserLike()

data/common/legacy-challenge-everything.yaml

@@ -1,5 +1,4 @@
 # Generic catchall rule
 - name: generic-browser
-  user_agent_regex: >-
-    Mozilla|Opera
+  expression: userAgent.isBrowserLike()
   action: CHALLENGE

lib/policy/expressions/environment.go

@@ -1,7 +1,11 @@
 package expressions
 
 import (
+	"strings"
+
 	"github.com/google/cel-go/cel"
+	"github.com/google/cel-go/common/types"
+	"github.com/google/cel-go/common/types/ref"
 	"github.com/google/cel-go/ext"
 )
 
@@ -20,6 +24,8 @@ func NewEnvironment() (*cel.Env, error) {
 		cel.DefaultUTCTimeZone(true),
 
 		// Variables exposed to CEL programs:
+
+		// Request metadata
 		cel.Variable("remoteAddress", cel.StringType),
 		cel.Variable("host", cel.StringType),
 		cel.Variable("method", cel.StringType),
@@ -28,11 +34,37 @@ func NewEnvironment() (*cel.Env, error) {
 		cel.Variable("query", cel.MapType(cel.StringType, cel.StringType)),
 		cel.Variable("headers", cel.MapType(cel.StringType, cel.StringType)),
 
+		// System load metadata
 		cel.Variable("load_1m", cel.DoubleType),
 		cel.Variable("load_5m", cel.DoubleType),
 		cel.Variable("load_15m", cel.DoubleType),
 
 		// Functions exposed to CEL programs:
+
+		// userAgent.isBrowserLike() method, used to detect if a user agent is likely a browser
+		// based on shibboleth words in the User-Agent string.
+		cel.Function("isBrowserLike",
+			cel.MemberOverload("userAgent_isBrowserLike_string",
+				[]*cel.Type{cel.StringType},
+				cel.BoolType,
+				cel.UnaryBinding(func(userAgentVal ref.Val) ref.Val {
+					var userAgent string
+					switch v := userAgentVal.Value().(type) {
+					case string:
+						userAgent = v
+					default:
+						return types.NewErr("invalid type %T", userAgentVal)
+					}
+
+					switch {
+					case strings.Contains(userAgent, "Mozilla"), strings.Contains(userAgent, "Opera"), strings.Contains(userAgent, "Gecko"), strings.Contains(userAgent, "WebKit"), strings.Contains(userAgent, "Apple"), strings.Contains(userAgent, "Chrome"), strings.Contains(userAgent, "Windows"), strings.Contains(userAgent, "Linux"):
+						return types.Bool(true)
+					default:
+						return types.Bool(false)
+					}
+				}),
+			),
+		),
 	)
 }