Merge branch 'reverse-proxy-allow-match-or' into reverse-proxy-crowdsec-appsec

# Conflicts:
#	proxy/internal/auth/middleware.go
#	proxy/internal/auth/middleware_test.go
#	proxy/internal/auth/tunnel_lookup_test.go
#	proxy/management_integration_test.go
#	proxy/server.go
#	shared/management/proto/proxy_service.pb.go

management/internals/modules/reverseproxy/domain/domain.go

@@ -1,5 +1,13 @@
 package domain
 
+import "time"
+
+// ValidationTTL is the time available to validate a custom domain registration.
+const ValidationTTL = 48 * time.Hour
+
+// ID identifies a custom domain registration.
+type ID string
+
 type Type string
 
 const (
@@ -8,12 +16,13 @@ const (
 )
 
 type Domain struct {
-	ID            string `gorm:"unique;primaryKey;autoIncrement"`
-	Domain        string `gorm:"unique"` // Domain records must be unique, this avoids domain reuse across accounts.
-	AccountID     string `gorm:"index"`
-	TargetCluster string // The proxy cluster this domain should be validated against
-	Type          Type   `gorm:"-"`
-	Validated     bool
+	ID                  string `gorm:"unique;primaryKey;autoIncrement"`
+	Domain              string `gorm:"unique"` // Domain records must be unique, this avoids domain reuse across accounts.
+	AccountID           string `gorm:"index"`
+	TargetCluster       string // The proxy cluster this domain should be validated against
+	Type                Type   `gorm:"-"`
+	Validated           bool
+	ValidationExpiresAt *time.Time `gorm:"index"`
 	// SupportsCustomPorts is populated at query time for free domains from the
 	// proxy cluster capabilities. Not persisted.
 	SupportsCustomPorts *bool `gorm:"-"`
@@ -39,7 +48,12 @@ func (d *Domain) EventMeta() map[string]any {
 	}
 }
 
+// Copy returns a copy with an independent validation deadline.
 func (d *Domain) Copy() *Domain {
 	dCopy := *d
+	if d.ValidationExpiresAt != nil {
+		expiresAt := *d.ValidationExpiresAt
+		dCopy.ValidationExpiresAt = &expiresAt
+	}
 	return &dCopy
 }

management/internals/modules/reverseproxy/domain/manager/domain_test.go

@@ -66,8 +66,8 @@ func TestExtractClusterFromFreeDomain(t *testing.T) {
 
 func TestExtractClusterFromCustomDomains(t *testing.T) {
 	customDomains := []*domain.Domain{
-		{Domain: "example.com", TargetCluster: "eu1.proxy.netbird.io"},
-		{Domain: "proxy.corp.io", TargetCluster: "us1.proxy.netbird.io"},
+		{Domain: "example.com", TargetCluster: "eu1.proxy.netbird.io", Validated: true},
+		{Domain: "proxy.corp.io", TargetCluster: "us1.proxy.netbird.io", Validated: true},
 	}
 
 	tests := []struct {
@@ -120,19 +120,49 @@ func TestExtractClusterFromCustomDomains(t *testing.T) {
 
 	for _, tc := range tests {
 		t.Run(tc.name, func(t *testing.T) {
-			cluster, ok := extractClusterFromCustomDomains(tc.domain, customDomains)
-			assert.Equal(t, tc.wantOK, ok)
-			if ok {
-				assert.Equal(t, tc.wantVal, cluster)
+			cluster, match := extractClusterFromCustomDomains(tc.domain, customDomains)
+			if !tc.wantOK {
+				assert.Equal(t, customDomainNoMatch, match, "unrelated domain should not match any custom domain")
+				return
 			}
+			assert.Equal(t, customDomainValidated, match, "validated custom domain should resolve a cluster")
+			assert.Equal(t, tc.wantVal, cluster)
 		})
 	}
 }
 
+// An unvalidated row must never yield a cluster: the account has not shown it
+// controls the name, so no service may be bound to it.
+func TestExtractClusterFromCustomDomains_UnvalidatedDomainRefused(t *testing.T) {
+	customDomains := []*domain.Domain{
+		{Domain: "example.com", TargetCluster: "eu1.proxy.netbird.io", Validated: false},
+	}
+
+	for _, serviceDomain := range []string{"example.com", "app.example.com"} {
+		t.Run(serviceDomain, func(t *testing.T) {
+			cluster, match := extractClusterFromCustomDomains(serviceDomain, customDomains)
+			assert.Equal(t, customDomainUnvalidated, match, "unvalidated row must be reported as such")
+			assert.Empty(t, cluster, "unvalidated row must not resolve a cluster")
+		})
+	}
+}
+
+// A more specific unvalidated row must not shadow a validated parent domain.
+func TestExtractClusterFromCustomDomains_ValidatedParentWinsOverUnvalidatedChild(t *testing.T) {
+	customDomains := []*domain.Domain{
+		{Domain: "example.com", TargetCluster: "cluster-generic", Validated: true},
+		{Domain: "app.example.com", TargetCluster: "cluster-app", Validated: false},
+	}
+
+	cluster, match := extractClusterFromCustomDomains("app.example.com", customDomains)
+	assert.Equal(t, customDomainValidated, match)
+	assert.Equal(t, "cluster-generic", cluster, "validated parent domain should provide the cluster")
+}
+
 func TestExtractClusterFromCustomDomains_OverlappingDomains(t *testing.T) {
 	customDomains := []*domain.Domain{
-		{Domain: "example.com", TargetCluster: "cluster-generic"},
-		{Domain: "app.example.com", TargetCluster: "cluster-app"},
+		{Domain: "example.com", TargetCluster: "cluster-generic", Validated: true},
+		{Domain: "app.example.com", TargetCluster: "cluster-app", Validated: true},
 	}
 
 	tests := []struct {
@@ -164,8 +194,8 @@ func TestExtractClusterFromCustomDomains_OverlappingDomains(t *testing.T) {
 
 	for _, tc := range tests {
 		t.Run(tc.name, func(t *testing.T) {
-			cluster, ok := extractClusterFromCustomDomains(tc.domain, customDomains)
-			assert.True(t, ok)
+			cluster, match := extractClusterFromCustomDomains(tc.domain, customDomains)
+			assert.Equal(t, customDomainValidated, match)
 			assert.Equal(t, tc.wantVal, cluster)
 		})
 	}

management/internals/modules/reverseproxy/domain/manager/expiration.go

@@ -0,0 +1,73 @@
+package manager
+
+import (
+	"context"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/management/internals/modules/reverseproxy/domain"
+	"github.com/netbirdio/netbird/management/server/activity"
+)
+
+const (
+	validationCleanupInterval = 60 * time.Minute
+	validationCleanupBatch    = 100
+)
+
+// RunValidationCleanup removes expired registrations on startup and hourly until cancellation.
+func (m Manager) RunValidationCleanup(ctx context.Context) {
+	ticker := time.NewTicker(validationCleanupInterval)
+	defer ticker.Stop()
+	for {
+		m.cleanupExpiredDomains(ctx, time.Now().UTC())
+		select {
+		case <-ctx.Done():
+			return
+		case <-ticker.C:
+		}
+	}
+}
+
+func (m Manager) cleanupExpiredDomains(ctx context.Context, now time.Time) {
+	var afterID domain.ID
+	for ctx.Err() == nil {
+		domains, err := m.store.GetExpiredCustomDomains(ctx, now, afterID, validationCleanupBatch)
+		if err != nil {
+			if ctx.Err() == nil {
+				log.WithContext(ctx).WithError(err).Error("list expired custom domain registrations")
+			}
+			return
+		}
+		for _, d := range domains {
+			if ctx.Err() != nil {
+				return
+			}
+			m.deleteExpiredDomain(ctx, d, now)
+			afterID = domain.ID(d.ID)
+		}
+		if len(domains) < validationCleanupBatch {
+			return
+		}
+	}
+}
+
+func (m Manager) deleteExpiredDomain(ctx context.Context, d *domain.Domain, now time.Time) {
+	deleted, err := m.store.DeleteExpiredCustomDomain(ctx, d, now)
+	if err != nil {
+		if ctx.Err() == nil {
+			log.WithContext(ctx).WithFields(log.Fields{"accountID": d.AccountID, "domainID": d.ID}).
+				WithError(err).Warn("could not expire custom domain registration")
+		}
+		return
+	}
+	if !deleted {
+		return
+	}
+	meta := d.EventMeta()
+	if d.ValidationExpiresAt != nil {
+		meta["validation_expires_at"] = d.ValidationExpiresAt.UTC().Format(time.RFC3339)
+	}
+	m.accountManager.StoreEvent(ctx, activity.SystemInitiator, d.ID, d.AccountID,
+		activity.CustomDomainValidationExpired, meta)
+}

management/internals/modules/reverseproxy/domain/manager/expiration_test.go

@@ -0,0 +1,274 @@
+package manager
+
+import (
+	"context"
+	"fmt"
+	"sync"
+	"testing"
+	"testing/synctest"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/management/internals/modules/reverseproxy/domain"
+	rpservice "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/service"
+	"github.com/netbirdio/netbird/management/server/activity"
+	"github.com/netbirdio/netbird/management/server/mock_server"
+	nbstore "github.com/netbirdio/netbird/management/server/store"
+)
+
+func TestValidateDomain_ExpiredRegistration(t *testing.T) {
+	env := setupDomainTest(t)
+	ctx := context.Background()
+	d, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "expired.example.com", testCluster)
+	require.NoError(t, err)
+	expiresAt := time.Now().Add(-time.Second)
+	db := env.store.(*nbstore.SqlStore).GetDB()
+	require.NoError(t, db.Model(&domain.Domain{}).Where("id = ?", d.ID).
+		Update("validation_expires_at", expiresAt).Error)
+	env.resolver.set("validation.expired.example.com", testCluster)
+
+	env.manager.ValidateDomain(ctx, accountA, accountAUser, d.ID)
+
+	stored := storedDomain(t, env.store, accountA, d.Domain)
+	require.NotNil(t, stored)
+	assert.False(t, stored.Validated, "an expired registration must not become usable before cleanup runs")
+}
+
+func TestCreateDomain_ValidationDeadline(t *testing.T) {
+	env := setupClockDomainTest(t)
+	synctest.Test(t, func(t *testing.T) {
+		ctx := context.Background()
+		createdAt := time.Now().UTC()
+		d, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "pending.example.com", testCluster)
+		require.NoError(t, err)
+		require.NotNil(t, d.ValidationExpiresAt)
+		assert.Equal(t, createdAt.Add(48*time.Hour), *d.ValidationExpiresAt, "new registrations get 48 hours")
+
+		time.Sleep(time.Hour)
+		env.manager.ValidateDomain(ctx, accountA, accountAUser, d.ID)
+		stored := storedDomain(t, env.store, accountA, d.Domain)
+		require.NotNil(t, stored)
+		require.NotNil(t, stored.ValidationExpiresAt)
+		assert.WithinDuration(t, *d.ValidationExpiresAt, *stored.ValidationExpiresAt, 0, "failed validation must not extend the deadline")
+	})
+}
+
+func TestCleanupExpiredDomains_Boundaries(t *testing.T) {
+	env := setupDomainTest(t)
+	events := captureDomainEvents(env)
+	ctx := context.Background()
+	now := time.Now().UTC().Truncate(time.Second)
+	tests := []struct {
+		name      string
+		expiresAt time.Time
+		validated bool
+		deleted   bool
+	}{
+		{"expired", now.Add(-time.Second), false, true},
+		{"deadline", now, false, true},
+		{"pending", now.Add(time.Second), false, false},
+		{"validated", now.Add(-time.Hour), true, false},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			d := createExpiringDomain(t, env, tt.name+".example.com", tt.expiresAt)
+			if tt.validated {
+				require.NoError(t, env.store.(*nbstore.SqlStore).GetDB().Model(d).Update("validated", true).Error)
+			}
+			env.manager.cleanupExpiredDomains(ctx, now)
+			stored := storedDomain(t, env.store, accountA, d.Domain)
+			if !tt.deleted {
+				assert.NotNil(t, stored, "pending and validated registrations must survive cleanup")
+				return
+			}
+			assert.Nil(t, stored, "expired unused registrations must be removed")
+			replacement, err := env.manager.CreateDomain(ctx, accountB, accountBUser, d.Domain, testCluster)
+			require.NoError(t, err)
+			assert.NotEqual(t, d.ID, replacement.ID, "the released name must receive a fresh registration")
+			assert.False(t, replacement.Validated, "the new account must validate its own registration")
+		})
+	}
+	got := events.get()
+	require.Len(t, got, 2, "only successful expiration deletions emit events")
+	for _, event := range got {
+		assert.Equal(t, activity.CustomDomainValidationExpired, event.Activity, "use the requested expiration event")
+		assert.Equal(t, activity.SystemInitiator, event.InitiatorID, "cleanup is attributed to the system")
+		assert.Equal(t, accountA, event.AccountID, "expiration belongs to the original account")
+		assert.NotEmpty(t, event.TargetID, "retain the deleted domain ID")
+		assert.NotEmpty(t, event.Meta["domain"], "retain the deleted domain name")
+		assert.NotEmpty(t, event.Meta["validation_expires_at"], "include the validation deadline")
+	}
+}
+
+func TestCleanupExpiredDomains_ContinuesPastProtectedBatch(t *testing.T) {
+	env := setupDomainTest(t)
+	ctx := context.Background()
+	now := time.Now().UTC()
+	for i := range validationCleanupBatch {
+		d := createExpiringDomain(t, env, fmt.Sprintf("protected-%d.example.com", i), now.Add(-time.Hour))
+		require.NoError(t, env.store.CreateService(ctx, &rpservice.Service{
+			ID: fmt.Sprintf("service-%d", i), AccountID: accountA, Domain: "app." + d.Domain,
+		}))
+	}
+	unprotected := createExpiringDomain(t, env, "unused.example.com", now.Add(-time.Hour))
+	env.manager.cleanupExpiredDomains(ctx, now)
+	assert.Nil(t, storedDomain(t, env.store, accountA, unprotected.Domain), "protected registrations must not starve later batches")
+	remaining, err := env.store.ListCustomDomains(ctx, accountA)
+	require.NoError(t, err)
+	assert.Len(t, remaining, validationCleanupBatch, "all registrations with dependent services must survive")
+}
+
+func TestCleanupExpiredDomains_ConcurrentWorkers(t *testing.T) {
+	env := setupDomainTest(t)
+	events := captureDomainEvents(env)
+	now := time.Now().UTC()
+	d := createExpiringDomain(t, env, "concurrent.example.com", now.Add(-time.Hour))
+	var workers sync.WaitGroup
+	for range 2 {
+		workers.Go(func() { env.manager.cleanupExpiredDomains(context.Background(), now) })
+	}
+	workers.Wait()
+	assert.Nil(t, storedDomain(t, env.store, accountA, d.Domain), "one worker must remove the expired registration")
+	assert.Len(t, events.get(), 1, "only the worker that deletes the row may emit the event")
+}
+
+func TestRunValidationCleanup_HourlyAndRestart(t *testing.T) {
+	env := setupClockDomainTest(t)
+	synctest.Test(t, func(t *testing.T) {
+		events := captureDomainEvents(env)
+		now := time.Now().UTC()
+		startup := createExpiringDomain(t, env, "startup.example.com", now.Add(-time.Hour))
+		hourly := createExpiringDomain(t, env, "hourly.example.com", now.Add(time.Minute))
+		ctx, cancel := context.WithCancel(context.Background())
+		done := make(chan struct{})
+		go func() {
+			defer close(done)
+			env.manager.RunValidationCleanup(ctx)
+		}()
+		synctest.Wait()
+		assert.Nil(t, storedDomain(t, env.store, accountA, startup.Domain), "startup must collect overdue registrations")
+		time.Sleep(59 * time.Minute)
+		synctest.Wait()
+		assert.NotNil(t, storedDomain(t, env.store, accountA, hourly.Domain), "cleanup must wait for the 60-minute interval")
+		time.Sleep(time.Minute)
+		synctest.Wait()
+		assert.Nil(t, storedDomain(t, env.store, accountA, hourly.Domain), "the hourly scan must collect expired registrations")
+		cancel()
+		<-done
+
+		offline := createExpiringDomain(t, env, "offline.example.com", time.Now().UTC().Add(time.Minute))
+		time.Sleep(2 * time.Hour)
+		assert.NotNil(t, storedDomain(t, env.store, accountA, offline.Domain), "a stopped worker must not continue deleting")
+		ctx, cancel = context.WithCancel(context.Background())
+		done = make(chan struct{})
+		go func() {
+			defer close(done)
+			env.manager.RunValidationCleanup(ctx)
+		}()
+		synctest.Wait()
+		assert.Nil(t, storedDomain(t, env.store, accountA, offline.Domain), "restart must use the persisted deadline")
+		cancel()
+		<-done
+		assert.Len(t, events.get(), 3, "each deletion should emit an expiration event")
+	})
+}
+
+type blockingDomainResolver struct {
+	started chan struct{}
+	release chan struct{}
+}
+
+func (r blockingDomainResolver) LookupCNAME(context.Context, string) (string, error) {
+	close(r.started)
+	<-r.release
+	return testCluster + ".", nil
+}
+
+func TestValidateDomain_DeadlinePassesDuringLookup(t *testing.T) {
+	for _, cleanup := range []bool{false, true} {
+		t.Run(fmt.Sprintf("cleanup=%t", cleanup), func(t *testing.T) {
+			env := setupClockDomainTest(t)
+			synctest.Test(t, func(t *testing.T) {
+				events := captureDomainEvents(env)
+				ctx := context.Background()
+				d, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "late.example.com", testCluster)
+				require.NoError(t, err)
+				resolver := blockingDomainResolver{started: make(chan struct{}), release: make(chan struct{})}
+				env.manager.validator.Resolver = resolver
+				done := make(chan struct{})
+				go func() {
+					defer close(done)
+					env.manager.ValidateDomain(ctx, accountA, accountAUser, d.ID)
+				}()
+				<-resolver.started
+				time.Sleep(48 * time.Hour)
+				if cleanup {
+					env.manager.cleanupExpiredDomains(ctx, time.Now().UTC())
+					_, err = env.store.CreateCustomDomain(ctx, accountB, d.Domain, testCluster, false)
+					require.NoError(t, err)
+				}
+				close(resolver.release)
+				<-done
+				owner := accountA
+				if cleanup {
+					assert.Nil(t, storedDomain(t, env.store, accountA, d.Domain), "late validation must not restore the old claim")
+					owner = accountB
+				}
+				stored := storedDomain(t, env.store, owner, d.Domain)
+				require.NotNil(t, stored)
+				assert.False(t, stored.Validated, "late validation must not validate either claim")
+				for _, event := range events.get() {
+					assert.NotEqual(t, activity.DomainValidated, event.Activity, "a rejected write must not emit a validation event")
+				}
+			})
+		})
+	}
+}
+
+func setupClockDomainTest(t *testing.T) *domainTestEnv {
+	t.Helper()
+	// Network driver watchers cannot share cancellation channels across synctest bubbles.
+	// Store boundary and concurrency tests still exercise the selected database engine.
+	t.Setenv("NETBIRD_STORE_ENGINE", "sqlite")
+	return setupDomainTest(t)
+}
+
+func createExpiringDomain(t *testing.T, env *domainTestEnv, name string, expiresAt time.Time) *domain.Domain {
+	t.Helper()
+	d, err := env.store.CreateCustomDomain(context.Background(), accountA, name, testCluster, false)
+	require.NoError(t, err)
+	require.NoError(t, env.store.(*nbstore.SqlStore).GetDB().Model(d).Update("validation_expires_at", expiresAt).Error)
+	d.ValidationExpiresAt = &expiresAt
+	return d
+}
+
+type domainEvents struct {
+	mu     sync.Mutex
+	events []*activity.Event
+}
+
+func captureDomainEvents(env *domainTestEnv) *domainEvents {
+	events := &domainEvents{}
+	env.manager.accountManager = &mock_server.MockAccountManager{
+		StoreEventFunc: func(_ context.Context, initiator, target, account string, code activity.ActivityDescriber, meta map[string]any) {
+			if code == activity.DomainAdded {
+				return
+			}
+			events.mu.Lock()
+			defer events.mu.Unlock()
+			events.events = append(events.events, &activity.Event{
+				InitiatorID: initiator, TargetID: target, AccountID: account,
+				Activity: code.(activity.Activity), Meta: meta,
+			})
+		},
+	}
+	return events
+}
+
+func (e *domainEvents) get() []*activity.Event {
+	e.mu.Lock()
+	defer e.mu.Unlock()
+	return append([]*activity.Event(nil), e.events...)
+}

management/internals/modules/reverseproxy/domain/manager/manager.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net"
 	"strings"
+	"time"
 
 	log "github.com/sirupsen/logrus"
 
@@ -18,6 +19,7 @@ import (
 	"github.com/netbirdio/netbird/management/server/permissions/operations"
 	nbstore "github.com/netbirdio/netbird/management/server/store"
 	"github.com/netbirdio/netbird/management/server/types"
+	nbdomain "github.com/netbirdio/netbird/shared/management/domain"
 	"github.com/netbirdio/netbird/shared/management/status"
 )
 
@@ -26,11 +28,14 @@ type store interface {
 	GetAgentNetworkSettings(ctx context.Context, lockStrength nbstore.LockingStrength, accountID string) (*agentnetworkTypes.Settings, error)
 
 	GetCustomDomain(ctx context.Context, accountID string, domainID string) (*domain.Domain, error)
+	GetCustomDomainByName(ctx context.Context, domainName string) (*domain.Domain, error)
 	ListFreeDomains(ctx context.Context, accountID string) ([]string, error)
 	ListCustomDomains(ctx context.Context, accountID string) ([]*domain.Domain, error)
 	CreateCustomDomain(ctx context.Context, accountID string, domainName string, targetCluster string, validated bool) (*domain.Domain, error)
 	UpdateCustomDomain(ctx context.Context, accountID string, d *domain.Domain) (*domain.Domain, error)
 	DeleteCustomDomain(ctx context.Context, accountID string, domainID string) error
+	GetExpiredCustomDomains(ctx context.Context, now time.Time, afterID domain.ID, limit int) ([]*domain.Domain, error)
+	DeleteExpiredCustomDomain(ctx context.Context, d *domain.Domain, now time.Time) (bool, error)
 }
 
 type proxyManager interface {
@@ -107,12 +112,13 @@ func (m Manager) GetDomains(ctx context.Context, accountID, userID string) ([]*d
 	// Add custom domains.
 	for _, d := range domains {
 		cd := &domain.Domain{
-			ID:            d.ID,
-			Domain:        d.Domain,
-			AccountID:     accountID,
-			TargetCluster: d.TargetCluster,
-			Type:          domain.TypeCustom,
-			Validated:     d.Validated,
+			ID:                  d.ID,
+			Domain:              d.Domain,
+			AccountID:           accountID,
+			TargetCluster:       d.TargetCluster,
+			Type:                domain.TypeCustom,
+			Validated:           d.Validated,
+			ValidationExpiresAt: d.ValidationExpiresAt,
 		}
 		if d.TargetCluster != "" {
 			cd.SupportsCustomPorts = m.proxyManager.ClusterSupportsCustomPorts(ctx, d.TargetCluster)
@@ -128,6 +134,7 @@ func (m Manager) GetDomains(ctx context.Context, accountID, userID string) ([]*d
 	return ret, nil
 }
 
+// CreateDomain registers a normalized custom domain and attempts DNS validation.
 func (m Manager) CreateDomain(ctx context.Context, accountID, userID, domainName, targetCluster string) (*domain.Domain, error) {
 	ok, ctx, err := m.permissionsManager.ValidateUserPermissions(ctx, accountID, userID, modules.Services, operations.Create)
 	if err != nil {
@@ -137,6 +144,15 @@ func (m Manager) CreateDomain(ctx context.Context, accountID, userID, domainName
 		return nil, status.NewPermissionDeniedError()
 	}
 
+	parsed, err := nbdomain.FromString(strings.TrimSuffix(domainName, "."))
+	if err != nil {
+		return nil, status.Errorf(status.InvalidArgument, "invalid domain: %v", err)
+	}
+	domainName = parsed.PunycodeString()
+	if !nbdomain.IsValidDomainNoWildcard(domainName) {
+		return nil, status.Errorf(status.InvalidArgument, "invalid domain format")
+	}
+
 	// Verify the target cluster is in the available clusters for this account
 	allowList, err := m.getClusterAllowList(ctx, accountID)
 	if err != nil {
@@ -153,6 +169,10 @@ func (m Manager) CreateDomain(ctx context.Context, accountID, userID, domainName
 		return nil, fmt.Errorf("target cluster %s is not available", targetCluster)
 	}
 
+	if err := m.checkDomainAvailable(ctx, domainName); err != nil {
+		return nil, err
+	}
+
 	// Attempt an initial validation against the specified cluster only
 	var validated bool
 	if m.validator.IsValid(ctx, domainName, []string{targetCluster}) {
@@ -169,6 +189,23 @@ func (m Manager) CreateDomain(ctx context.Context, accountID, userID, domainName
 	return d, nil
 }
 
+// checkDomainAvailable reports whether the domain is free to claim. The unique
+// index on the column is the real guard; this turns the violation into a
+// conflict the caller can act on instead of a database error, and says nothing
+// about which account holds the domain.
+func (m Manager) checkDomainAvailable(ctx context.Context, domainName string) error {
+	_, err := m.store.GetCustomDomainByName(ctx, domainName)
+	if err == nil {
+		return status.Errorf(status.AlreadyExists, "domain %s is already registered", domainName)
+	}
+
+	if sErr, ok := status.FromError(err); ok && sErr.Type() == status.NotFound {
+		return nil
+	}
+
+	return fmt.Errorf("look up domain: %w", err)
+}
+
 func (m Manager) DeleteDomain(ctx context.Context, accountID, userID, domainID string) error {
 	ok, ctx, err := m.permissionsManager.ValidateUserPermissions(ctx, accountID, userID, modules.Services, operations.Delete)
 	if err != nil {
@@ -206,7 +243,9 @@ func (m Manager) ValidateDomain(ctx context.Context, accountID, userID, domainID
 		log.WithFields(log.Fields{
 			"accountID": accountID,
 			"domainID":  domainID,
-		}).WithError(err).Error("validate domain")
+			"userID":    userID,
+		}).Error("validate domain: permission denied")
+		return
 	}
 
 	log.WithFields(log.Fields{
@@ -222,6 +261,14 @@ func (m Manager) ValidateDomain(ctx context.Context, accountID, userID, domainID
 		}).WithError(err).Error("get custom domain from store")
 		return
 	}
+	if d.Validated {
+		return
+	}
+	if d.ValidationExpiresAt == nil || !time.Now().Before(*d.ValidationExpiresAt) {
+		log.WithFields(log.Fields{"accountID": accountID, "domainID": domainID}).
+			Debug("custom domain validation window has expired")
+		return
+	}
 
 	// Validate only against the domain's target cluster
 	targetCluster := d.TargetCluster
@@ -242,20 +289,21 @@ func (m Manager) ValidateDomain(ctx context.Context, accountID, userID, domainID
 	}).Info("validating domain against target cluster")
 
 	if m.validator.IsValid(context.Background(), d.Domain, []string{targetCluster}) {
-		log.WithFields(log.Fields{
-			"accountID": accountID,
-			"domainID":  domainID,
-			"domain":    d.Domain,
-		}).Info("domain validated successfully")
 		d.Validated = true
 		if _, err := m.store.UpdateCustomDomain(context.Background(), accountID, d); err != nil {
-			log.WithFields(log.Fields{
+			entry := log.WithFields(log.Fields{
 				"accountID": accountID,
 				"domainID":  domainID,
-				"domain":    d.Domain,
-			}).WithError(err).Error("update custom domain in store")
+			}).WithError(err)
+			if sErr, ok := status.FromError(err); ok && sErr.Type() == status.PreconditionFailed {
+				entry.Debug("custom domain registration is no longer pending validation")
+				return
+			}
+			entry.Error("update custom domain in store")
 			return
 		}
+		log.WithFields(log.Fields{"accountID": accountID, "domainID": domainID}).
+			Info("custom domain validated successfully")
 
 		m.accountManager.StoreEvent(context.Background(), userID, domainID, accountID, activity.DomainValidated, d.EventMeta())
 	} else {
@@ -301,9 +349,12 @@ func (m Manager) DeriveClusterFromDomain(ctx context.Context, accountID, domain
 		return "", fmt.Errorf("list custom domains: %w", err)
 	}
 
-	targetCluster, valid := extractClusterFromCustomDomains(domain, customDomains)
-	if valid {
+	targetCluster, match := extractClusterFromCustomDomains(domain, customDomains)
+	switch match {
+	case customDomainValidated:
 		return targetCluster, nil
+	case customDomainUnvalidated:
+		return "", status.Errorf(status.PreconditionFailed, "domain %s is not validated", domain)
 	}
 
 	return "", fmt.Errorf("domain %s does not match any available proxy cluster", domain)
@@ -366,19 +417,46 @@ func (m Manager) reservedGatewayAddress(ctx context.Context, accountID string) (
 	return settings.ProxyAddress, nil
 }
 
-func extractClusterFromCustomDomains(serviceDomain string, customDomains []*domain.Domain) (string, bool) {
+// customDomainMatch describes how a service domain relates to the account's
+// custom domain rows.
+type customDomainMatch int
+
+const (
+	customDomainNoMatch customDomainMatch = iota
+	customDomainUnvalidated
+	customDomainValidated
+)
+
+// extractClusterFromCustomDomains finds the longest custom domain covering the
+// service domain and reports its target cluster. Only a validated row yields a
+// cluster: until the CNAME check has passed the account has not shown it
+// controls the name, so no traffic may be routed for it.
+func extractClusterFromCustomDomains(serviceDomain string, customDomains []*domain.Domain) (string, customDomainMatch) {
 	bestCluster := ""
 	bestLen := -1
+	matched := false
 	for _, cd := range customDomains {
 		if serviceDomain != cd.Domain && !strings.HasSuffix(serviceDomain, "."+cd.Domain) {
 			continue
 		}
+		matched = true
+		if !cd.Validated {
+			continue
+		}
 		if l := len(cd.Domain); l > bestLen {
 			bestLen = l
 			bestCluster = cd.TargetCluster
 		}
 	}
-	return bestCluster, bestLen >= 0
+
+	switch {
+	case bestLen >= 0:
+		return bestCluster, customDomainValidated
+	case matched:
+		return "", customDomainUnvalidated
+	default:
+		return "", customDomainNoMatch
+	}
 }
 
 // ExtractClusterFromFreeDomain extracts the cluster address from a free domain.

management/internals/modules/reverseproxy/domain/manager/manager_realstore_test.go

@@ -0,0 +1,321 @@
+package manager
+
+import (
+	"context"
+	"fmt"
+	"sync"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"go.opentelemetry.io/otel/metric/noop"
+
+	"github.com/netbirdio/netbird/management/internals/modules/reverseproxy/domain"
+	proxymanager "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/proxy/manager"
+	"github.com/netbirdio/netbird/management/server/activity"
+	"github.com/netbirdio/netbird/management/server/mock_server"
+	"github.com/netbirdio/netbird/management/server/permissions"
+	nbstore "github.com/netbirdio/netbird/management/server/store"
+	"github.com/netbirdio/netbird/management/server/types"
+	"github.com/netbirdio/netbird/shared/management/status"
+)
+
+const (
+	testCluster    = "eu.proxy.test"
+	accountA       = "account-a"
+	accountAUser   = "account-a-admin"
+	accountB       = "account-b"
+	accountBUser   = "account-b-admin"
+	accountAMember = "account-a-member"
+)
+
+// stubResolver answers CNAME lookups from a table the test controls, so a
+// domain can point at the cluster or nowhere without touching a real resolver.
+type stubResolver struct {
+	mu     sync.Mutex
+	cnames map[string]string
+}
+
+func (r *stubResolver) LookupCNAME(_ context.Context, host string) (string, error) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+
+	cname, ok := r.cnames[host]
+	if !ok {
+		return "", fmt.Errorf("lookup %s: no such host", host)
+	}
+	return cname + ".", nil
+}
+
+func (r *stubResolver) set(host, cname string) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	r.cnames[host] = cname
+}
+
+type domainTestEnv struct {
+	manager  Manager
+	store    nbstore.Store
+	resolver *stubResolver
+}
+
+// setupDomainTest builds the domain manager on a real SQLite store with two
+// accounts and one active public proxy cluster.
+func setupDomainTest(t *testing.T) *domainTestEnv {
+	t.Helper()
+
+	ctx := context.Background()
+	testStore, cleanup, err := nbstore.NewTestStoreFromSQL(ctx, "", t.TempDir())
+	require.NoError(t, err)
+	t.Cleanup(cleanup)
+
+	for accountID, userID := range map[string]string{accountA: accountAUser, accountB: accountBUser} {
+		users := map[string]*types.User{
+			userID: {
+				Id:        userID,
+				AccountID: accountID,
+				Role:      types.UserRoleAdmin,
+			},
+		}
+		if accountID == accountA {
+			// A real member of the account whose role denies Services:Create, so
+			// permission denial is exercised as ok=false rather than as a lookup
+			// error for a user who is not in the account at all.
+			users[accountAMember] = &types.User{
+				Id:        accountAMember,
+				AccountID: accountID,
+				Role:      types.UserRoleUser,
+			}
+		}
+
+		require.NoError(t, testStore.SaveAccount(ctx, &types.Account{
+			Id:        accountID,
+			CreatedBy: userID,
+			Settings:  &types.Settings{},
+			Users:     users,
+		}))
+	}
+
+	proxyMgr, err := proxymanager.NewManager(testStore, noop.NewMeterProvider().Meter(""))
+	require.NoError(t, err)
+
+	_, err = proxyMgr.Connect(ctx, "proxy-1", "session-1", testCluster, "127.0.0.1", nil, nil)
+	require.NoError(t, err)
+
+	resolver := &stubResolver{cnames: make(map[string]string)}
+
+	mgr := Manager{
+		store:              testStore,
+		proxyManager:       proxyMgr,
+		validator:          domain.Validator{Resolver: resolver},
+		permissionsManager: permissions.NewManager(testStore),
+		accountManager: &mock_server.MockAccountManager{
+			StoreEventFunc: func(context.Context, string, string, string, activity.ActivityDescriber, map[string]any) {},
+		},
+	}
+
+	return &domainTestEnv{manager: mgr, store: testStore, resolver: resolver}
+}
+
+// storedDomain reads a domain row back through the store so assertions are made
+// on what was persisted rather than on the value the manager returned.
+func storedDomain(t *testing.T, s nbstore.Store, accountID, domainName string) *domain.Domain {
+	t.Helper()
+
+	domains, err := s.ListCustomDomains(context.Background(), accountID)
+	require.NoError(t, err)
+	for _, d := range domains {
+		if d.Domain == domainName {
+			return d
+		}
+	}
+	return nil
+}
+
+// A domain whose CNAME check fails is stored unvalidated and must not resolve a
+// cluster, which is what service creation gates on.
+func TestCreateDomain_FailedLookupIsNotServable(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+
+	created, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "apps.example.com", testCluster)
+	require.NoError(t, err)
+	assert.False(t, created.Validated, "a domain whose CNAME lookup fails must not be created validated")
+
+	stored := storedDomain(t, env.store, accountA, "apps.example.com")
+	require.NotNil(t, stored, "domain row should exist")
+	assert.False(t, stored.Validated, "persisted row must be unvalidated")
+
+	cluster, err := env.manager.DeriveClusterFromDomain(ctx, accountA, "apps.example.com")
+	require.Error(t, err, "an unvalidated domain must not resolve a cluster")
+	assert.Empty(t, cluster)
+	assert.Contains(t, err.Error(), "not validated", "error should tell the caller what to fix")
+
+	sErr, ok := status.FromError(err)
+	require.True(t, ok, "error should be a typed status error")
+	assert.Equal(t, status.PreconditionFailed, sErr.Type())
+
+	_, err = env.manager.DeriveClusterFromDomain(ctx, accountA, "sub.apps.example.com")
+	assert.Error(t, err, "subdomains of an unvalidated custom domain are not servable either")
+}
+
+// A second account claiming a registered domain gets a clean conflict, not a
+// database error surfaced as a 500.
+func TestCreateDomain_DuplicateIsAConflict(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+
+	_, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "shared.example.com", testCluster)
+	require.NoError(t, err)
+
+	_, err = env.manager.CreateDomain(ctx, accountB, accountBUser, "shared.example.com", testCluster)
+	require.Error(t, err)
+
+	sErr, ok := status.FromError(err)
+	require.True(t, ok, "conflict must be a typed status error, not a raw database error")
+	assert.Equal(t, status.AlreadyExists, sErr.Type(), "conflict should map to 409, not 500")
+	assert.NotContains(t, sErr.Message, accountA, "the response must not reveal the holding account")
+
+	assert.Nil(t, storedDomain(t, env.store, accountB, "shared.example.com"), "no row should be written on conflict")
+}
+
+// The same account re-adding one of its own domains is a conflict too.
+func TestCreateDomain_SameAccountDuplicateIsAConflict(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+
+	_, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "dup.example.com", testCluster)
+	require.NoError(t, err)
+
+	_, err = env.manager.CreateDomain(ctx, accountA, accountAUser, "dup.example.com", testCluster)
+	require.Error(t, err)
+
+	sErr, ok := status.FromError(err)
+	require.True(t, ok)
+	assert.Equal(t, status.AlreadyExists, sErr.Type())
+}
+
+// The negative control: a validated domain still derives its cluster, for the
+// bare name and for subdomains, exactly as before.
+func TestCreateDomain_ValidatedDomainDerivesCluster(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+	env.resolver.set("validation.valid.example.com", testCluster)
+
+	created, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "valid.example.com", testCluster)
+	require.NoError(t, err)
+	require.True(t, created.Validated, "a matching CNAME should validate on create")
+
+	cluster, err := env.manager.DeriveClusterFromDomain(ctx, accountA, "valid.example.com")
+	require.NoError(t, err)
+	assert.Equal(t, testCluster, cluster)
+
+	cluster, err = env.manager.DeriveClusterFromDomain(ctx, accountA, "app.valid.example.com")
+	require.NoError(t, err)
+	assert.Equal(t, testCluster, cluster, "subdomains of a validated custom domain resolve too")
+}
+
+// Validating a domain flips the gate: the same lookup that failed before now
+// resolves a cluster.
+func TestValidateDomain_UnlocksClusterDerivation(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+
+	created, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "later.example.com", testCluster)
+	require.NoError(t, err)
+	require.False(t, created.Validated)
+
+	_, err = env.manager.DeriveClusterFromDomain(ctx, accountA, "later.example.com")
+	require.Error(t, err)
+
+	env.resolver.set("validation.later.example.com", testCluster)
+	env.manager.ValidateDomain(ctx, accountA, accountAUser, created.ID)
+
+	require.True(t, storedDomain(t, env.store, accountA, "later.example.com").Validated)
+
+	cluster, err := env.manager.DeriveClusterFromDomain(ctx, accountA, "later.example.com")
+	require.NoError(t, err)
+	assert.Equal(t, testCluster, cluster)
+}
+
+// Free cluster domains are unaffected by the custom domain gate.
+func TestDeriveClusterFromDomain_FreeDomainUnaffected(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+
+	cluster, err := env.manager.DeriveClusterFromDomain(ctx, accountA, "myapp.abc123."+testCluster)
+	require.NoError(t, err)
+	assert.Equal(t, testCluster, cluster)
+}
+
+// The manager pre-check exists to turn a conflict into a 409, but the unique
+// index on the column is what actually guarantees the domain is claimed once.
+//
+// Two requests can clear the pre-check concurrently and race to the insert.
+// Inserting twice through the store reaches the same code path the loser of
+// that race takes, without the nondeterminism of driving it from goroutines,
+// and the loser must still see a conflict rather than an internal error.
+func TestStore_DuplicateDomainRejectedByIndexAsConflict(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+
+	_, err := env.store.CreateCustomDomain(ctx, accountA, "indexed.example.com", testCluster, false)
+	require.NoError(t, err)
+
+	_, err = env.store.CreateCustomDomain(ctx, accountB, "indexed.example.com", testCluster, false)
+	require.Error(t, err, "the unique index must reject the same domain in a second account")
+
+	sErr, ok := status.FromError(err)
+	require.True(t, ok, "the losing insert must return a typed status error")
+	assert.Equal(t, status.AlreadyExists, sErr.Type(), "a lost race is a 409, not a 500")
+}
+
+// Validation is what decides whether a domain routes traffic, so a caller
+// without permission to it must not be able to flip the flag. The check logged
+// the denial and then carried on, which was inert while nothing read Validated
+// and is not once cluster derivation gates on it.
+func TestValidateDomain_PermissionDeniedDoesNotValidate(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+
+	created, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "guarded.example.com", testCluster)
+	require.NoError(t, err)
+	require.False(t, created.Validated)
+
+	// The CNAME is in place, so the only thing standing between this caller and
+	// a validated domain is the permission check.
+	env.resolver.set("validation.guarded.example.com", testCluster)
+
+	env.manager.ValidateDomain(ctx, accountA, accountAMember, created.ID)
+
+	stored := storedDomain(t, env.store, accountA, "guarded.example.com")
+	require.NotNil(t, stored)
+	assert.False(t, stored.Validated, "a caller without permission must not validate the domain")
+
+	_, err = env.manager.DeriveClusterFromDomain(ctx, accountA, "guarded.example.com")
+	assert.Error(t, err, "the domain must still be unservable")
+}
+
+// A validation finishing after deletion must reject the stale write, without
+// restoring the registration or reporting successful validation.
+func TestUpdateCustomDomain_DoesNotResurrectDeletedDomain(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+
+	created, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "racy.example.com", testCluster)
+	require.NoError(t, err)
+
+	stale := storedDomain(t, env.store, accountA, "racy.example.com")
+	require.NotNil(t, stale)
+
+	require.NoError(t, env.manager.DeleteDomain(ctx, accountA, accountAUser, created.ID))
+	require.Nil(t, storedDomain(t, env.store, accountA, "racy.example.com"), "the domain should be gone")
+
+	// What an in-flight validation would write once its CNAME check succeeded.
+	stale.Validated = true
+	_, err = env.store.UpdateCustomDomain(ctx, accountA, stale)
+	require.Error(t, err, "a deleted registration must reject a late validation")
+
+	assert.Nil(t, storedDomain(t, env.store, accountA, "racy.example.com"),
+		"a late validation write must not recreate a deleted domain")
+}

management/internals/modules/reverseproxy/domain/manager/manager_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"testing"
+	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -188,6 +189,10 @@ func (s *stubStore) GetCustomDomain(context.Context, string, string) (*domain.Do
 	panic("not used in allow-list tests")
 }
 
+func (s *stubStore) GetCustomDomainByName(context.Context, string) (*domain.Domain, error) {
+	panic("not used in allow-list tests")
+}
+
 func (s *stubStore) ListFreeDomains(context.Context, string) ([]string, error) {
 	panic("not used in allow-list tests")
 }
@@ -208,6 +213,14 @@ func (s *stubStore) DeleteCustomDomain(context.Context, string, string) error {
 	panic("not used in allow-list tests")
 }
 
+func (s *stubStore) GetExpiredCustomDomains(context.Context, time.Time, domain.ID, int) ([]*domain.Domain, error) {
+	panic("not used in allow-list tests")
+}
+
+func (s *stubStore) DeleteExpiredCustomDomain(context.Context, *domain.Domain, time.Time) (bool, error) {
+	panic("not used in allow-list tests")
+}
+
 // TestGetClusterAllowList_DedicatedGatewayAddressExcluded pins invariant (B)'s
 // chokepoint: a self-addressed settings pin reserves the account's gateway
 // address, so it is dropped from the allow list — which, because the

management/internals/modules/reverseproxy/domain/manager/normalization_test.go

@@ -0,0 +1,83 @@
+package manager
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/shared/management/status"
+)
+
+func TestCreateDomain_NormalizesName(t *testing.T) {
+	for _, tt := range []struct {
+		name      string
+		input     string
+		canonical string
+	}{
+		{"mixed case", "Apps.Example.COM", "apps.example.com"},
+		{"unicode", "münchen.example.com", "xn--mnchen-3ya.example.com"},
+		{"trailing dot", "apps.example.com.", "apps.example.com"},
+		{"underscore", "My_App.example.com", "my_app.example.com"},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			ctx := context.Background()
+			env := setupDomainTest(t)
+			env.resolver.set("validation."+tt.canonical, testCluster)
+
+			created, err := env.manager.CreateDomain(ctx, accountA, accountAUser, tt.input, testCluster)
+			require.NoError(t, err)
+			assert.Equal(t, tt.canonical, created.Domain, "the response must use the normalized name")
+			assert.True(t, created.Validated, "the CNAME lookup must use the normalized name")
+			stored, err := env.store.GetCustomDomain(ctx, accountA, created.ID)
+			require.NoError(t, err)
+			assert.Equal(t, tt.canonical, stored.Domain, "the database must retain the normalized name")
+
+			_, err = env.manager.CreateDomain(ctx, accountB, accountBUser, tt.canonical, testCluster)
+			require.Error(t, err)
+			sErr, ok := status.FromError(err)
+			require.True(t, ok, "an equivalent name must return a typed conflict")
+			assert.Equal(t, status.AlreadyExists, sErr.Type(), "normalization must precede the availability check")
+		})
+	}
+}
+
+func TestCreateDomain_NormalizedNameCanValidateLater(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+	created, err := env.manager.CreateDomain(ctx, accountA, accountAUser, "Apps.Example.COM.", testCluster)
+	require.NoError(t, err)
+	require.False(t, created.Validated, "a missing CNAME must leave the normalized registration pending")
+
+	env.resolver.set("validation.apps.example.com", testCluster)
+	env.manager.ValidateDomain(ctx, accountA, accountAUser, created.ID)
+	stored, err := env.store.GetCustomDomain(ctx, accountA, created.ID)
+	require.NoError(t, err)
+	assert.Equal(t, "apps.example.com", stored.Domain, "retrying validation must retain the normalized name")
+	assert.True(t, stored.Validated, "later validation must look up the normalized name")
+}
+
+func TestCreateDomain_RejectsInvalidName(t *testing.T) {
+	ctx := context.Background()
+	env := setupDomainTest(t)
+	for _, name := range []string{
+		"", ".", "app..example.com", "app.example.com..", "-app.example.com",
+		"app%.example.com", "app!.example.com", "*.example.com", "app example.com",
+		"https://example.com", strings.Repeat("a", 64) + ".example.com",
+	} {
+		t.Run(name, func(t *testing.T) {
+			// A matching DNS response must not make a malformed name acceptable.
+			env.resolver.set("validation."+name, testCluster)
+			_, err := env.manager.CreateDomain(ctx, accountA, accountAUser, name, testCluster)
+			require.Error(t, err)
+			sErr, ok := status.FromError(err)
+			require.True(t, ok, "invalid names must return a typed client error")
+			assert.Equal(t, status.InvalidArgument, sErr.Type(), "malformed names must be rejected before storage")
+		})
+	}
+	stored, err := env.store.ListCustomDomains(ctx, accountA)
+	require.NoError(t, err)
+	assert.Empty(t, stored, "invalid registration attempts must not reserve any names")
+}

management/internals/modules/reverseproxy/service/manager/domain_validation_test.go

@@ -0,0 +1,127 @@
+package manager
+
+import (
+	"context"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"go.opentelemetry.io/otel/metric/noop"
+
+	domainmanager "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/domain/manager"
+	proxymanager "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/proxy/manager"
+	rpservice "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/service"
+	"github.com/netbirdio/netbird/management/server/activity"
+	"github.com/netbirdio/netbird/management/server/mock_server"
+	"github.com/netbirdio/netbird/management/server/permissions"
+	"github.com/netbirdio/netbird/management/server/store"
+	"github.com/netbirdio/netbird/shared/management/status"
+)
+
+const validationTestCluster = "eu.proxy.test"
+
+// withRealDomainManager swaps the stub cluster deriver for the real domain
+// manager backed by the same store, so service creation is gated by the actual
+// domain rows rather than by a test double that always agrees.
+func withRealDomainManager(t *testing.T, mgr *Manager, testStore store.Store) {
+	t.Helper()
+
+	ctx := context.Background()
+	proxyMgr, err := proxymanager.NewManager(testStore, noop.NewMeterProvider().Meter(""))
+	require.NoError(t, err)
+
+	_, err = proxyMgr.Connect(ctx, "proxy-1", "session-1", validationTestCluster, "127.0.0.1", nil, nil)
+	require.NoError(t, err)
+
+	accountMgr := &mock_server.MockAccountManager{
+		StoreEventFunc: func(context.Context, string, string, string, activity.ActivityDescriber, map[string]any) {},
+	}
+	mgr.clusterDeriver = domainmanager.NewManager(testStore, proxyMgr, permissions.NewManager(testStore), accountMgr)
+}
+
+func newTestService(domain string) *rpservice.Service {
+	return &rpservice.Service{
+		Name:    "test-service",
+		Domain:  domain,
+		Enabled: true,
+		Mode:    rpservice.ModeHTTP,
+		Targets: []*rpservice.Target{{
+			Host:       "10.0.0.1",
+			Port:       8080,
+			Protocol:   "http",
+			TargetId:   testPeerID,
+			TargetType: "peer",
+			Enabled:    true,
+		}},
+	}
+}
+
+// A service must not bind to a domain the account has not validated, and
+// nothing may be persisted for the attempt.
+func TestCreateService_RefusesUnvalidatedDomain(t *testing.T) {
+	ctx := context.Background()
+	mgr, testStore := setupIntegrationTest(t)
+	withRealDomainManager(t, mgr, testStore)
+
+	_, err := testStore.CreateCustomDomain(ctx, testAccountID, "unproven.example.com", validationTestCluster, false)
+	require.NoError(t, err)
+
+	_, err = mgr.CreateService(ctx, testAccountID, testUserID, newTestService("unproven.example.com"))
+	require.Error(t, err, "an unvalidated domain must not bind a service")
+	assert.Contains(t, err.Error(), "not validated", "the API error should name the actual problem")
+
+	sErr, ok := status.FromError(err)
+	require.True(t, ok, "error should be a typed status error")
+	assert.Equal(t, status.PreconditionFailed, sErr.Type())
+
+	services, err := testStore.GetAccountServices(ctx, store.LockingStrengthNone, testAccountID)
+	require.NoError(t, err)
+	assert.Empty(t, services, "no service row should be written for a refused domain")
+}
+
+// The negative control: a validated domain still binds a service and derives
+// its cluster exactly as before.
+func TestCreateService_ValidatedDomainBindsService(t *testing.T) {
+	ctx := context.Background()
+	mgr, testStore := setupIntegrationTest(t)
+	withRealDomainManager(t, mgr, testStore)
+
+	_, err := testStore.CreateCustomDomain(ctx, testAccountID, "proven.example.com", validationTestCluster, true)
+	require.NoError(t, err)
+
+	created, err := mgr.CreateService(ctx, testAccountID, testUserID, newTestService("app.proven.example.com"))
+	require.NoError(t, err)
+	assert.Equal(t, validationTestCluster, created.ProxyCluster, "service should bind to the domain's target cluster")
+
+	services, err := testStore.GetAccountServices(ctx, store.LockingStrengthNone, testAccountID)
+	require.NoError(t, err)
+	require.Len(t, services, 1, "the service should be persisted")
+	assert.Equal(t, "app.proven.example.com", services[0].Domain)
+}
+
+// An update must not be a way around the creation gate: moving a live service
+// onto an unvalidated domain has to fail rather than silently keep the old
+// cluster and start serving the new hostname.
+func TestUpdateService_RefusesMoveToUnvalidatedDomain(t *testing.T) {
+	ctx := context.Background()
+	mgr, testStore := setupIntegrationTest(t)
+	withRealDomainManager(t, mgr, testStore)
+
+	_, err := testStore.CreateCustomDomain(ctx, testAccountID, "proven.example.com", validationTestCluster, true)
+	require.NoError(t, err)
+	_, err = testStore.CreateCustomDomain(ctx, testAccountID, "unproven.example.com", validationTestCluster, false)
+	require.NoError(t, err)
+
+	created, err := mgr.CreateService(ctx, testAccountID, testUserID, newTestService("app.proven.example.com"))
+	require.NoError(t, err)
+
+	moved := *created
+	moved.Domain = "app.unproven.example.com"
+	_, err = mgr.UpdateService(ctx, testAccountID, testUserID, &moved)
+	require.Error(t, err, "moving to an unvalidated domain must fail")
+	assert.Contains(t, err.Error(), "not validated")
+
+	stored, err := testStore.GetServiceByID(ctx, store.LockingStrengthNone, testAccountID, created.ID)
+	require.NoError(t, err)
+	assert.Equal(t, "app.proven.example.com", stored.Domain, "the service must keep its original domain")
+}

management/internals/modules/reverseproxy/service/manager/manager.go

@@ -608,16 +608,19 @@ func (m *Manager) resolveEffectiveCluster(ctx context.Context, accountID string,
 		return existing.ProxyCluster, nil
 	}
 
-	if m.clusterDeriver != nil {
-		derived, err := m.clusterDeriver.DeriveClusterFromDomain(ctx, accountID, svc.Domain)
-		if err != nil {
-			log.WithError(err).Warnf("could not derive cluster from domain %s", svc.Domain)
-		} else {
-			return derived, nil
-		}
+	if m.clusterDeriver == nil {
+		return existing.ProxyCluster, nil
 	}
 
-	return existing.ProxyCluster, nil
+	// Falling back to the old cluster here would let an update move a service
+	// onto a domain the account has not validated, bypassing the check that
+	// creation makes.
+	derived, err := m.clusterDeriver.DeriveClusterFromDomain(ctx, accountID, svc.Domain)
+	if err != nil {
+		return "", status.Errorf(status.PreconditionFailed, "could not derive cluster from domain %s: %v", svc.Domain, err)
+	}
+
+	return derived, nil
 }
 
 func (m *Manager) executeServiceUpdate(ctx context.Context, transaction store.Store, accountID string, service *service.Service, updateInfo *serviceUpdateInfo, customPorts *bool, effectiveCluster string) error {

management/internals/modules/reverseproxy/service/manager/manager_test.go

@@ -7,11 +7,10 @@ import (
 	"testing"
 	"time"
 
-	cachestore "github.com/eko/gocache/lib/v4/store"
-	"go.uber.org/mock/gomock"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"go.opentelemetry.io/otel/metric/noop"
+	"go.uber.org/mock/gomock"
 
 	"github.com/netbirdio/netbird/management/internals/modules/reverseproxy/proxy"
 	proxymanager "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/proxy/manager"
@@ -31,7 +30,7 @@ import (
 	"github.com/netbirdio/netbird/shared/management/status"
 )
 
-func testCacheStore(t *testing.T) cachestore.StoreInterface {
+func testCacheStore(t *testing.T) nbcache.Store {
 	t.Helper()
 	s, err := nbcache.NewStore(context.Background(), 30*time.Minute, 10*time.Minute, 100)
 	require.NoError(t, err)
@@ -295,6 +294,7 @@ func TestPersistNewService(t *testing.T) {
 		assert.Equal(t, status.AlreadyExists, sErr.Type())
 	})
 }
+
 func TestPreserveExistingAuthSecrets(t *testing.T) {
 	mgr := &Manager{}
 

management/internals/modules/reverseproxy/service/service.go

@@ -55,6 +55,8 @@ const (
 	SourceEphemeral = "ephemeral"
 )
 
+var ErrUnsupportedIPAddressUpstreamHost = errors.New("unsupported ip address for a direct upstream host")
+
 type TargetOptions struct {
 	SkipTLSVerify      bool              `json:"skip_tls_verify"`
 	RequestTimeout     time.Duration     `json:"request_timeout,omitempty"`
@@ -406,6 +408,7 @@ func (s *Service) ToProtoMapping(operation Operation, authToken string, oidcConf
 
 	if s.Auth.BearerAuth != nil && s.Auth.BearerAuth.Enabled {
 		auth.Oidc = true
+		auth.AllowedGroupIds = append([]string(nil), s.Auth.BearerAuth.DistributionGroups...)
 	}
 
 	for _, h := range s.Auth.HeaderAuths {
@@ -1002,8 +1005,8 @@ func (s *Service) validateHTTPTargets() error {
 				return err
 			}
 		case TargetTypeSubnet:
-			if target.Host == "" {
-				return fmt.Errorf("target %d has empty host but target_type is %q", i, target.TargetType)
+			if err := validateSubnetTarget(i, target); err != nil {
+				return err
 			}
 		case TargetTypeCluster:
 			if err := validateClusterTarget(i, target); err != nil {
@@ -1026,6 +1029,34 @@ func (s *Service) validateHTTPTargets() error {
 	return nil
 }
 
+func validateSubnetTarget(idx int, target *Target) error {
+	host := strings.TrimSpace(target.Host)
+	if host == "" {
+		return fmt.Errorf("target %d has empty host but target_type is %q", idx, target.TargetType)
+	}
+	if strings.ContainsAny(host, " \t/") {
+		return fmt.Errorf("target %d: host %q contains invalid characters", idx, host)
+	}
+	if _, _, err := net.SplitHostPort(host); err == nil {
+		return fmt.Errorf("target %d: host %q must not include a port (set target.port instead)", idx, host)
+	}
+	noBrackets := strings.TrimSuffix(strings.TrimPrefix(host, "["), "]")
+	maybeip, err := netip.ParseAddr(noBrackets)
+	if err != nil { // not an ip
+		return nil //nolint:nilerr
+	}
+	if maybeip.Zone() != "" {
+		return fmt.Errorf("invalid direct upstream host ip %s %w", maybeip.String(), ErrUnsupportedIPAddressUpstreamHost)
+	}
+	if !target.Options.DirectUpstream {
+		return nil
+	}
+	if maybeip.IsLoopback() || maybeip.IsMulticast() || maybeip.IsLinkLocalUnicast() {
+		return fmt.Errorf("invalid direct upstream host ip %s %w", maybeip.String(), ErrUnsupportedIPAddressUpstreamHost)
+	}
+	return nil
+}
+
 // validateClusterTarget cluster targets should not have empty hosts and should have direct upstream enabled.
 func validateClusterTarget(idx int, target *Target) error {
 	host := strings.TrimSpace(target.Host)
@@ -1060,6 +1091,15 @@ func validateDirectUpstreamHost(idx int, target *Target) error {
 	if _, _, err := net.SplitHostPort(host); err == nil {
 		return fmt.Errorf("target %d: host %q must not include a port (set target.port instead)", idx, host)
 	}
+	noBrackets := strings.TrimSuffix(strings.TrimPrefix(host, "["), "]")
+	maybeip, err := netip.ParseAddr(noBrackets)
+	if err != nil { // not an ip
+		return nil //nolint:nilerr
+	}
+	if maybeip.Zone() != "" || maybeip.IsLoopback() || maybeip.IsMulticast() || maybeip.IsLinkLocalUnicast() {
+		return fmt.Errorf("invalid direct upstream host ip %s %w", maybeip.String(), ErrUnsupportedIPAddressUpstreamHost)
+	}
+
 	return nil
 }
 

management/internals/modules/reverseproxy/service/service_test.go

@@ -227,6 +227,64 @@ func TestValidateTargetOptions_CustomHeaders(t *testing.T) {
 	})
 }
 
+func TestValidate_DirectUpstreamHost(t *testing.T) {
+	target := Target{TargetId: "id-1", TargetType: TargetTypePeer, Host: "10.0.0.1", Port: 80, Protocol: "http", Enabled: true, Options: TargetOptions{DirectUpstream: true}}
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "127.0.0.2")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.NotNil(t, validateDirectUpstreamHost(0, targetWithHost(&target, "127.0.0.2:80")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "::1")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "::1%lo0")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "[::1]")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.NotNil(t, validateDirectUpstreamHost(0, targetWithHost(&target, "[::1]:80")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "[::1%lo0]")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.NotNil(t, validateDirectUpstreamHost(0, targetWithHost(&target, "[::1%lo0]:80")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "169.254.100.100")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "fe80::1")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "[fe80::1]")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "224.100.100.100")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "ff00::ffff")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateDirectUpstreamHost(0, targetWithHost(&target, "[ff00::ffff]")), ErrUnsupportedIPAddressUpstreamHost)
+
+	// empty host
+	assert.Nil(t, validateDirectUpstreamHost(0, &Target{Options: TargetOptions{DirectUpstream: true}, Host: "   "}))
+	// host with a space
+	assert.NotNil(t, validateDirectUpstreamHost(0, &Target{Options: TargetOptions{DirectUpstream: true}, Host: "with space"}))
+	// host with a tab
+	assert.NotNil(t, validateDirectUpstreamHost(0, &Target{Options: TargetOptions{DirectUpstream: true}, Host: "with\ttab"}))
+	// host with a slash
+	assert.NotNil(t, validateDirectUpstreamHost(0, &Target{Options: TargetOptions{DirectUpstream: true}, Host: "with/slash"}))
+}
+
+func TestValidate_ValidateSubnetTarget(t *testing.T) {
+	target := Target{TargetId: "id-1", TargetType: TargetTypeSubnet, Host: "10.0.0.1", Port: 80, Protocol: "http", Enabled: true, Options: TargetOptions{DirectUpstream: true}}
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "127.0.0.2")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.NotNil(t, validateSubnetTarget(0, targetWithHost(&target, "127.0.0.2:80")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "::1")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.NotNil(t, validateSubnetTarget(0, targetWithHost(&target, "[::1]:80")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "::1%lo0")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "[::1%lo0]")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.NotNil(t, validateSubnetTarget(0, targetWithHost(&target, "[::1%lo0]:80")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "169.254.100.100")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "fe80::1")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "[fe80::1]")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "224.100.100.100")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "ff00::ffff")), ErrUnsupportedIPAddressUpstreamHost)
+	assert.ErrorIs(t, validateSubnetTarget(0, targetWithHost(&target, "[ff00::ffff]")), ErrUnsupportedIPAddressUpstreamHost)
+
+	// empty host
+	assert.NotNil(t, validateSubnetTarget(0, &Target{Options: TargetOptions{DirectUpstream: true}, Host: "   "}))
+	// host with a space
+	assert.NotNil(t, validateSubnetTarget(0, &Target{Options: TargetOptions{DirectUpstream: true}, Host: "with space"}))
+	// host with a tab
+	assert.NotNil(t, validateSubnetTarget(0, &Target{Options: TargetOptions{DirectUpstream: true}, Host: "with\ttab"}))
+	// host with a slash
+	assert.NotNil(t, validateSubnetTarget(0, &Target{Options: TargetOptions{DirectUpstream: true}, Host: "with/slash"}))
+}
+
+func targetWithHost(t *Target, host string) *Target {
+	t.Host = host
+	return t
+}
+
 func TestToProtoMapping_TargetOptions(t *testing.T) {
 	rp := &Service{
 		ID:        "svc-1",
@@ -261,6 +319,44 @@ func TestToProtoMapping_TargetOptions(t *testing.T) {
 	assert.Equal(t, int64(30), opts.RequestTimeout.Seconds)
 }
 
+// TestToProtoMapping_AllowedGroupIds covers the list the proxy gates session
+// cookies on: without it the proxy can only check a cookie's signature, which
+// makes a token minted for a user outside the groups a bearer credential.
+func TestToProtoMapping_AllowedGroupIds(t *testing.T) {
+	t.Run("distribution groups reach the proxy", func(t *testing.T) {
+		rp := &Service{
+			ID:        "svc-1",
+			AccountID: "acc-1",
+			Domain:    "example.com",
+			Auth: AuthConfig{
+				BearerAuth: &BearerAuthConfig{
+					Enabled:            true,
+					DistributionGroups: []string{"grp-1", "grp-2"},
+				},
+			},
+		}
+		pm := rp.ToProtoMapping(Create, "token", proxy.OIDCValidationConfig{})
+
+		assert.True(t, pm.GetAuth().GetOidc())
+		assert.Equal(t, []string{"grp-1", "grp-2"}, pm.GetAuth().GetAllowedGroupIds())
+	})
+
+	t.Run("a service open to the account carries no groups", func(t *testing.T) {
+		rp := &Service{
+			ID:        "svc-1",
+			AccountID: "acc-1",
+			Domain:    "example.com",
+			Auth: AuthConfig{
+				BearerAuth: &BearerAuthConfig{Enabled: true},
+			},
+		}
+		pm := rp.ToProtoMapping(Create, "token", proxy.OIDCValidationConfig{})
+
+		assert.True(t, pm.GetAuth().GetOidc())
+		assert.Empty(t, pm.GetAuth().GetAllowedGroupIds(), "an empty list must not restrict access")
+	})
+}
+
 func TestToProtoMapping_NoOptionsWhenDefault(t *testing.T) {
 	rp := &Service{
 		ID:        "svc-1",