diff --git a/Documentation/configuration.md b/Documentation/configuration.md
index a5ea6637f0..2a59019640 100644
--- a/Documentation/configuration.md
+++ b/Documentation/configuration.md
@@ -19,6 +19,9 @@ The value of the config is a JSON dictionary with the following keys:
 * `EnableIPv6` (bool): Enables ipv6 support
   Defaults to `false`
 
+* `EnableNFTables` (bool): (EXPERIMENTAL) If set to true, flannel uses nftables instead of iptables to masquerade the traffic.
+   Default to `false`
+
 * `SubnetLen` (integer): The size of the subnet allocated to each host.
    Defaults to 24 (i.e. /24) unless `Network` was configured to be smaller than a /22 in which case it is two less than the network.
 
@@ -128,3 +131,22 @@ FLANNEL_IPMASQ=true
 ## IPv6 only
 
 To use an IPv6-only environment use the same configuration of the Dual-stack section to enable IPv6 and add "EnableIPv4": false in the net-conf.json of the kube-flannel-cfg ConfigMap. In case of IPv6-only setup, please use the docker.io IPv6-only endpoint as described in the following link: https://www.docker.com/blog/beta-ipv6-support-on-docker-hub-registry/
+
+## nftables mode
+To enable `nftables` mode in flannel, set `EnableNFTables` to true in flannel configuration.
+
+Note: to test with kube-proxy, use kubeadm with the following configuration:
+```yaml
+apiVersion: kubeadm.k8s.io/v1beta3
+kind: ClusterConfiguration
+kubernetesVersion: v1.29.0
+controllerManager:
+  extraArgs:
+    feature-gates: NFTablesProxyMode=true
+---
+apiVersion: kubeproxy.config.k8s.io/v1alpha1
+kind: KubeProxyConfiguration
+mode: "nftables"
+featureGates:
+  NFTablesProxyMode: true
+```
diff --git a/Documentation/kube-flannel.yml b/Documentation/kube-flannel.yml
index 9f7b2eb531..2bfe29d479 100644
--- a/Documentation/kube-flannel.yml
+++ b/Documentation/kube-flannel.yml
@@ -98,6 +98,7 @@ data:
   net-conf.json: |
     {
       "Network": "10.244.0.0/16",
+      "EnableNFTables": false,
       "Backend": {
         "Type": "vxlan"
       }
diff --git a/Makefile b/Makefile
index eaf698a58a..6ef7823fcf 100644
--- a/Makefile
+++ b/Makefile
@@ -6,7 +6,7 @@ QEMU_VERSION=v3.0.0
 BASH_UNIT_VERSION=v2.3.0
 
 # Default tag and architecture. Can be overridden
-TAG?=$(shell git describe --tags --always)
+TAG?=$(shell git describe --tags --dirty --always)
 ARCH?=amd64
 # Only enable CGO (and build the UDP backend) on AMD64
 ifeq ($(ARCH),amd64)
diff --git a/e2e/Dockerfile b/e2e/Dockerfile
index 41f6bf6916..09cf9bba6c 100644
--- a/e2e/Dockerfile
+++ b/e2e/Dockerfile
@@ -11,6 +11,7 @@ RUN set -x \
     curl \
     tar gzip\
     iptables \
+    nftables \
     iproute2 \
     iputils \
  && if [ "${ARCH?required}" != "amd64" ]; then \
diff --git a/e2e/run-e2e-tests.sh b/e2e/run-e2e-tests.sh
index d23dfa3bbe..b93eb8f621 100644
--- a/e2e/run-e2e-tests.sh
+++ b/e2e/run-e2e-tests.sh
@@ -38,11 +38,12 @@ EOF
 
 write-flannel-conf(){
     local backend=$1
+    local enable_nftables=$2
     cp ../Documentation/kube-flannel.yml ./kube-flannel.yml
     yq -i 'select(.kind == "DaemonSet").spec.template.spec.containers[0].image |= strenv(FLANNEL_IMAGE)' ./kube-flannel.yml
     yq -i 'select(.kind == "DaemonSet").spec.template.spec.initContainers[1].image |= strenv(FLANNEL_IMAGE)' ./kube-flannel.yml
 
-    export flannel_conf="{ \"Network\": \"$FLANNEL_NET\", \"Backend\": { \"Type\": \"${backend}\" } }"
+    export flannel_conf="{ \"Network\": \"$FLANNEL_NET\", \"Backend\": { \"Type\": \"${backend}\" }, \"EnableNFTables\": ${enable_nftables} }"
 
     yq -i 'select(.metadata.name == "kube-flannel-cfg").data."net-conf.json" |= strenv(flannel_conf)' ./kube-flannel.yml
 
@@ -55,10 +56,11 @@ write-flannel-conf(){
 # This is not used at the moment since github runners don't support dual-stack networking
 write-flannel-conf-dual-stack(){
     local backend=$1
+    local enable_nftables=$2
     cp ../Documentation/kube-flannel.yml ./kube-flannel.yml
     yq -i 'select(.kind == "DaemonSet").spec.template.spec.containers[0].image |= strenv(FLANNEL_IMAGE)' ./kube-flannel.yml
 
-    export flannel_conf="{ \"EnableIPv6\": true, \"Network\": \"$FLANNEL_NET\", \"IPv6Network\":\"${FLANNEL_IP6NET}\", \"Backend\": { \"Type\": \"${backend}\" } }"
+    export flannel_conf="{ \"EnableIPv6\": true, \"Network\": \"$FLANNEL_NET\", \"IPv6Network\":\"${FLANNEL_IP6NET}\", \"Backend\": { \"Type\": \"${backend}\" }, \"EnableNFTables\": ${enable_nftables} }"
 
     yq -i 'select(.metadata.name == "kube-flannel-cfg").data."net-conf.json" |= strenv(flannel_conf)' ./kube-flannel.yml
 }
@@ -67,6 +69,10 @@ install-flannel() {
     kubectl --kubeconfig="${HOME}/.kube/config" apply -f ./kube-flannel.yml
 }
 
+delete-flannel() {
+    kubectl --kubeconfig="${HOME}/.kube/config" delete -f ./kube-flannel.yml
+}
+
 get_pod_ip() {
     local pod_name=$1
     kubectl --kubeconfig="${HOME}/.kube/config" get pod ${pod_name} --template '{{.status.podIP}}'
@@ -125,8 +131,9 @@ perf() {
 
 prepare_test() {
     local backend=$1
+    local enable_nftables=${2:-false}
     # install flannel version to test
-    write-flannel-conf ${backend} 
+    write-flannel-conf ${backend} ${enable_nftables}
     
     install-flannel
     # wait for nodes to be ready
@@ -150,18 +157,32 @@ test_vxlan() {
     prepare_test vxlan
     pings
     check_iptables
+    delete-flannel
+    check_iptables_removed
+}
+
+test_vxlan_nft() {
+    prepare_test vxlan true
+    pings
+    check_nftables
+    delete-flannel
+    check_nftables_removed
 }
 
 test_wireguard() {
     prepare_test wireguard
     pings
     check_iptables
+    delete-flannel
+    check_iptables_removed
 }
 
 test_host-gw() {
     prepare_test host-gw
     pings
     check_iptables
+    delete-flannel
+    check_iptables_removed
 }
 
 if [[ ${ARCH} == "amd64" ]]; then
@@ -169,6 +190,8 @@ test_udp() {
     prepare_test udp
     pings
     check_iptables
+    delete-flannel
+    check_iptables_removed
 }
 fi
 
@@ -176,6 +199,8 @@ test_ipip() {
     prepare_test ipip
     pings
     check_iptables
+    delete-flannel
+    check_iptables_removed
 }
 
 test_perf_vxlan() {
@@ -260,3 +285,94 @@ $(docker exec --privileged local-worker /usr/sbin/iptables -t filter -S FLANNEL-
                 "$(docker exec --privileged local-leader /usr/sbin/iptables -t filter -S FORWARD)
 $(docker exec --privileged local-leader /usr/sbin/iptables -t filter -S FLANNEL-FWD)" "Host 2 has not expected forward rules"
 }
+
+check_iptables_removed() {
+  local worker_podcidr=$(get_pod_cidr local-worker)
+  local leader_pod_cidr=$(get_pod_cidr local-leader)
+  read -r -d '' POSTROUTING_RULES_WORKER << EOM
+-N FLANNEL-POSTRTG
+EOM
+  read -r -d '' POSTROUTING_RULES_LEADER << EOM
+-N FLANNEL-POSTRTG
+EOM
+  read -r -d '' FORWARD_RULES << EOM
+-P FORWARD ACCEPT
+-A FORWARD -m conntrack --ctstate NEW -m comment --comment "kubernetes load balancer firewall" -j KUBE-PROXY-FIREWALL
+-A FORWARD -m comment --comment "kubernetes forwarding rules" -j KUBE-FORWARD
+-A FORWARD -m conntrack --ctstate NEW -m comment --comment "kubernetes service portals" -j KUBE-SERVICES
+-A FORWARD -m conntrack --ctstate NEW -m comment --comment "kubernetes externally-visible service portals" -j KUBE-EXTERNAL-SERVICES
+-N FLANNEL-FWD
+EOM
+# check that masquerade & forward rules have been removed
+  assert_equals "$POSTROUTING_RULES_WORKER" \
+                "$(docker exec --privileged local-worker /usr/sbin/iptables -t nat -S POSTROUTING | grep FLANNEL)$(docker exec --privileged local-worker /usr/sbin/iptables -t nat -S FLANNEL-POSTRTG)" "Host 1 has not expected postrouting rules"
+  assert_equals "$POSTROUTING_RULES_LEADER" \
+                "$(docker exec --privileged local-leader /usr/sbin/iptables -t nat -S POSTROUTING | grep FLANNEL)$(docker exec --privileged local-leader /usr/sbin/iptables -t nat -S FLANNEL-POSTRTG)" "Host 2 has not expected postrouting rules"
+  assert_equals "$FORWARD_RULES" \
+                "$(docker exec --privileged local-worker /usr/sbin/iptables -t filter -S FORWARD)
+$(docker exec --privileged local-worker /usr/sbin/iptables -t filter -S FLANNEL-FWD -w 5)" "Host 1 has not expected forward rules"
+  assert_equals "$FORWARD_RULES" \
+                "$(docker exec --privileged local-leader /usr/sbin/iptables -t filter -S FORWARD)
+$(docker exec --privileged local-leader /usr/sbin/iptables -t filter -S FLANNEL-FWD)" "Host 2 has not expected forward rules"
+}
+
+###nftables
+check_nftables() {
+  local worker_podcidr=$(get_pod_cidr local-worker)
+  local leader_podcidr=$(get_pod_cidr local-leader)
+  read -d '' POSTROUTING_RULES_WORKER << EOM
+table ip flannel-ipv4 {
+	chain postrtg {
+		type nat hook postrouting priority srcnat; policy accept;
+		meta mark 0x00004000 return
+		ip saddr ${worker_podcidr} ip daddr 10.42.0.0/16 return
+		ip saddr 10.42.0.0/16 ip daddr ${worker_podcidr} return
+		ip saddr != ${worker_podcidr} ip daddr 10.42.0.0/16 return
+		ip saddr 10.42.0.0/16 ip daddr != 224.0.0.0/4 masquerade fully-random
+		ip saddr != 10.42.0.0/16 ip daddr 10.42.0.0/16 masquerade fully-random
+	}
+}
+EOM
+  read -r -d '' POSTROUTING_RULES_LEADER << EOM
+table ip flannel-ipv4 {
+	chain postrtg {
+		type nat hook postrouting priority srcnat; policy accept;
+		meta mark 0x00004000 return
+		ip saddr ${leader_podcidr} ip daddr 10.42.0.0/16 return
+		ip saddr 10.42.0.0/16 ip daddr ${leader_podcidr} return
+		ip saddr != ${leader_podcidr} ip daddr 10.42.0.0/16 return
+		ip saddr 10.42.0.0/16 ip daddr != 224.0.0.0/4 masquerade fully-random
+		ip saddr != 10.42.0.0/16 ip daddr 10.42.0.0/16 masquerade fully-random
+	}
+}
+EOM
+  read -r -d '' FORWARD_RULES << EOM
+table ip flannel-ipv4 {
+	chain forward {
+		type filter hook input priority filter; policy accept;
+		ip saddr 10.42.0.0/16 accept
+		ip daddr 10.42.0.0/16 accept
+	}
+}
+EOM
+  # check masquerade & forward rules
+  assert_equals "$POSTROUTING_RULES_WORKER" \
+                "$(docker exec --privileged local-worker /usr/sbin/nft list chain flannel-ipv4 postrtg)" "Node worker does not have expected postrouting rules"
+  assert_equals "$POSTROUTING_RULES_LEADER" \
+                "$(docker exec --privileged local-leader /usr/sbin/nft list chain flannel-ipv4 postrtg)" "Node leader does not have expected postrouting rules"
+  assert_equals "$FORWARD_RULES" \
+                "$(docker exec --privileged local-worker /usr/sbin/nft list chain flannel-ipv4 forward)" "Node worker does not have expected forward rules"
+  assert_equals "$FORWARD_RULES" \
+                "$(docker exec --privileged local-leader /usr/sbin/nft list chain flannel-ipv4 forward)" "Node leader does not have expected forward rules"
+}
+
+  # check masquerade & forward rules
+  assert_equals "" \
+                "$(docker exec --privileged local-worker /usr/sbin/nft list chain flannel-ipv4 postrtg)" "Node worker has unexpected postrouting rules"
+  assert_equals "" \
+                "$(docker exec --privileged local-leader /usr/sbin/nft list chain flannel-ipv4 postrtg)" "Node leader has unexpected postrouting rules"
+  assert_equals "" \
+                "$(docker exec --privileged local-worker /usr/sbin/nft list chain flannel-ipv4 forward)" "Node worker has unexpected forward rules"
+  assert_equals "" \
+                "$(docker exec --privileged local-leader /usr/sbin/nft list chain flannel-ipv4 forward)" "Node leader has unexpected forward rules"
+}
\ No newline at end of file
diff --git a/go.mod b/go.mod
index 6d9205d63d..d41b5a0674 100644
--- a/go.mod
+++ b/go.mod
@@ -35,6 +35,7 @@ require (
 	github.com/avast/retry-go/v4 v4.5.1
 	github.com/tencentcloud/tencentcloud-sdk-go/tencentcloud/common v1.0.872
 	github.com/tencentcloud/tencentcloud-sdk-go/tencentcloud/vpc v1.0.872
+	sigs.k8s.io/knftables v0.0.14
 )
 
 require (
diff --git a/go.sum b/go.sum
index 520673ae92..8293d45f32 100644
--- a/go.sum
+++ b/go.sum
@@ -255,6 +255,8 @@ github.com/kr/pty v1.1.1/go.mod h1:pFQYn66WHrOpPYNljwOMqo10TkYh1fy3cYio2l3bCsQ=
 github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
 github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
 github.com/kr/text v0.2.0/go.mod h1:eLer722TekiGuMkidMxC/pM04lWEeraHUUmBw8l2grE=
+github.com/lithammer/dedent v1.1.0 h1:VNzHMVCBNG1j0fh3OrsFRkVUwStdDArbgBWoPAffktY=
+github.com/lithammer/dedent v1.1.0/go.mod h1:jrXYCQtgg0nJiN+StA2KgR7w6CiQNv9Fd/Z9BP0jIOc=
 github.com/mailru/easyjson v0.0.0-20190614124828-94de47d64c63/go.mod h1:C1wdFJiN94OJF2b5HbByQZoLdCWB1Yqtg26g4irojpc=
 github.com/mailru/easyjson v0.0.0-20190626092158-b2ccc519800e/go.mod h1:C1wdFJiN94OJF2b5HbByQZoLdCWB1Yqtg26g4irojpc=
 github.com/mailru/easyjson v0.7.6 h1:8yTIVnZgCoiM1TgqoeTl+LfU5Jg6/xL3QhGQnimLYnA=
@@ -781,6 +783,8 @@ rsc.io/quote/v3 v3.1.0/go.mod h1:yEA65RcK8LyAZtP9Kv3t0HmxON59tX3rD+tICJqUlj0=
 rsc.io/sampler v1.3.0/go.mod h1:T1hPZKmBbMNahiBKFy5HrXp6adAjACjK9JXDnKaTXpA=
 sigs.k8s.io/json v0.0.0-20220713155537-f223a00ba0e2 h1:iXTIw73aPyC+oRdyqqvVJuloN1p0AC/kzH07hu3NE+k=
 sigs.k8s.io/json v0.0.0-20220713155537-f223a00ba0e2/go.mod h1:B8JuhiUyNFVKdsE8h686QcCxMaH6HrOAZj4vswFpcB0=
+sigs.k8s.io/knftables v0.0.14 h1:VzKQoDMCGBOH8c85sGrWSXSPCS0XrIpEfOlcCLBXiC0=
+sigs.k8s.io/knftables v0.0.14/go.mod h1:f/5ZLKYEUPUhVjUCg6l80ACdL7CIIyeL0DxfgojGRTk=
 sigs.k8s.io/structured-merge-diff/v4 v4.2.3 h1:PRbqxJClWWYMNV1dhaG4NsibJbArud9kFxnAMREiWFE=
 sigs.k8s.io/structured-merge-diff/v4 v4.2.3/go.mod h1:qjx8mGObPmV2aSZepjQjbmb2ihdVs8cGKBraizNC69E=
 sigs.k8s.io/yaml v1.3.0 h1:a2VclLzOGrwOHDiV8EfBGhvjHvP46CtW5j6POvhYGGo=
diff --git a/images/Dockerfile b/images/Dockerfile
index fd60008878..3c8d248120 100644
--- a/images/Dockerfile
+++ b/images/Dockerfile
@@ -27,7 +27,7 @@ RUN export GOOS=$(xx-info os) &&\
 
 FROM alpine:20231219
 RUN apk update && apk upgrade
-RUN apk add --no-cache iproute2 ca-certificates iptables strongswan iptables-legacy && update-ca-certificates
+RUN apk add --no-cache iproute2 ca-certificates nftables iptables strongswan iptables-legacy && update-ca-certificates
 RUN apk add wireguard-tools --no-cache --repository http://dl-cdn.alpinelinux.org/alpine/edge/community
 COPY --from=build /build/dist/flanneld /opt/bin/flanneld
 COPY dist/mk-docker-opts.sh /opt/bin/
diff --git a/main.go b/main.go
index 94a95ba80e..702a2b476a 100644
--- a/main.go
+++ b/main.go
@@ -36,6 +36,7 @@ import (
 	"github.com/flannel-io/flannel/pkg/subnet/kube"
 	"github.com/flannel-io/flannel/pkg/trafficmngr"
 	"github.com/flannel-io/flannel/pkg/trafficmngr/iptables"
+	"github.com/flannel-io/flannel/pkg/trafficmngr/nftables"
 	"github.com/flannel-io/flannel/pkg/version"
 	"golang.org/x/net/context"
 	log "k8s.io/klog/v2"
@@ -336,7 +337,15 @@ func main() {
 	}
 
 	//Create TrafficManager and instanciate it based on whether we use iptables or nftables
-	trafficMngr := newTrafficManager()
+	trafficMngr := newTrafficManager(config.EnableNFTables)
+	err = trafficMngr.Init(ctx, &wg)
+	if err != nil {
+		log.Error(err)
+		cancel()
+		wg.Wait()
+		os.Exit(1)
+	}
+
 	flannelIPv4Net := ip.IP4Net{}
 	flannelIpv6Net := ip.IP6Net{}
 	if config.EnableIPv4 {
@@ -365,7 +374,8 @@ func main() {
 		prevIPv6Networks := ReadIP6CIDRsFromSubnetFile(opts.subnetFile, "FLANNEL_IPV6_NETWORK")
 		prevIPv6Subnet := ReadIP6CIDRFromSubnetFile(opts.subnetFile, "FLANNEL_IPV6_SUBNET")
 
-		err = trafficMngr.SetupAndEnsureMasqRules(flannelIPv4Net, prevSubnet,
+		err = trafficMngr.SetupAndEnsureMasqRules(ctx,
+			flannelIPv4Net, prevSubnet,
 			prevNetworks,
 			flannelIpv6Net, prevIPv6Subnet,
 			prevIPv6Networks,
@@ -383,7 +393,7 @@ func main() {
 	// In Docker 1.12 and earlier, the default FORWARD chain policy was ACCEPT.
 	// In Docker 1.13 and later, Docker sets the default policy of the FORWARD chain to DROP.
 	if opts.iptablesForwardRules {
-		trafficMngr.SetupAndEnsureForwardRules(
+		trafficMngr.SetupAndEnsureForwardRules(ctx,
 			flannelIPv4Net,
 			flannelIpv6Net,
 			opts.iptablesResyncSeconds)
@@ -569,6 +579,13 @@ func ReadIP6CIDRsFromSubnetFile(path string, CIDRKey string) []ip.IP6Net {
 	return prevCIDRs
 }
 
-func newTrafficManager() trafficmngr.TrafficManager {
-	return iptables.IPTablesManager{}
+func newTrafficManager(useNftables bool) trafficmngr.TrafficManager {
+	if useNftables {
+		log.Info("Starting flannel in nftables mode")
+		return &nftables.NFTablesManager{}
+	} else {
+		log.Info("Starting flannel in iptables mode")
+		return &iptables.IPTablesManager{}
+
+	}
 }
diff --git a/pkg/subnet/config.go b/pkg/subnet/config.go
index 5b8357cec4..b7a1afe270 100644
--- a/pkg/subnet/config.go
+++ b/pkg/subnet/config.go
@@ -27,20 +27,21 @@ import (
 )
 
 type Config struct {
-	EnableIPv4    bool
-	EnableIPv6    bool
-	Network       ip.IP4Net
-	IPv6Network   ip.IP6Net
-	Networks      []ip.IP4Net
-	IPv6Networks  []ip.IP6Net
-	SubnetMin     ip.IP4
-	SubnetMax     ip.IP4
-	IPv6SubnetMin *ip.IP6
-	IPv6SubnetMax *ip.IP6
-	SubnetLen     uint
-	IPv6SubnetLen uint
-	BackendType   string          `json:"-"`
-	Backend       json.RawMessage `json:",omitempty"`
+	EnableIPv4     bool
+	EnableIPv6     bool
+	EnableNFTables bool
+	Network        ip.IP4Net
+	IPv6Network    ip.IP6Net
+	Networks       []ip.IP4Net
+	IPv6Networks   []ip.IP6Net
+	SubnetMin      ip.IP4
+	SubnetMax      ip.IP4
+	IPv6SubnetMin  *ip.IP6
+	IPv6SubnetMax  *ip.IP6
+	SubnetLen      uint
+	IPv6SubnetLen  uint
+	BackendType    string          `json:"-"`
+	Backend        json.RawMessage `json:",omitempty"`
 }
 
 func parseBackendType(be json.RawMessage) (string, error) {
diff --git a/pkg/trafficmngr/iptables/iptables.go b/pkg/trafficmngr/iptables/iptables.go
index a16a607782..7fa93aefcc 100644
--- a/pkg/trafficmngr/iptables/iptables.go
+++ b/pkg/trafficmngr/iptables/iptables.go
@@ -17,7 +17,9 @@
 package iptables
 
 import (
+	"context"
 	"fmt"
+	"sync"
 	"time"
 
 	"github.com/coreos/go-iptables/iptables"
@@ -40,11 +42,67 @@ type IPTablesError interface {
 	Error() string
 }
 
-type IPTablesManager struct{}
+type IPTablesManager struct {
+	ipv4Rules []trafficmngr.IPTablesRule
+	ipv6Rules []trafficmngr.IPTablesRule
+}
+
+func (iptm *IPTablesManager) Init(ctx context.Context, wg *sync.WaitGroup) error {
+	iptm.ipv4Rules = make([]trafficmngr.IPTablesRule, 0, 10)
+	iptm.ipv6Rules = make([]trafficmngr.IPTablesRule, 0, 10)
+	wg.Add(1)
+	go func() {
+		<-ctx.Done()
+		time.Sleep(time.Second)
+		err := iptm.cleanUp()
+		if err != nil {
+			log.Errorf("iptables: error while cleaning-up: %v", err)
+		}
+		wg.Done()
+	}()
 
-const kubeProxyMark string = "0x4000/0x4000"
+	return nil
+}
+
+func (iptm *IPTablesManager) cleanUp() error {
+	if len(iptm.ipv4Rules) > 0 {
+		ipt, err := iptables.New()
+		if err != nil {
+			// if we can't find iptables, give up and return
+			return fmt.Errorf("failed to setup IPTables. iptables binary was not found: %v", err)
+		}
+		iptRestore, err := NewIPTablesRestoreWithProtocol(iptables.ProtocolIPv4)
+		if err != nil {
+			// if we can't find iptables-restore, give up and return
+			return fmt.Errorf("failed to setup IPTables. iptables-restore binary was not found: %v", err)
+		}
+		log.Info("iptables (ipv4): cleaning-up before exiting flannel...")
+		err = teardownIPTables(ipt, iptRestore, iptm.ipv4Rules)
+		if err != nil {
+			log.Errorf("Failed to tear down IPTables: %v", err)
+		}
+	}
+	if len(iptm.ipv6Rules) > 0 {
+		ipt, err := iptables.NewWithProtocol(iptables.ProtocolIPv6)
+		if err != nil {
+			// if we can't find iptables, give up and return
+			return fmt.Errorf("failed to setup IPTables. iptables binary was not found: %v", err)
+		}
+		iptRestore, err := NewIPTablesRestoreWithProtocol(iptables.ProtocolIPv6)
+		if err != nil {
+			// if we can't find iptables-restore, give up and return
+			return fmt.Errorf("failed to setup IPTables. iptables-restore binary was not found: %v", err)
+		}
+		log.Info("iptables (ipv6): cleaning-up before exiting flannel...")
+		err = teardownIPTables(ipt, iptRestore, iptm.ipv6Rules)
+		if err != nil {
+			log.Errorf("Failed to tear down IPTables: %v", err)
+		}
+	}
+	return nil
+}
 
-func (iptm IPTablesManager) SetupAndEnsureMasqRules(flannelIPv4Net, prevSubnet ip.IP4Net,
+func (iptm *IPTablesManager) SetupAndEnsureMasqRules(ctx context.Context, flannelIPv4Net, prevSubnet ip.IP4Net,
 	prevNetworks []ip.IP4Net,
 	flannelIPv6Net, prevIPv6Subnet ip.IP6Net,
 	prevIPv6Networks []ip.IP6Net,
@@ -77,7 +135,7 @@ func (iptm IPTablesManager) SetupAndEnsureMasqRules(flannelIPv4Net, prevSubnet i
 		getRules := func() []trafficmngr.IPTablesRule {
 			return iptm.masqRules([]ip.IP4Net{flannelIPv4Net}, currentlease)
 		}
-		go iptm.setupAndEnsureIP4Tables(getRules, resyncPeriod)
+		go iptm.setupAndEnsureIP4Tables(ctx, getRules, resyncPeriod)
 	}
 	if flannelIPv6Net.String() != "" {
 		//Find the cidr in FLANNEL_IPV6_NETWORK which contains the podCIDR (i.e. FLANNEL_IPV6_SUBNET) of this node
@@ -106,12 +164,12 @@ func (iptm IPTablesManager) SetupAndEnsureMasqRules(flannelIPv4Net, prevSubnet i
 		getRules := func() []trafficmngr.IPTablesRule {
 			return iptm.masqIP6Rules([]ip.IP6Net{flannelIPv6Net}, currentlease)
 		}
-		go iptm.setupAndEnsureIP6Tables(getRules, resyncPeriod)
+		go iptm.setupAndEnsureIP6Tables(ctx, getRules, resyncPeriod)
 	}
 	return nil
 }
 
-func (iptm IPTablesManager) masqRules(cluster_cidrs []ip.IP4Net, lease *lease.Lease) []trafficmngr.IPTablesRule {
+func (iptm *IPTablesManager) masqRules(cluster_cidrs []ip.IP4Net, lease *lease.Lease) []trafficmngr.IPTablesRule {
 	pod_cidr := lease.Subnet.String()
 	ipt, err := iptables.New()
 	supports_random_fully := false
@@ -122,7 +180,7 @@ func (iptm IPTablesManager) masqRules(cluster_cidrs []ip.IP4Net, lease *lease.Le
 	// This rule ensure that the flannel iptables rules are executed before other rules on the node
 	rules[0] = trafficmngr.IPTablesRule{Table: "nat", Action: "-A", Chain: "POSTROUTING", Rulespec: []string{"-m", "comment", "--comment", "flanneld masq", "-j", "FLANNEL-POSTRTG"}}
 	// This rule will not masquerade traffic marked by the kube-proxy to avoid double NAT bug on some kernel version
-	rules[1] = trafficmngr.IPTablesRule{Table: "nat", Action: "-A", Chain: "FLANNEL-POSTRTG", Rulespec: []string{"-m", "mark", "--mark", kubeProxyMark, "-m", "comment", "--comment", "flanneld masq", "-j", "RETURN"}}
+	rules[1] = trafficmngr.IPTablesRule{Table: "nat", Action: "-A", Chain: "FLANNEL-POSTRTG", Rulespec: []string{"-m", "mark", "--mark", trafficmngr.KubeProxyMark, "-m", "comment", "--comment", "flanneld masq", "-j", "RETURN"}}
 	for _, ccidr := range cluster_cidrs {
 		cluster_cidr := ccidr.String()
 		// This rule makes sure we don't NAT traffic within overlay network (e.g. coming out of docker0), for any of the cluster_cidrs
@@ -157,7 +215,7 @@ func (iptm IPTablesManager) masqRules(cluster_cidrs []ip.IP4Net, lease *lease.Le
 	return rules
 }
 
-func (iptm IPTablesManager) masqIP6Rules(cluster_cidrs []ip.IP6Net, lease *lease.Lease) []trafficmngr.IPTablesRule {
+func (iptm *IPTablesManager) masqIP6Rules(cluster_cidrs []ip.IP6Net, lease *lease.Lease) []trafficmngr.IPTablesRule {
 	pod_cidr := lease.IPv6Subnet.String()
 	ipt, err := iptables.NewWithProtocol(iptables.ProtocolIPv6)
 	supports_random_fully := false
@@ -169,7 +227,7 @@ func (iptm IPTablesManager) masqIP6Rules(cluster_cidrs []ip.IP6Net, lease *lease
 	// This rule ensure that the flannel iptables rules are executed before other rules on the node
 	rules[0] = trafficmngr.IPTablesRule{Table: "nat", Action: "-A", Chain: "POSTROUTING", Rulespec: []string{"-m", "comment", "--comment", "flanneld masq", "-j", "FLANNEL-POSTRTG"}}
 	// This rule will not masquerade traffic marked by the kube-proxy to avoid double NAT bug on some kernel version
-	rules[1] = trafficmngr.IPTablesRule{Table: "nat", Action: "-A", Chain: "FLANNEL-POSTRTG", Rulespec: []string{"-m", "mark", "--mark", kubeProxyMark, "-m", "comment", "--comment", "flanneld masq", "-j", "RETURN"}}
+	rules[1] = trafficmngr.IPTablesRule{Table: "nat", Action: "-A", Chain: "FLANNEL-POSTRTG", Rulespec: []string{"-m", "mark", "--mark", trafficmngr.KubeProxyMark, "-m", "comment", "--comment", "flanneld masq", "-j", "RETURN"}}
 
 	for _, ccidr := range cluster_cidrs {
 		cluster_cidr := ccidr.String()
@@ -208,14 +266,14 @@ func (iptm IPTablesManager) masqIP6Rules(cluster_cidrs []ip.IP6Net, lease *lease
 	return rules
 }
 
-func (iptm IPTablesManager) SetupAndEnsureForwardRules(flannelIPv4Network ip.IP4Net, flannelIPv6Network ip.IP6Net, resyncPeriod int) {
+func (iptm *IPTablesManager) SetupAndEnsureForwardRules(ctx context.Context, flannelIPv4Network ip.IP4Net, flannelIPv6Network ip.IP6Net, resyncPeriod int) {
 	if flannelIPv4Network.String() != "" {
 		log.Infof("Changing default FORWARD chain policy to ACCEPT")
 		iptm.CreateIP4Chain("filter", "FLANNEL-FWD")
 		getRules := func() []trafficmngr.IPTablesRule {
 			return iptm.forwardRules(flannelIPv4Network.String())
 		}
-		go iptm.setupAndEnsureIP4Tables(getRules, resyncPeriod)
+		go iptm.setupAndEnsureIP4Tables(ctx, getRules, resyncPeriod)
 	}
 	if flannelIPv6Network.String() != "" {
 		log.Infof("IPv6: Changing default FORWARD chain policy to ACCEPT")
@@ -223,11 +281,11 @@ func (iptm IPTablesManager) SetupAndEnsureForwardRules(flannelIPv4Network ip.IP4
 		getRules := func() []trafficmngr.IPTablesRule {
 			return iptm.forwardRules(flannelIPv6Network.String())
 		}
-		go iptm.setupAndEnsureIP6Tables(getRules, resyncPeriod)
+		go iptm.setupAndEnsureIP6Tables(ctx, getRules, resyncPeriod)
 	}
 }
 
-func (iptm IPTablesManager) forwardRules(flannelNetwork string) []trafficmngr.IPTablesRule {
+func (iptm *IPTablesManager) forwardRules(flannelNetwork string) []trafficmngr.IPTablesRule {
 	return []trafficmngr.IPTablesRule{
 		// This rule ensure that the flannel iptables rules are executed before other rules on the node
 		{Table: "filter", Action: "-A", Chain: "FORWARD", Rulespec: []string{"-m", "comment", "--comment", "flanneld forward", "-j", "FLANNEL-FWD"}},
@@ -237,7 +295,7 @@ func (iptm IPTablesManager) forwardRules(flannelNetwork string) []trafficmngr.IP
 	}
 }
 
-func (iptm IPTablesManager) CreateIP4Chain(table, chain string) {
+func (iptm *IPTablesManager) CreateIP4Chain(table, chain string) {
 	ipt, err := iptables.New()
 	if err != nil {
 		// if we can't find iptables, give up and return
@@ -252,7 +310,7 @@ func (iptm IPTablesManager) CreateIP4Chain(table, chain string) {
 	}
 }
 
-func (iptm IPTablesManager) CreateIP6Chain(table, chain string) {
+func (iptm *IPTablesManager) CreateIP6Chain(table, chain string) {
 	ipt, err := iptables.NewWithProtocol(iptables.ProtocolIPv6)
 	if err != nil {
 		// if we can't find iptables, give up and return
@@ -367,9 +425,8 @@ func ipTablesBootstrap(ipt IPTables, iptRestore IPTablesRestore, rules []traffic
 	return nil
 }
 
-func (iptm IPTablesManager) setupAndEnsureIP4Tables(getRules func() []trafficmngr.IPTablesRule, resyncPeriod int) {
+func (iptm *IPTablesManager) setupAndEnsureIP4Tables(ctx context.Context, getRules func() []trafficmngr.IPTablesRule, resyncPeriod int) {
 	rules := getRules()
-	log.Infof("generated %d rules", len(rules))
 	ipt, err := iptables.New()
 	if err != nil {
 		// if we can't find iptables, give up and return
@@ -389,24 +446,23 @@ func (iptm IPTablesManager) setupAndEnsureIP4Tables(getRules func() []trafficmng
 		log.Errorf("Failed to bootstrap IPTables: %v", err)
 	}
 
-	defer func() {
-		err := teardownIPTables(ipt, iptRestore, rules)
-		if err != nil {
-			log.Errorf("Failed to tear down IPTables: %v", err)
-		}
-	}()
-
+	iptm.ipv4Rules = append(iptm.ipv4Rules, rules...)
 	for {
-		// Ensure that all the iptables rules exist every 5 seconds
-		if err := ensureIPTables(ipt, iptRestore, getRules()); err != nil {
-			log.Errorf("Failed to ensure iptables rules: %v", err)
+		select {
+		case <-ctx.Done():
+			//clean-up is setup in Init
+			return
+		case <-time.After(time.Duration(resyncPeriod) * time.Second):
+			// Ensure that all the iptables rules exist every 5 seconds
+			if err := ensureIPTables(ipt, iptRestore, rules); err != nil {
+				log.Errorf("Failed to ensure iptables rules: %v", err)
+			}
 		}
 
-		time.Sleep(time.Duration(resyncPeriod) * time.Second)
 	}
 }
 
-func (iptm IPTablesManager) setupAndEnsureIP6Tables(getRules func() []trafficmngr.IPTablesRule, resyncPeriod int) {
+func (iptm *IPTablesManager) setupAndEnsureIP6Tables(ctx context.Context, getRules func() []trafficmngr.IPTablesRule, resyncPeriod int) {
 	rules := getRules()
 	ipt, err := iptables.NewWithProtocol(iptables.ProtocolIPv6)
 	if err != nil {
@@ -426,26 +482,24 @@ func (iptm IPTablesManager) setupAndEnsureIP6Tables(getRules func() []trafficmng
 		// if we can't find iptables, give up and return
 		log.Errorf("Failed to bootstrap IPTables: %v", err)
 	}
-
-	defer func() {
-		err := teardownIPTables(ipt, iptRestore, rules)
-		if err != nil {
-			log.Errorf("Failed to tear down IPTables: %v", err)
-		}
-	}()
+	iptm.ipv6Rules = append(iptm.ipv6Rules, rules...)
 
 	for {
-		// Ensure that all the iptables rules exist every 5 seconds
-		if err := ensureIPTables(ipt, iptRestore, getRules()); err != nil {
-			log.Errorf("Failed to ensure iptables rules: %v", err)
+		select {
+		case <-ctx.Done():
+			//clean-up is setup in Init
+			return
+		case <-time.After(time.Duration(resyncPeriod) * time.Second):
+			// Ensure that all the iptables rules exist every 5 seconds
+			if err := ensureIPTables(ipt, iptRestore, getRules()); err != nil {
+				log.Errorf("Failed to ensure iptables rules: %v", err)
+			}
 		}
-
-		time.Sleep(time.Duration(resyncPeriod) * time.Second)
 	}
 }
 
 // deleteIP4Tables delete specified iptables rules
-func (iptm IPTablesManager) deleteIP4Tables(rules []trafficmngr.IPTablesRule) error {
+func (iptm *IPTablesManager) deleteIP4Tables(rules []trafficmngr.IPTablesRule) error {
 	ipt, err := iptables.New()
 	if err != nil {
 		// if we can't find iptables, give up and return
@@ -467,7 +521,7 @@ func (iptm IPTablesManager) deleteIP4Tables(rules []trafficmngr.IPTablesRule) er
 }
 
 // deleteIP6Tables delete specified iptables rules
-func (iptm IPTablesManager) deleteIP6Tables(rules []trafficmngr.IPTablesRule) error {
+func (iptm *IPTablesManager) deleteIP6Tables(rules []trafficmngr.IPTablesRule) error {
 	ipt, err := iptables.NewWithProtocol(iptables.ProtocolIPv6)
 	if err != nil {
 		// if we can't find iptables, give up and return
@@ -538,6 +592,7 @@ func teardownIPTables(ipt IPTables, iptr IPTablesRestore, rules []trafficmngr.IP
 			// this shouldn't ever happen
 			return fmt.Errorf("failed to check rule existence: %v", err)
 		}
+
 		if exists {
 			if _, ok := tablesRules[rule.Table]; !ok {
 				tablesRules[rule.Table] = []IPTablesRestoreRuleSpec{}
diff --git a/pkg/trafficmngr/iptables/iptables_restore.go b/pkg/trafficmngr/iptables/iptables_restore.go
index 75973a6212..aa6e42a587 100644
--- a/pkg/trafficmngr/iptables/iptables_restore.go
+++ b/pkg/trafficmngr/iptables/iptables_restore.go
@@ -23,6 +23,7 @@ import (
 	"os/exec"
 	"regexp"
 	"strconv"
+	"sync"
 
 	"github.com/coreos/go-iptables/iptables"
 	log "k8s.io/klog/v2"
@@ -46,6 +47,10 @@ type ipTablesRestore struct {
 	path    string
 	proto   iptables.Protocol
 	hasWait bool
+	// ipTablesRestore needs a mutex to ensure that two avoid
+	// collisions between two goroutines calling ApplyWithoutFlush in parallel.
+	// This could result in the second call accidentally restoring a rule removed by the first
+	mu sync.Mutex
 }
 
 // IPTablesRestoreRules represents iptables-restore table block
@@ -75,12 +80,15 @@ func NewIPTablesRestoreWithProtocol(protocol iptables.Protocol) (IPTablesRestore
 		path:    path,
 		proto:   protocol,
 		hasWait: hasWait,
+		mu:      sync.Mutex{},
 	}
 	return &ipt, nil
 }
 
 // ApplyWithoutFlush apply without flush chains
 func (iptr *ipTablesRestore) ApplyWithoutFlush(rules IPTablesRestoreRules) error {
+	iptr.mu.Lock()
+	defer iptr.mu.Unlock()
 	payload := buildIPTablesRestorePayload(rules)
 
 	log.V(6).Infof("trying to run with payload %s", payload)
diff --git a/pkg/trafficmngr/iptables/iptables_windows.go b/pkg/trafficmngr/iptables/iptables_windows.go
index c63900b4b9..4d55415a06 100644
--- a/pkg/trafficmngr/iptables/iptables_windows.go
+++ b/pkg/trafficmngr/iptables/iptables_windows.go
@@ -15,8 +15,14 @@
 package iptables
 
 import (
+	"context"
+	"sync"
+
+	log "k8s.io/klog/v2"
+
 	"github.com/flannel-io/flannel/pkg/ip"
 	"github.com/flannel-io/flannel/pkg/lease"
+	"github.com/flannel-io/flannel/pkg/trafficmngr"
 )
 
 type IPTablesManager struct{}
@@ -29,14 +35,19 @@ type IPTables interface {
 	Exists(table string, chain string, rulespec ...string) (bool, error)
 }
 
-func (iptm IPTablesManager) SetupAndEnsureForwardRules(flannelIPv4Network ip.IP4Net, flannelIPv6Network ip.IP6Net, resyncPeriod int) {
+func (iptm IPTablesManager) Init(ctx context.Context, wg *sync.WaitGroup) error {
+	return trafficmngr.ErrUnimplemented
+}
+
+func (iptm *IPTablesManager) SetupAndEnsureForwardRules(ctx context.Context, flannelIPv4Network ip.IP4Net, flannelIPv6Network ip.IP6Net, resyncPeriod int) {
 }
 
-func (iptm IPTablesManager) SetupAndEnsureMasqRules(flannelIPv4Net, prevSubnet ip.IP4Net,
+func (iptm *IPTablesManager) SetupAndEnsureMasqRules(ctx context.Context, flannelIPv4Net, prevSubnet ip.IP4Net,
 	prevNetworks []ip.IP4Net,
 	flannelIPv6Net, prevIPv6Subnet ip.IP6Net,
 	prevIPv6Networks []ip.IP6Net,
 	currentlease *lease.Lease,
 	resyncPeriod int) error {
+	log.Error(trafficmngr.ErrUnimplemented)
 	return nil
 }
diff --git a/pkg/trafficmngr/nftables/nftables.go b/pkg/trafficmngr/nftables/nftables.go
new file mode 100644
index 0000000000..41ab505ca6
--- /dev/null
+++ b/pkg/trafficmngr/nftables/nftables.go
@@ -0,0 +1,317 @@
+// Copyright 2024 flannel authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//go:build !windows
+// +build !windows
+
+package nftables
+
+import (
+	"context"
+	"fmt"
+	"sync"
+	"time"
+
+	log "k8s.io/klog/v2"
+
+	"github.com/flannel-io/flannel/pkg/ip"
+	"github.com/flannel-io/flannel/pkg/lease"
+	"sigs.k8s.io/knftables"
+)
+
+const (
+	ipv4Table    = "flannel-ipv4"
+	ipv6Table    = "flannel-ipv6"
+	forwardChain = "forward"
+	postrtgChain = "postrtg"
+	//maximum delay in second to clean-up when the context is cancelled
+	cleanUpDeadline = 15
+)
+
+type NFTablesManager struct {
+	nftv4 knftables.Interface
+	nftv6 knftables.Interface
+}
+
+func (nftm *NFTablesManager) Init(ctx context.Context, wg *sync.WaitGroup) error {
+	var err error
+	nftm.nftv4, err = initTable(ctx, knftables.IPv4Family, ipv4Table)
+	if err != nil {
+		return err
+	}
+	nftm.nftv6, err = initTable(ctx, knftables.IPv6Family, ipv6Table)
+	if err != nil {
+		return err
+	}
+
+	wg.Add(1)
+	go func() {
+		<-ctx.Done()
+		log.Info("Cleaning-up flannel tables...")
+
+		cleanupCtx, cleanUpCancelFunc := context.WithTimeout(context.Background(), cleanUpDeadline*time.Second)
+		defer cleanUpCancelFunc()
+		err := nftm.cleanUp(cleanupCtx)
+		log.Errorf("nftables: error while cleaning-up: %v", err)
+		wg.Done()
+	}()
+	return nil
+}
+
+// create a new table and returns the interface to interact with it
+func initTable(ctx context.Context, ipFamily knftables.Family, name string) (knftables.Interface, error) {
+	nft, err := knftables.New(ipFamily, name)
+	if err != nil {
+		return nil, fmt.Errorf("no nftables support: %v", err)
+	}
+	tx := nft.NewTransaction()
+
+	tx.Add(&knftables.Table{
+		Comment: knftables.PtrTo("rules for " + name),
+	})
+	err = nft.Run(ctx, tx)
+	if err != nil {
+		return nil, fmt.Errorf("nftables: couldn't initialise table %s: %v", name, err)
+	}
+	return nft, nil
+}
+
+// It is needed when using nftables? accept seems to be the default
+// warning: never add a default 'drop' policy on the forwarChain as it breaks connectivity to the node
+func (nftm *NFTablesManager) SetupAndEnsureForwardRules(ctx context.Context,
+	flannelIPv4Network ip.IP4Net, flannelIPv6Network ip.IP6Net, resyncPeriod int) {
+	if !flannelIPv4Network.Empty() {
+		log.Infof("Changing default FORWARD chain policy to ACCEPT")
+		tx := nftm.nftv4.NewTransaction()
+
+		//TODO how to express that the default is drop?
+		tx.Add(&knftables.Chain{
+			Name:     forwardChain,
+			Comment:  knftables.PtrTo("chain to accept flannel traffic"),
+			Type:     knftables.PtrTo(knftables.FilterType),
+			Hook:     knftables.PtrTo(knftables.InputHook),
+			Priority: knftables.PtrTo(knftables.FilterPriority),
+		})
+		tx.Flush(&knftables.Chain{
+			Name: forwardChain,
+		})
+
+		tx.Add(&knftables.Rule{
+			Chain: forwardChain,
+			Rule: knftables.Concat(
+				"ip saddr", flannelIPv4Network.String(),
+				"accept",
+			),
+		})
+		tx.Add(&knftables.Rule{
+			Chain: forwardChain,
+			Rule: knftables.Concat(
+				"ip daddr", flannelIPv4Network.String(),
+				"accept",
+			),
+		})
+		err := nftm.nftv4.Run(ctx, tx)
+		if err != nil {
+			log.Errorf("nftables: couldn't setup forward rules: %v", err)
+		}
+	}
+	if !flannelIPv6Network.Empty() {
+		log.Infof("Changing default FORWARD chain policy to ACCEPT (ipv6)")
+		tx := nftm.nftv6.NewTransaction()
+
+		//TODO how to express that the default is drop?
+		tx.Add(&knftables.Chain{
+			Name:     forwardChain,
+			Comment:  knftables.PtrTo("chain to accept flannel traffic"),
+			Type:     knftables.PtrTo(knftables.FilterType),
+			Hook:     knftables.PtrTo(knftables.InputHook),
+			Priority: knftables.PtrTo(knftables.FilterPriority),
+		})
+		tx.Flush(&knftables.Chain{
+			Name: forwardChain,
+		})
+
+		tx.Add(&knftables.Rule{
+			Chain: forwardChain,
+			Rule: knftables.Concat(
+				"ip6 saddr", flannelIPv6Network.String(),
+				"accept",
+			),
+		})
+		tx.Add(&knftables.Rule{
+			Chain: forwardChain,
+			Rule: knftables.Concat(
+				"ip6 daddr", flannelIPv6Network.String(),
+				"accept",
+			),
+		})
+		err := nftm.nftv6.Run(ctx, tx)
+		if err != nil {
+			log.Errorf("nftables: couldn't setup forward rules (ipv6): %v", err)
+		}
+	}
+}
+
+func (nftm *NFTablesManager) SetupAndEnsureMasqRules(ctx context.Context, flannelIPv4Net, prevSubnet ip.IP4Net,
+	prevNetworks []ip.IP4Net,
+	flannelIPv6Net, prevIPv6Subnet ip.IP6Net,
+	prevIPv6Networks []ip.IP6Net,
+	currentlease *lease.Lease,
+	resyncPeriod int) error {
+	if !flannelIPv4Net.Empty() {
+		log.Infof("nftables: setting up masking rules (ipv4)")
+		tx := nftm.nftv4.NewTransaction()
+
+		tx.Add(&knftables.Chain{
+			Name:     postrtgChain,
+			Comment:  knftables.PtrTo("chain to manage traffic masquerading by flannel"),
+			Type:     knftables.PtrTo(knftables.NATType),
+			Hook:     knftables.PtrTo(knftables.PostroutingHook),
+			Priority: knftables.PtrTo(knftables.SNATPriority),
+		})
+		// make sure that the chain is empty before adding our rules
+		// => no need for the check and recycle part of iptables.go
+		tx.Flush(&knftables.Chain{
+			Name: postrtgChain,
+		})
+		err := nftm.addMasqRules(ctx, tx, flannelIPv4Net.String(), currentlease.Subnet.String(), knftables.IPv4Family)
+		if err != nil {
+			return fmt.Errorf("nftables: couldn't setup masq rules: %v", err)
+		}
+		err = nftm.nftv4.Run(ctx, tx)
+		if err != nil {
+			return fmt.Errorf("nftables: couldn't setup masq rules: %v", err)
+		}
+	}
+	if !flannelIPv6Net.Empty() {
+		log.Infof("nftables: setting up masking rules (ipv6)")
+		tx := nftm.nftv6.NewTransaction()
+
+		tx.Add(&knftables.Chain{
+			Name:     postrtgChain,
+			Comment:  knftables.PtrTo("chain to manage traffic masquerading by flannel"),
+			Type:     knftables.PtrTo(knftables.NATType),
+			Hook:     knftables.PtrTo(knftables.PostroutingHook),
+			Priority: knftables.PtrTo(knftables.SNATPriority),
+		})
+		// make sure that the chain is empty before adding our rules
+		// => no need for the check and recycle part of iptables.go
+		tx.Flush(&knftables.Chain{
+			Name: postrtgChain,
+		})
+		err := nftm.addMasqRules(ctx, tx, flannelIPv6Net.String(), currentlease.IPv6Subnet.String(), knftables.IPv6Family)
+		if err != nil {
+			return fmt.Errorf("nftables: couldn't setup masq rules: %v", err)
+		}
+		err = nftm.nftv6.Run(ctx, tx)
+		if err != nil {
+			return fmt.Errorf("nftables: couldn't setup masq rules: %v", err)
+		}
+	}
+	return nil
+}
+
+// add required masking rules to transaction tx
+func (nftm *NFTablesManager) addMasqRules(ctx context.Context,
+	tx *knftables.Transaction,
+	clusterCidr, podCidr string,
+	family knftables.Family) error {
+	masquerade := "masquerade fully-random"
+	if !nftm.checkRandomfully(ctx) {
+		masquerade = "masquerade"
+	}
+
+	multicastCidr := "224.0.0.0/4"
+	if family == knftables.IPv6Family {
+		multicastCidr = "ff00::/8"
+	}
+	// This rule will not masquerade traffic marked
+	// by the kube-proxy to avoid double NAT bug on some kernel version
+	tx.Add(&knftables.Rule{
+		Chain: postrtgChain,
+		Rule: knftables.Concat(
+			"meta mark", "0x4000", //TODO_TF: check the correct value when deploying kube-proxy
+			"return",
+		),
+	})
+	// don't NAT traffic within overlay network
+	tx.Add(&knftables.Rule{
+		Chain: postrtgChain,
+		Rule: knftables.Concat(
+			family, "saddr", podCidr,
+			family, "daddr", clusterCidr,
+			"return",
+		),
+	})
+	tx.Add(&knftables.Rule{
+		Chain: postrtgChain,
+		Rule: knftables.Concat(
+			family, "saddr", clusterCidr,
+			family, "daddr", podCidr,
+			"return",
+		),
+	})
+	// Prevent performing Masquerade on external traffic which arrives from a Node that owns the container/pod IP address
+	tx.Add(&knftables.Rule{
+		Chain: postrtgChain,
+		Rule: knftables.Concat(
+			family, "saddr", "!=", podCidr,
+			family, "daddr", clusterCidr,
+			"return",
+		),
+	})
+	// NAT if it's not multicast traffic
+	tx.Add(&knftables.Rule{
+		Chain: postrtgChain,
+		Rule: knftables.Concat(
+			family, "saddr", clusterCidr,
+			family, "daddr", "!=", multicastCidr,
+			masquerade,
+		),
+	})
+	// Masquerade anything headed towards flannel from the host
+	tx.Add(&knftables.Rule{
+		Chain: postrtgChain,
+		Rule: knftables.Concat(
+			family, "saddr", "!=", clusterCidr,
+			family, "daddr", clusterCidr,
+			masquerade,
+		),
+	})
+	return nil
+}
+
+// clean-up all nftables states created by flannel by deleting all related tables
+func (nftm *NFTablesManager) cleanUp(ctx context.Context) error {
+	nft, err := knftables.New(knftables.IPv4Family, ipv4Table)
+	if err == nil {
+		tx := nft.NewTransaction()
+		tx.Delete(&knftables.Table{})
+		err = nft.Run(ctx, tx)
+	}
+	if err != nil {
+		return fmt.Errorf("nftables: couldn't delete table: %v", err)
+	}
+
+	nft, err = knftables.New(knftables.IPv6Family, ipv6Table)
+	if err == nil {
+		tx := nft.NewTransaction()
+		tx.Delete(&knftables.Table{})
+		err = nft.Run(ctx, tx)
+	}
+	if err != nil {
+		return fmt.Errorf("nftables (ipv6): couldn't delete table: %v", err)
+	}
+
+	return nil
+}
diff --git a/pkg/trafficmngr/nftables/nftables_windows.go b/pkg/trafficmngr/nftables/nftables_windows.go
new file mode 100644
index 0000000000..b66034b5b7
--- /dev/null
+++ b/pkg/trafficmngr/nftables/nftables_windows.go
@@ -0,0 +1,48 @@
+// Copyright 2024 flannel authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package nftables
+
+import (
+	"context"
+	"sync"
+
+	log "k8s.io/klog/v2"
+
+	"github.com/flannel-io/flannel/pkg/ip"
+	"github.com/flannel-io/flannel/pkg/lease"
+	"github.com/flannel-io/flannel/pkg/trafficmngr"
+)
+
+type NFTablesManager struct {
+}
+
+func (nftm *NFTablesManager) Init(ctx context.Context, wg *sync.WaitGroup) error {
+	log.Error(trafficmngr.ErrUnimplemented)
+	return nil
+}
+
+func (nftm *NFTablesManager) SetupAndEnsureForwardRules(ctx context.Context,
+	flannelIPv4Network ip.IP4Net, flannelIPv6Network ip.IP6Net, resyncPeriod int) {
+}
+
+func (nftm *NFTablesManager) SetupAndEnsureMasqRules(ctx context.Context, flannelIPv4Net, prevSubnet ip.IP4Net,
+	prevNetworks []ip.IP4Net,
+	flannelIPv6Net, prevIPv6Subnet ip.IP6Net,
+	prevIPv6Networks []ip.IP6Net,
+	currentlease *lease.Lease,
+	resyncPeriod int) error {
+	log.Error(trafficmngr.ErrUnimplemented)
+	return nil
+}
diff --git a/pkg/trafficmngr/nftables/utils.go b/pkg/trafficmngr/nftables/utils.go
new file mode 100644
index 0000000000..fa3f269361
--- /dev/null
+++ b/pkg/trafficmngr/nftables/utils.go
@@ -0,0 +1,58 @@
+// Copyright 2024 flannel authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//go:build !windows
+// +build !windows
+
+package nftables
+
+import (
+	"context"
+
+	log "k8s.io/klog/v2"
+	"sigs.k8s.io/knftables"
+)
+
+const (
+	masqueradeTestTable = "masqueradeTest"
+)
+
+// check whether masquerade fully-random is supported by the kernel
+func (nftm *NFTablesManager) checkRandomfully(ctx context.Context) bool {
+	result := true
+	tx := nftm.nftv4.NewTransaction()
+	tx.Add(&knftables.Chain{
+		Name:     masqueradeTestTable,
+		Comment:  knftables.PtrTo("chain to test if masquerade random fully is supported"),
+		Type:     knftables.PtrTo(knftables.NATType),
+		Hook:     knftables.PtrTo(knftables.PostroutingHook),
+		Priority: knftables.PtrTo(knftables.SNATPriority),
+	})
+	tx.Flush(&knftables.Chain{
+		Name: masqueradeTestTable,
+	})
+	// Masquerade anything headed towards flannel from the host
+	tx.Add(&knftables.Rule{
+		Chain: masqueradeTestTable,
+		Rule: knftables.Concat(
+			"ip saddr", "!=", "127.0.0.1",
+			"masquerade fully-random",
+		),
+	})
+	err := nftm.nftv4.Check(ctx, tx)
+	if err != nil {
+		log.Warningf("nftables: random fully unsupported")
+		result = false
+	}
+	return result
+}
diff --git a/pkg/trafficmngr/trafficmngr.go b/pkg/trafficmngr/trafficmngr.go
index 7ae2ef7ae0..99f401c4c5 100644
--- a/pkg/trafficmngr/trafficmngr.go
+++ b/pkg/trafficmngr/trafficmngr.go
@@ -15,6 +15,10 @@
 package trafficmngr
 
 import (
+	"context"
+	"errors"
+	"sync"
+
 	"github.com/flannel-io/flannel/pkg/ip"
 	"github.com/flannel-io/flannel/pkg/lease"
 )
@@ -26,19 +30,27 @@ type IPTablesRule struct {
 	Rulespec []string
 }
 
+var (
+	ErrUnimplemented = errors.New("unimplemented")
+)
+
+const KubeProxyMark string = "0x4000/0x4000"
+
 type TrafficManager interface {
+	// Initialize the TrafficManager, including the go routine to clean-up when flanneld is closed
+	Init(ctx context.Context, wg *sync.WaitGroup) error
 	// Install kernel rules to forward the traffic to and from the flannel network range.
 	// This is done for IPv4 and/or IPv6 based on whether flannelIPv4Network and flannelIPv6Network are set.
 	// SetupAndEnsureForwardRules starts a go routine that
 	// rewrites these rules every resyncPeriod seconds if needed
-	SetupAndEnsureForwardRules(flannelIPv4Network ip.IP4Net, flannelIPv6Network ip.IP6Net, resyncPeriod int)
+	SetupAndEnsureForwardRules(ctx context.Context, flannelIPv4Network ip.IP4Net, flannelIPv6Network ip.IP6Net, resyncPeriod int)
 	// Install kernel rules to setup NATing of packets sent to the flannel interface
 	// This is done for IPv4 and/or IPv6 based on whether flannelIPv4Network and flannelIPv6Network are set.
 	// prevSubnet,prevNetworks, prevIPv6Subnet, prevIPv6Networks are used
 	// to determine whether the existing rules need to be replaced.
 	// SetupAndEnsureMasqRules starts a go routine that
 	// rewrites these rules every resyncPeriod seconds if needed
-	SetupAndEnsureMasqRules(
+	SetupAndEnsureMasqRules(ctx context.Context,
 		flannelIPv4Net, prevSubnet ip.IP4Net,
 		prevNetworks []ip.IP4Net,
 		flannelIPv6Net, prevIPv6Subnet ip.IP6Net,