feat(waf): 新增 UA 检查节点 ua_check

支持要求携带 UA、浏览器/OS 白名单 and-or 匹配，以及优先屏蔽爬虫与非正常 UA。

internal/apps/agent/nginx/manager_test.go

@@ -782,6 +782,9 @@ func TestManagedWAFLuaExecutesCompiledGraphWithoutRequestIO(t *testing.T) {
 	if !strings.Contains(openRestyWAFRuntimeLua, `node.type == "ip_match"`) {
 		t.Fatal("expected WAF runtime to execute compiled IP match nodes")
 	}
+	if !strings.Contains(openRestyWAFRuntimeLua, `node.type == "ua_check"`) {
+		t.Fatal("expected WAF runtime to execute compiled UA check nodes")
+	}
 	checkStart := strings.Index(openRestyWAFRuntimeLua, "function _M.check()")
 	if checkStart < 0 || strings.Contains(openRestyWAFRuntimeLua[checkStart:], "io.open") {
 		t.Fatal("expected WAF request path not to perform file I/O")

internal/apps/agent/nginx/waf_runtime.lua

@@ -294,6 +294,108 @@ local function matches_ip_values(config, ip)
     return false
 end
 
+local function ua_trim(value)
+    return (string.gsub(value or "", "^%s*(.-)%s*$", "%1"))
+end
+
+local function ua_label_in(items, value)
+    if type(items) ~= "table" or not value then return false end
+    for _, item in ipairs(items) do
+        if tostring(item) == value then return true end
+    end
+    return false
+end
+
+local function match_ua_rules(ua_lower, rules, fallback)
+    if ua_lower == "" then return "Unknown" end
+    for _, rule in ipairs(rules) do
+        local matched = false
+        for _, token in ipairs(rule.contains or {}) do
+            if string.find(ua_lower, token, 1, true) then
+                matched = true
+                break
+            end
+        end
+        if not matched and type(rule.all_of) == "table" and #rule.all_of > 0 then
+            matched = true
+            for _, token in ipairs(rule.all_of) do
+                if not string.find(ua_lower, token, 1, true) then
+                    matched = false
+                    break
+                end
+            end
+        end
+        if matched then
+            local excluded = false
+            for _, token in ipairs(rule.none_of or {}) do
+                if string.find(ua_lower, token, 1, true) then
+                    excluded = true
+                    break
+                end
+            end
+            if not excluded then return rule.label end
+        end
+    end
+    return fallback
+end
+
+-- Mirrors internal/repository/analytics/browser.go browserRules / osRules.
+local browser_rules = {
+    { label = "WeChat", contains = { "micromessenger" } },
+    { label = "Postman", contains = { "postman" } },
+    { label = "CLI", contains = { "curl/", "wget/" } },
+    { label = "Edge", contains = { "edg/", "edgios/", "edga/" } },
+    { label = "Opera", contains = { "opr/", "opera" } },
+    { label = "Firefox", contains = { "firefox", "fxios" } },
+    { label = "Chrome", contains = { "crios", "chrome" }, none_of = { "chromium" } },
+    { label = "Chromium", contains = { "chromium" } },
+    { label = "Safari", contains = { "safari" } },
+    { label = "Bot", contains = { "bot", "spider", "crawler", "slurp" } },
+}
+
+local os_rules = {
+    { label = "Android", contains = { "android" } },
+    { label = "iOS", contains = { "iphone", "ipad", "ipod", "ios" } },
+    { label = "Windows", contains = { "windows" } },
+    { label = "macOS", contains = { "mac os x", "macintosh", "macos" } },
+    { label = "Chrome OS", contains = { "cros" } },
+    { label = "Linux", contains = { "linux" } },
+    { label = "Bot", contains = { "bot", "spider", "crawler" } },
+}
+
+local function parse_browser_name(ua)
+    return match_ua_rules(string.lower(ua or ""), browser_rules, "Other")
+end
+
+local function parse_os_name(ua)
+    return match_ua_rules(string.lower(ua or ""), os_rules, "Other")
+end
+
+local function matches_ua_check(config)
+    config = config or {}
+    local ua = ua_trim(ngx.var.http_user_agent or "")
+    if config.require_ua and ua == "" then return false end
+    local browser = parse_browser_name(ua)
+    local os_name = parse_os_name(ua)
+    if config.block_common_bots and (browser == "Bot" or os_name == "Bot") then return false end
+    if config.block_abnormal_ua and (browser == "Bot" or browser == "Other" or browser == "Unknown") then
+        return false
+    end
+    local browsers = array_or_empty(config.browsers)
+    local operating_systems = array_or_empty(config.operating_systems)
+    local has_browsers = #browsers > 0
+    local has_os = #operating_systems > 0
+    if not has_browsers and not has_os then return true end
+    local browser_ok = ua_label_in(browsers, browser)
+    local os_ok = ua_label_in(operating_systems, os_name)
+    if has_browsers and not has_os then return browser_ok end
+    if has_os and not has_browsers then return os_ok end
+    local mode = config.match_mode
+    if mode ~= "and" and mode ~= "or" then mode = "or" end
+    if mode == "and" then return browser_ok and os_ok end
+    return browser_ok or os_ok
+end
+
 local function fail_closed(reason)
     local dict = ngx.shared and ngx.shared.openflare_waf_config
     if not dict or not dict.add or dict:add("_damaged_graph_logged", true, 60) then
@@ -347,6 +449,8 @@ local function execute_graph(graph)
             local region_required = type(config.regions) == "table" and #config.regions > 0
             local country, region = geo_lookup(ngx.var.remote_addr or "", region_required)
             handle = (list_contains(config.countries, country) or list_contains(config.regions, region)) and "true" or "false"
+        elseif node.type == "ua_check" then
+            handle = matches_ua_check(node.config or {}) and "true" or "false"
         elseif node.type == "pow" then
             if pow_runtime.evaluate(node.config or {}) ~= true then
                 return { kind = "takeover" }

internal/apps/agent/nginx/waf_runtime_spec.lua

@@ -79,8 +79,15 @@ local function load_runtime(config, options)
     return runtime
 end
 
-local function reset_request(site, ip, uri, is_internal)
-    ngx.var = { openflare_waf_site = site, remote_addr = ip or "192.0.2.1", uri = uri or "/", request_id = "request-1", openflare_internal = is_internal == true }
+local function reset_request(site, ip, uri, is_internal, user_agent)
+    ngx.var = {
+        openflare_waf_site = site,
+        remote_addr = ip or "192.0.2.1",
+        uri = uri or "/",
+        request_id = "request-1",
+        openflare_internal = is_internal == true,
+        http_user_agent = user_agent,
+    }
     ngx.ctx = {}
     ngx.header = {}
     output = {}
@@ -531,6 +538,98 @@ local function test_request_path_has_no_file_io()
     io.open = original_open
 end
 
+local function test_ua_check_require_block_and_whitelist()
+    local chrome_ua = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36"
+    local safari_ios_ua = "Mozilla/5.0 (iPhone; CPU iPhone OS 17_0 like Mac OS X) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/17.0 Mobile/15E148 Safari/604.1"
+    local bot_ua = "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)"
+    local weird_ua = "TotallyUnknownClient/1.0"
+
+    local function ua_graph(config)
+        return graph({
+            start = start_to("ua"),
+            ua = node("ua_check", config, { ["true"] = "allow", ["false"] = "blocked" }),
+            blocked = node("block", { status_code = 403, response_body = "ua blocked" }),
+            allow = node("allow"),
+        })
+    end
+
+    local runtime = load_runtime({
+        rule_groups = { rule(1, false, ua_graph({ require_ua = true })) },
+        bindings = { binding("ua-site", { 1 }) },
+    })
+    reset_request("ua-site", nil, nil, nil, nil)
+    runtime.check()
+    assert_equal(output.exit, 403, "missing UA with require_ua should block")
+    reset_request("ua-site", nil, nil, nil, chrome_ua)
+    output = {}
+    runtime.check()
+    assert_equal(output.exit, nil, "present UA with require_ua should allow")
+
+    runtime = load_runtime({
+        rule_groups = { rule(1, false, ua_graph({ block_common_bots = true })) },
+        bindings = { binding("ua-site", { 1 }) },
+    })
+    reset_request("ua-site", nil, nil, nil, bot_ua)
+    runtime.check()
+    assert_equal(output.exit, 403, "common bot should be blocked")
+
+    runtime = load_runtime({
+        rule_groups = { rule(1, false, ua_graph({ block_abnormal_ua = true })) },
+        bindings = { binding("ua-site", { 1 }) },
+    })
+    reset_request("ua-site", nil, nil, nil, weird_ua)
+    runtime.check()
+    assert_equal(output.exit, 403, "abnormal UA should be blocked")
+    reset_request("ua-site", nil, nil, nil, chrome_ua)
+    output = {}
+    runtime.check()
+    assert_equal(output.exit, nil, "normal browser should pass abnormal check")
+
+    runtime = load_runtime({
+        rule_groups = { rule(1, false, ua_graph({ browsers = { "Chrome" }, match_mode = "or" })) },
+        bindings = { binding("ua-site", { 1 }) },
+    })
+    reset_request("ua-site", nil, nil, nil, safari_ios_ua)
+    runtime.check()
+    assert_equal(output.exit, 403, "Safari should miss Chrome whitelist")
+    reset_request("ua-site", nil, nil, nil, chrome_ua)
+    output = {}
+    runtime.check()
+    assert_equal(output.exit, nil, "Chrome should hit whitelist")
+
+    runtime = load_runtime({
+        rule_groups = { rule(1, false, ua_graph({
+            browsers = { "Chrome" },
+            operating_systems = { "iOS" },
+            match_mode = "and",
+        })) },
+        bindings = { binding("ua-site", { 1 }) },
+    })
+    reset_request("ua-site", nil, nil, nil, chrome_ua)
+    runtime.check()
+    assert_equal(output.exit, 403, "Chrome desktop should fail Chrome+iOS and")
+    reset_request("ua-site", nil, nil, nil, safari_ios_ua)
+    output = {}
+    runtime.check()
+    assert_equal(output.exit, 403, "Safari iOS should fail Chrome+iOS and")
+
+    runtime = load_runtime({
+        rule_groups = { rule(1, false, ua_graph({
+            browsers = { "Chrome" },
+            operating_systems = { "iOS" },
+            match_mode = "or",
+        })) },
+        bindings = { binding("ua-site", { 1 }) },
+    })
+    reset_request("ua-site", nil, nil, nil, chrome_ua)
+    runtime.check()
+    assert_equal(output.exit, nil, "Chrome desktop should pass Chrome|iOS or")
+    reset_request("ua-site", nil, nil, nil, safari_ios_ua)
+    output = {}
+    runtime.check()
+    assert_equal(output.exit, nil, "Safari iOS should pass Chrome|iOS or")
+end
+
 test_ip_true_and_false()
 test_ipv6_exact_cidr_and_group()
 test_geo_true_and_false()
@@ -546,5 +645,6 @@ test_block_config_and_rule_order()
 test_damaged_graphs_fail_closed()
 test_null_binding_ids_are_treated_as_empty()
 test_request_path_has_no_file_io()
+test_ua_check_require_block_and_whitelist()
 
 return true

internal/apps/openflare/waf/graph_compile.go

@@ -88,6 +88,17 @@ func compileRuleNodeConfig(node RuleNode) (any, error) {
 	case RuleNodePoW:
 		var config PoWNodeConfig
 		return config, decodeStrictConfig(node.Config, &config)
+	case RuleNodeUACheck:
+		var config UACheckConfig
+		if err := decodeStrictConfig(node.Config, &config); err != nil {
+			return nil, err
+		}
+		config.Browsers = sortedUniqueStrings(config.Browsers)
+		config.OperatingSystems = sortedUniqueStrings(config.OperatingSystems)
+		if config.MatchMode == "" {
+			config.MatchMode = UACheckMatchModeOr
+		}
+		return config, nil
 	case RuleNodeBlock:
 		var config BlockNodeConfig
 		return config, decodeStrictConfig(node.Config, &config)

internal/apps/openflare/waf/graph_compile_test.go

@@ -42,6 +42,39 @@ func TestCompileRuleGraph(t *testing.T) {
 	}
 }
 
+func TestCompileUACheckConfigNormalizesListsAndMatchMode(t *testing.T) {
+	graph := RuleGraph{SchemaVersion: RuleGraphSchemaVersion, Nodes: []RuleNode{
+		{ID: "start", Type: RuleNodeStart, Config: rawConfig(`{}`)},
+		{ID: "ua", Type: RuleNodeUACheck, Config: rawConfig(`{"browsers":["Safari","Chrome","Chrome"],"operating_systems":["iOS","Android"],"require_ua":true,"block_common_bots":true}`)},
+		{ID: "allow", Type: RuleNodeAllow, Config: rawConfig(`{}`)},
+		{ID: "block", Type: RuleNodeBlock, Config: rawConfig(`{"status_code":403}`)},
+	}, Edges: []RuleEdge{
+		{ID: "e1", Source: "start", SourceHandle: "next", Target: "ua"},
+		{ID: "e2", Source: "ua", SourceHandle: "true", Target: "allow"},
+		{ID: "e3", Source: "ua", SourceHandle: "false", Target: "block"},
+	}}
+	compiled, err := CompileRuleGraph(graph)
+	if err != nil {
+		t.Fatalf("CompileRuleGraph() error = %v", err)
+	}
+	cfg, ok := compiled.Nodes["ua"].Config.(UACheckConfig)
+	if !ok {
+		t.Fatalf("config type = %T", compiled.Nodes["ua"].Config)
+	}
+	if !reflect.DeepEqual(cfg.Browsers, []string{"Chrome", "Safari"}) {
+		t.Fatalf("browsers = %#v", cfg.Browsers)
+	}
+	if !reflect.DeepEqual(cfg.OperatingSystems, []string{"Android", "iOS"}) {
+		t.Fatalf("os = %#v", cfg.OperatingSystems)
+	}
+	if cfg.MatchMode != UACheckMatchModeOr {
+		t.Fatalf("match_mode = %q, want or", cfg.MatchMode)
+	}
+	if !cfg.RequireUA || !cfg.BlockCommonBots || cfg.BlockAbnormalUA {
+		t.Fatalf("flags = %#v", cfg)
+	}
+}
+
 func TestCompileRuleGraphIsDeterministicForNodeAndEdgeOrder(t *testing.T) {
 	first := RuleGraph{SchemaVersion: RuleGraphSchemaVersion, Nodes: []RuleNode{
 		{ID: "start", Type: RuleNodeStart, Config: rawConfig(`{}`)},

internal/apps/openflare/waf/graph_types.go

@@ -24,6 +24,8 @@ const (
 	RuleNodeGeoMatch RuleNodeType = "geo_match"
 	// RuleNodePoW runs a proof-of-work challenge before continuing.
 	RuleNodePoW RuleNodeType = "pow"
+	// RuleNodeUACheck branches on User-Agent presence, classification, and lists.
+	RuleNodeUACheck RuleNodeType = "ua_check"
 )
 
 // RuleGraph is the editor-facing representation of an executable WAF graph.
@@ -83,6 +85,22 @@ type BlockNodeConfig struct {
 	ResponseBody string `json:"response_body,omitempty"`
 }
 
+// UACheckConfig configures User-Agent presence, whitelist, and block switches.
+type UACheckConfig struct {
+	RequireUA        bool     `json:"require_ua"`
+	Browsers         []string `json:"browsers,omitempty"`
+	OperatingSystems []string `json:"operating_systems,omitempty"`
+	MatchMode        string   `json:"match_mode,omitempty"`
+	BlockCommonBots  bool     `json:"block_common_bots"`
+	BlockAbnormalUA  bool     `json:"block_abnormal_ua"`
+}
+
+// UA check match modes.
+const (
+	UACheckMatchModeAnd = "and"
+	UACheckMatchModeOr  = "or"
+)
+
 // DefaultRuleGraph returns the minimal start-to-allow graph.
 func DefaultRuleGraph() RuleGraph {
 	return RuleGraph{SchemaVersion: RuleGraphSchemaVersion, Nodes: []RuleNode{

internal/apps/openflare/waf/graph_validate.go

@@ -87,7 +87,7 @@ func validateRuleGraphNodes(ctx context.Context, graphNodes []RuleNode, ipGroupE
 			startID = node.ID
 		case RuleNodeAllow:
 			allowCount++
-		case RuleNodeBlock, RuleNodeIPMatch, RuleNodeGeoMatch, RuleNodePoW:
+		case RuleNodeBlock, RuleNodeIPMatch, RuleNodeGeoMatch, RuleNodePoW, RuleNodeUACheck:
 		default:
 			return nil, "", fmt.Errorf("节点 %s 的类型 %s 未知", node.ID, node.Type)
 		}
@@ -180,6 +180,8 @@ func validateRuleNodeConfig(ctx context.Context, node RuleNode, exists func(cont
 		return validateGeoMatchNodeConfig(node)
 	case RuleNodePoW:
 		return validatePoWNodeConfig(node)
+	case RuleNodeUACheck:
+		return validateUACheckNodeConfig(node)
 	case RuleNodeBlock:
 		return validateBlockNodeConfig(node)
 	}
@@ -283,6 +285,42 @@ func validateBlockNodeConfig(node RuleNode) error {
 	return nil
 }
 
+func validateUACheckNodeConfig(node RuleNode) error {
+	var cfg UACheckConfig
+	if err := decodeNodeConfig(node, &cfg); err != nil {
+		return err
+	}
+	mode := cfg.MatchMode
+	if mode == "" {
+		mode = UACheckMatchModeOr
+	}
+	if mode != UACheckMatchModeAnd && mode != UACheckMatchModeOr {
+		return fmt.Errorf("节点 %s 的匹配模式必须为 and 或 or", node.ID)
+	}
+	for _, label := range cfg.Browsers {
+		if !uaBrowserLabels[label] {
+			return fmt.Errorf("节点 %s 的浏览器标签 %s 无效", node.ID, label)
+		}
+	}
+	for _, label := range cfg.OperatingSystems {
+		if !uaOSLabels[label] {
+			return fmt.Errorf("节点 %s 的操作系统标签 %s 无效", node.ID, label)
+		}
+	}
+	return nil
+}
+
+var uaBrowserLabels = map[string]bool{
+	"Chrome": true, "Safari": true, "Firefox": true, "Edge": true, "Opera": true,
+	"Chromium": true, "WeChat": true, "Postman": true, "CLI": true, "Bot": true,
+	"Unknown": true, "Other": true,
+}
+
+var uaOSLabels = map[string]bool{
+	"Android": true, "iOS": true, "Windows": true, "macOS": true, "Chrome OS": true,
+	"Linux": true, "Bot": true, "Unknown": true, "Other": true,
+}
+
 func decodeNodeConfig(node RuleNode, dst any) error {
 	if err := decodeStrictConfig(node.Config, dst); err != nil {
 		return fmt.Errorf("节点 %s 的配置无效: %w", node.ID, err)
@@ -321,7 +359,7 @@ func requiredHandles(t RuleNodeType) []string {
 	switch t {
 	case RuleNodeStart, RuleNodePoW:
 		return []string{"next"}
-	case RuleNodeIPMatch, RuleNodeGeoMatch:
+	case RuleNodeIPMatch, RuleNodeGeoMatch, RuleNodeUACheck:
 		return []string{"true", "false"}
 	default:
 		return nil

internal/apps/openflare/waf/graph_validate_test.go

@@ -77,6 +77,14 @@ func TestValidateRuleGraph(t *testing.T) {
 			g.Edges[1].SourceHandle = "next"
 			g.Edges = g.Edges[:2]
 		}, "节点 match-1 的 PoW 难度必须在 1-16 之间"},
+		{"invalid ua browser", func(g *RuleGraph) {
+			g.Nodes[1].Type = RuleNodeUACheck
+			g.Nodes[1].Config = rawConfig(`{"browsers":["NotABrowser"],"match_mode":"or"}`)
+		}, "节点 match-1 的浏览器标签 NotABrowser 无效"},
+		{"invalid ua match mode", func(g *RuleGraph) {
+			g.Nodes[1].Type = RuleNodeUACheck
+			g.Nodes[1].Config = rawConfig(`{"match_mode":"xor"}`)
+		}, "节点 match-1 的匹配模式必须为 and 或 or"},
 		{"unknown config field", func(g *RuleGraph) { g.Nodes[1].Config = rawConfig(`{"ips":[],"surprise":true}`) }, "节点 match-1 的配置无效"},
 		{"null config", func(g *RuleGraph) { g.Nodes[1].Config = rawConfig(`null`) }, "节点 match-1 的配置无效"},
 		{"too many nodes", func(g *RuleGraph) {