diff --git a/test/e2e/BUILD b/test/e2e/BUILD
index 3c4b2b16627..0120407bd48 100644
--- a/test/e2e/BUILD
+++ b/test/e2e/BUILD
@@ -34,6 +34,7 @@ go_test(
         "//test/e2e/lifecycle:go_default_library",
         "//test/e2e/lifecycle/bootstrap:go_default_library",
         "//test/e2e/network:go_default_library",
+        "//test/e2e/network/netpol:go_default_library",
         "//test/e2e/node:go_default_library",
         "//test/e2e/scheduling:go_default_library",
         "//test/e2e/storage:go_default_library",
diff --git a/test/e2e/e2e_test.go b/test/e2e/e2e_test.go
index 61f7f61727e..56307f4de61 100644
--- a/test/e2e/e2e_test.go
+++ b/test/e2e/e2e_test.go
@@ -51,6 +51,7 @@ import (
 	_ "k8s.io/kubernetes/test/e2e/lifecycle"
 	_ "k8s.io/kubernetes/test/e2e/lifecycle/bootstrap"
 	_ "k8s.io/kubernetes/test/e2e/network"
+	_ "k8s.io/kubernetes/test/e2e/network/netpol"
 	_ "k8s.io/kubernetes/test/e2e/node"
 	_ "k8s.io/kubernetes/test/e2e/scheduling"
 	_ "k8s.io/kubernetes/test/e2e/storage"
diff --git a/test/e2e/network/BUILD b/test/e2e/network/BUILD
index bdd1fd8bea5..e6998d0ba5d 100644
--- a/test/e2e/network/BUILD
+++ b/test/e2e/network/BUILD
@@ -110,6 +110,7 @@ filegroup(
     name = "all-srcs",
     srcs = [
         ":package-srcs",
+        "//test/e2e/network/netpol:all-srcs",
         "//test/e2e/network/scale:all-srcs",
     ],
     tags = ["automanaged"],
diff --git a/test/e2e/network/netpol/BUILD b/test/e2e/network/netpol/BUILD
new file mode 100644
index 00000000000..033ee9d0b3b
--- /dev/null
+++ b/test/e2e/network/netpol/BUILD
@@ -0,0 +1,47 @@
+package(default_visibility = ["//visibility:public"])
+
+load("@io_bazel_rules_go//go:def.bzl", "go_library")
+
+go_library(
+    name = "go_default_library",
+    srcs = [
+        "k8s_util.go",
+        "model.go",
+        "network_policy.go",
+        "policies.go",
+        "probe.go",
+        "reachability.go",
+        "test_helper.go",
+        "truthtable.go",
+    ],
+    importpath = "k8s.io/kubernetes/test/e2e/network/netpol",
+    deps = [
+        "//staging/src/k8s.io/api/core/v1:go_default_library",
+        "//staging/src/k8s.io/api/networking/v1:go_default_library",
+        "//staging/src/k8s.io/apimachinery/pkg/apis/meta/v1:go_default_library",
+        "//staging/src/k8s.io/apimachinery/pkg/util/intstr:go_default_library",
+        "//staging/src/k8s.io/apimachinery/pkg/util/wait:go_default_library",
+        "//staging/src/k8s.io/client-go/kubernetes:go_default_library",
+        "//test/e2e/framework:go_default_library",
+        "//test/e2e/framework/pod:go_default_library",
+        "//test/e2e/network:go_default_library",
+        "//test/utils/image:go_default_library",
+        "//vendor/github.com/onsi/ginkgo:go_default_library",
+        "//vendor/github.com/pkg/errors:go_default_library",
+        "//vendor/k8s.io/utils/net:go_default_library",
+        "//vendor/sigs.k8s.io/yaml:go_default_library",
+    ],
+)
+
+filegroup(
+    name = "package-srcs",
+    srcs = glob(["**"]),
+    tags = ["automanaged"],
+    visibility = ["//visibility:private"],
+)
+
+filegroup(
+    name = "all-srcs",
+    srcs = [":package-srcs"],
+    tags = ["automanaged"],
+)
diff --git a/test/e2e/network/netpol/k8s_util.go b/test/e2e/network/netpol/k8s_util.go
new file mode 100644
index 00000000000..264ebc2912e
--- /dev/null
+++ b/test/e2e/network/netpol/k8s_util.go
@@ -0,0 +1,351 @@
+/*
+Copyright 2020 The Kubernetes 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 netpol
+
+import (
+	"context"
+	"fmt"
+	"strings"
+	"sync"
+	"time"
+
+	e2epod "k8s.io/kubernetes/test/e2e/framework/pod"
+
+	"github.com/pkg/errors"
+	v1 "k8s.io/api/core/v1"
+	networkingv1 "k8s.io/api/networking/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	clientset "k8s.io/client-go/kubernetes"
+	"k8s.io/kubernetes/test/e2e/framework"
+)
+
+// Scenario provides a convenience interface to kube functionality that we leverage for polling NetworkPolicy connections.
+type Scenario struct {
+	mutex     sync.Mutex
+	podCache  map[string][]v1.Pod
+	framework *framework.Framework
+	ClientSet clientset.Interface
+}
+
+// NewScenario is a utility function that wraps creation of the stuff we're using for creating a cluster that expresses the global policy scenario we're testing.
+func NewScenario(framework *framework.Framework) *Scenario {
+	return &Scenario{
+		podCache:  map[string][]v1.Pod{},
+		framework: framework,
+		ClientSet: framework.ClientSet,
+	}
+}
+
+// InitializeCluster checks the state of the cluster, creating or updating namespaces and deployments as needed
+func (k *Scenario) InitializeCluster(model *Model) error {
+	var createdPods []*v1.Pod
+	for _, ns := range model.Namespaces {
+		_, err := k.CreateNamespace(ns.Spec())
+		if err != nil {
+			return err
+		}
+
+		for _, pod := range ns.Pods {
+			framework.Logf("creating/updating pod %s/%s", ns.Name, pod.Name)
+
+			kubePod, err := k.CreatePod(pod.KubePod())
+			if err != nil {
+				return err
+			}
+			createdPods = append(createdPods, kubePod)
+
+			_, err = k.CreateService(pod.Service())
+			if err != nil {
+				return err
+			}
+		}
+	}
+
+	for _, podString := range model.AllPodStrings() {
+		k8sPod, err := k.GetPodFromCache(podString.Namespace(), podString.PodName())
+		if err != nil {
+			return err
+		}
+		if k8sPod == nil {
+			return errors.Errorf("unable to find pod in ns %s with key/val pod=%s", podString.Namespace(), podString.PodName())
+		}
+		err = e2epod.WaitForPodNameRunningInNamespace(k.ClientSet, k8sPod.Name, k8sPod.Namespace)
+		if err != nil {
+			return errors.Wrapf(err, "unable to wait for pod %s/%s", podString.Namespace(), podString.PodName())
+		}
+	}
+
+	for _, createdPod := range createdPods {
+		err := e2epod.WaitForPodRunningInNamespace(k.ClientSet, createdPod)
+		if err != nil {
+			return errors.Wrapf(err, "unable to wait for pod %s/%s", createdPod.Namespace, createdPod.Name)
+		}
+	}
+
+	return nil
+}
+
+// GetPodFromCache returns a pod with the matching namespace and name
+func (k *Scenario) GetPodFromCache(ns string, name string) (*v1.Pod, error) {
+	pods, err := k.getPodsUncached(ns, "pod", name)
+	if err != nil {
+		return nil, err
+	}
+	if len(pods) == 0 {
+		return nil, nil
+	}
+	return &pods[0], nil
+}
+
+func (k *Scenario) getPodsUncached(ns string, key string, val string) ([]v1.Pod, error) {
+	v1PodList, err := k.ClientSet.CoreV1().Pods(ns).List(context.TODO(), metav1.ListOptions{
+		LabelSelector: fmt.Sprintf("%v=%v", key, val),
+	})
+	if err != nil {
+		return nil, errors.Wrapf(err, "unable to list Pods in ns %s with key/val %s=%s", ns, key, val)
+	}
+	return v1PodList.Items, nil
+}
+
+// GetPodsFromCacheByKeyVal returns an array of all Pods in the given namespace having a k/v label pair.
+func (k *Scenario) GetPodsFromCacheByKeyVal(ns string, key string, val string) ([]v1.Pod, error) {
+	k.mutex.Lock()
+	p, ok := k.podCache[fmt.Sprintf("%v_%v_%v", ns, key, val)]
+	k.mutex.Unlock()
+	if ok {
+		return p, nil
+	}
+
+	v1PodList, err := k.getPodsUncached(ns, key, val)
+	if err != nil {
+		return nil, err
+	}
+
+	k.mutex.Lock()
+	k.podCache[fmt.Sprintf("%v_%v_%v", ns, key, val)] = v1PodList
+	k.mutex.Unlock()
+
+	return v1PodList, nil
+}
+
+// GetPod gets a pod by namespace and name
+func (k *Scenario) GetPod(ns string, name string) (*v1.Pod, error) {
+	kubePod, err := k.ClientSet.CoreV1().Pods(ns).Get(context.TODO(), name, metav1.GetOptions{})
+	if err != nil {
+		return nil, errors.Wrapf(err, "unable to get pod %s/%s", ns, name)
+	}
+	return kubePod, nil
+}
+
+// Probe execs into a pod and checks its connectivity to another pod.
+func (k *Scenario) Probe(nsFrom string, podFrom string, containerFrom string, addrTo string, protocol v1.Protocol, toPort int) (bool, string, error) {
+	fromPods, err := k.GetPodsFromCacheByKeyVal(nsFrom, "pod", podFrom)
+	if err != nil {
+		return false, "", err
+	}
+	if len(fromPods) == 0 {
+		return false, "", errors.Errorf("pod %s/%s not found", nsFrom, podFrom)
+	}
+	fromPod := fromPods[0]
+
+	var cmd []string
+	switch protocol {
+	case v1.ProtocolSCTP:
+		cmd = []string{"/agnhost", "connect", fmt.Sprintf("%s:%d", addrTo, toPort), "--timeout=1s", "--protocol=sctp"}
+	case v1.ProtocolTCP:
+		cmd = []string{"/agnhost", "connect", fmt.Sprintf("%s:%d", addrTo, toPort), "--timeout=1s", "--protocol=tcp"}
+	case v1.ProtocolUDP:
+		cmd = []string{"nc", "-v", "-z", "-w", "1", "-u", addrTo, fmt.Sprintf("%d", toPort)}
+	default:
+		framework.Failf("protocol %s not supported", protocol)
+	}
+
+	commandDebugString := fmt.Sprintf("kubectl exec %s -c %s -n %s -- %s", fromPod.Name, containerFrom, fromPod.Namespace, strings.Join(cmd, " "))
+	stdout, stderr, err := k.ExecuteRemoteCommand(fromPod, containerFrom, cmd)
+	if err != nil {
+		framework.Logf("%s/%s -> %s: error when running command: err - %v /// stdout - %s /// stderr - %s", nsFrom, podFrom, addrTo, err, stdout, stderr)
+		return false, commandDebugString, nil
+	}
+	return true, commandDebugString, nil
+}
+
+// ExecuteRemoteCommand executes a remote shell command on the given pod
+func (k *Scenario) ExecuteRemoteCommand(pod v1.Pod, containerName string, command []string) (string, string, error) {
+	return k.framework.ExecWithOptions(framework.ExecOptions{
+		Command:            command,
+		Namespace:          pod.Namespace,
+		PodName:            pod.Name,
+		ContainerName:      containerName,
+		Stdin:              nil,
+		CaptureStdout:      true,
+		CaptureStderr:      true,
+		PreserveWhitespace: false,
+	})
+
+}
+
+// CreateNamespace is a convenience function for namespace setup
+func (k *Scenario) CreateNamespace(ns *v1.Namespace) (*v1.Namespace, error) {
+	createdNamespace, err := k.ClientSet.CoreV1().Namespaces().Create(context.TODO(), ns, metav1.CreateOptions{})
+	if err != nil {
+		return nil, errors.Wrapf(err, "unable to update namespace %s", ns.Name)
+	}
+	return createdNamespace, nil
+}
+
+// CreateService is a convenience function for service setup
+func (k *Scenario) CreateService(service *v1.Service) (*v1.Service, error) {
+	ns := service.Namespace
+	name := service.Name
+
+	createdService, err := k.ClientSet.CoreV1().Services(ns).Create(context.TODO(), service, metav1.CreateOptions{})
+	if err != nil {
+		return nil, errors.Wrapf(err, "unable to create service %s/%s", ns, name)
+	}
+	return createdService, nil
+}
+
+// CreatePod is a convenience function for pod setup
+func (k *Scenario) CreatePod(pod *v1.Pod) (*v1.Pod, error) {
+	ns := pod.Namespace
+	framework.Logf("creating pod %s/%s", ns, pod.Name)
+
+	createdPod, err := k.ClientSet.CoreV1().Pods(ns).Create(context.TODO(), pod, metav1.CreateOptions{})
+	if err != nil {
+		return nil, errors.Wrapf(err, "unable to update pod %s/%s", ns, pod.Name)
+	}
+	return createdPod, nil
+}
+
+// CleanNetworkPolicies is a convenience function for deleting network policies before startup of any new test.
+func (k *Scenario) CleanNetworkPolicies(namespaces []string) error {
+	for _, ns := range namespaces {
+		framework.Logf("deleting policies in %s ..........", ns)
+		l, err := k.ClientSet.NetworkingV1().NetworkPolicies(ns).List(context.TODO(), metav1.ListOptions{})
+		if err != nil {
+			return errors.Wrapf(err, "unable to list network policies in ns %s", ns)
+		}
+		for _, np := range l.Items {
+			framework.Logf("deleting network policy %s/%s", ns, np.Name)
+			err = k.ClientSet.NetworkingV1().NetworkPolicies(ns).Delete(context.TODO(), np.Name, metav1.DeleteOptions{})
+			if err != nil {
+				return errors.Wrapf(err, "unable to delete network policy %s/%s", ns, np.Name)
+			}
+		}
+	}
+	return nil
+}
+
+// ClearCache clears the kube pod cache
+func (k *Scenario) ClearCache() {
+	framework.Logf("Clearing pod cache")
+	k.mutex.Lock()
+	k.podCache = map[string][]v1.Pod{}
+	k.mutex.Unlock()
+	framework.Logf("Pod cache successfully cleared")
+}
+
+// CreateNetworkPolicy is a convenience function for creating netpols
+func (k *Scenario) CreateNetworkPolicy(ns string, netpol *networkingv1.NetworkPolicy) (*networkingv1.NetworkPolicy, error) {
+	framework.Logf("creating network policy %s/%s", ns, netpol.Name)
+	netpol.ObjectMeta.Namespace = ns
+	np, err := k.ClientSet.NetworkingV1().NetworkPolicies(ns).Create(context.TODO(), netpol, metav1.CreateOptions{})
+	if err != nil {
+		return nil, errors.Wrapf(err, "unable to create network policy %s/%s", ns, netpol.Name)
+	}
+	return np, nil
+}
+
+// UpdateNetworkPolicy is a convenience function for updating netpols
+func (k *Scenario) UpdateNetworkPolicy(ns string, netpol *networkingv1.NetworkPolicy) (*networkingv1.NetworkPolicy, error) {
+	framework.Logf("updating network policy %s/%s", ns, netpol.Name)
+	netpol.ObjectMeta.Namespace = ns
+	np, err := k.ClientSet.NetworkingV1().NetworkPolicies(ns).Update(context.TODO(), netpol, metav1.UpdateOptions{})
+	if err != nil {
+		return np, errors.Wrapf(err, "unable to update network policy %s/%s", ns, netpol.Name)
+	}
+	return np, nil
+}
+
+func (k *Scenario) getNamespace(ns string) (*v1.Namespace, error) {
+	selectedNameSpace, err := k.ClientSet.CoreV1().Namespaces().Get(context.TODO(), ns, metav1.GetOptions{})
+	if err != nil {
+		return nil, errors.Wrapf(err, "unable to get namespace %s", ns)
+	}
+	return selectedNameSpace, nil
+}
+
+func (k *Scenario) setNamespaceLabels(ns string, labels map[string]string) error {
+	selectedNameSpace, err := k.getNamespace(ns)
+	if err != nil {
+		return err
+	}
+	selectedNameSpace.ObjectMeta.Labels = labels
+	_, err = k.ClientSet.CoreV1().Namespaces().Update(context.TODO(), selectedNameSpace, metav1.UpdateOptions{})
+	return errors.Wrapf(err, "unable to update namespace %s", ns)
+}
+
+func (k *Scenario) deleteNamespaces(namespaces []string) error {
+	for _, ns := range namespaces {
+		err := k.ClientSet.CoreV1().Namespaces().Delete(context.TODO(), ns, metav1.DeleteOptions{})
+		if err != nil {
+			return errors.Wrapf(err, "unable to delete namespace %s", ns)
+		}
+	}
+	return nil
+}
+
+// waitForHTTPServers waits for all webservers to be up, on all protocols, and then validates them using the same probe logic as the rest of the suite.
+func (k *Scenario) waitForHTTPServers(model *Model) error {
+	const maxTries = 10
+	framework.Logf("waiting for HTTP servers (ports 80 and 81) to become ready")
+
+	testCases := map[string]*TestCase{}
+	for _, port := range model.Ports {
+		for _, protocol := range model.Protocols {
+			fromPort := 81
+			desc := fmt.Sprintf("%d->%d,%s", fromPort, port, protocol)
+			testCases[desc] = &TestCase{FromPort: fromPort, ToPort: int(port), Protocol: protocol}
+		}
+	}
+	notReady := map[string]bool{}
+	for caseName := range testCases {
+		notReady[caseName] = true
+	}
+
+	for i := 0; i < maxTries; i++ {
+		for caseName, testCase := range testCases {
+			if notReady[caseName] {
+				reachability := NewReachability(model.AllPods(), true)
+				testCase.Reachability = reachability
+				ProbePodToPodConnectivity(k, model, testCase)
+				_, wrong, _, _ := reachability.Summary(ignoreLoopback)
+				if wrong == 0 {
+					framework.Logf("server %s is ready", caseName)
+					delete(notReady, caseName)
+				} else {
+					framework.Logf("server %s is not ready", caseName)
+				}
+			}
+		}
+		if len(notReady) == 0 {
+			return nil
+		}
+		time.Sleep(waitInterval)
+	}
+	return errors.Errorf("after %d tries, %d HTTP servers are not ready", maxTries, len(notReady))
+}
diff --git a/test/e2e/network/netpol/model.go b/test/e2e/network/netpol/model.go
new file mode 100644
index 00000000000..3879f11ba2d
--- /dev/null
+++ b/test/e2e/network/netpol/model.go
@@ -0,0 +1,300 @@
+/*
+Copyright 2020 The Kubernetes 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 netpol
+
+import (
+	"fmt"
+	"strings"
+
+	"github.com/pkg/errors"
+	v1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/kubernetes/test/e2e/framework"
+	imageutils "k8s.io/kubernetes/test/utils/image"
+)
+
+// Model defines the namespaces, deployments, services, pods, containers and associated
+// data for network policy test cases and provides the source of truth
+type Model struct {
+	Namespaces    []*Namespace
+	allPodStrings *[]PodString
+	allPods       *[]*Pod
+	// the raw data
+	NamespaceNames []string
+	PodNames       []string
+	Ports          []int32
+	Protocols      []v1.Protocol
+	DNSDomain      string
+}
+
+// NewModel instantiates a model based on:
+// - namespaces
+// - pods
+// - ports to listen on
+// - protocols to listen on
+// The total number of pods is the number of namespaces x the number of pods per namespace.
+// The number of containers per pod is the number of ports x the number of protocols.
+// The *total* number of containers is namespaces x pods x ports x protocols.
+func NewModel(namespaces []string, podNames []string, ports []int32, protocols []v1.Protocol, dnsDomain string) *Model {
+	model := &Model{
+		NamespaceNames: namespaces,
+		PodNames:       podNames,
+		Ports:          ports,
+		Protocols:      protocols,
+		DNSDomain:      dnsDomain,
+	}
+	framework.Logf("DnsDomain %v", model.DNSDomain)
+
+	// build the entire "model" for the overall test, which means, building
+	// namespaces, pods, containers for each protocol.
+	for _, ns := range namespaces {
+		var pods []*Pod
+		for _, podName := range podNames {
+			var containers []*Container
+			for _, port := range ports {
+				for _, protocol := range protocols {
+					containers = append(containers, &Container{
+						Port:     port,
+						Protocol: protocol,
+					})
+				}
+			}
+			pods = append(pods, &Pod{
+				Namespace:  ns,
+				Name:       podName,
+				Containers: containers,
+			})
+		}
+		model.Namespaces = append(model.Namespaces, &Namespace{Name: ns, Pods: pods})
+	}
+	return model
+}
+
+// NewReachability instantiates a default-true reachability from the model's pods
+func (m *Model) NewReachability() *Reachability {
+	return NewReachability(m.AllPods(), true)
+}
+
+// AllPodStrings returns a slice of all pod strings
+func (m *Model) AllPodStrings() []PodString {
+	if m.allPodStrings == nil {
+		var pods []PodString
+		for _, ns := range m.Namespaces {
+			for _, pod := range ns.Pods {
+				pods = append(pods, pod.PodString())
+			}
+		}
+		m.allPodStrings = &pods
+	}
+	return *m.allPodStrings
+}
+
+// AllPods returns a slice of all pods
+func (m *Model) AllPods() []*Pod {
+	if m.allPods == nil {
+		var pods []*Pod
+		for _, ns := range m.Namespaces {
+			for _, pod := range ns.Pods {
+				pods = append(pods, pod)
+			}
+		}
+		m.allPods = &pods
+	}
+	return *m.allPods
+}
+
+// FindPod returns the pod of matching namespace and name, or an error
+func (m *Model) FindPod(ns string, name string) (*Pod, error) {
+	for _, namespace := range m.Namespaces {
+		for _, pod := range namespace.Pods {
+			if namespace.Name == ns && pod.Name == name {
+				return pod, nil
+			}
+		}
+	}
+	return nil, errors.Errorf("unable to find pod %s/%s", ns, name)
+}
+
+// Namespace is the abstract representation of what matters to network policy
+// tests for a namespace; i.e. it ignores kube implementation details
+type Namespace struct {
+	Name string
+	Pods []*Pod
+}
+
+// Spec builds a kubernetes namespace spec
+func (ns *Namespace) Spec() *v1.Namespace {
+	return &v1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:   ns.Name,
+			Labels: ns.LabelSelector(),
+		},
+	}
+}
+
+// LabelSelector returns the default labels that should be placed on a namespace
+// in order for it to be uniquely selectable by label selectors
+func (ns *Namespace) LabelSelector() map[string]string {
+	return map[string]string{"ns": ns.Name}
+}
+
+// Pod is the abstract representation of what matters to network policy tests for
+// a pod; i.e. it ignores kube implementation details
+type Pod struct {
+	Namespace  string
+	Name       string
+	Containers []*Container
+}
+
+// FindContainer returns the container matching port and protocol; otherwise, an error
+func (p *Pod) FindContainer(port int32, protocol v1.Protocol) (*Container, error) {
+	for _, cont := range p.Containers {
+		if cont.Port == port && cont.Protocol == protocol {
+			return cont, nil
+		}
+	}
+	return nil, errors.Errorf("unable to find container in pod %s/%s, port %d, protocol %s", p.Namespace, p.Name, port, protocol)
+}
+
+// PodString returns a corresponding pod string
+func (p *Pod) PodString() PodString {
+	return NewPodString(p.Namespace, p.Name)
+}
+
+// ContainerSpecs builds kubernetes container specs for the pod
+func (p *Pod) ContainerSpecs() []v1.Container {
+	var containers []v1.Container
+	for _, cont := range p.Containers {
+		containers = append(containers, cont.Spec())
+	}
+	return containers
+}
+
+func (p *Pod) labelSelectorKey() string {
+	return "pod"
+}
+
+func (p *Pod) labelSelectorValue() string {
+	return p.Name
+}
+
+// LabelSelector returns the default labels that should be placed on a pod/deployment
+// in order for it to be uniquely selectable by label selectors
+func (p *Pod) LabelSelector() map[string]string {
+	return map[string]string{
+		p.labelSelectorKey(): p.labelSelectorValue(),
+	}
+}
+
+// KubePod returns the kube pod
+func (p *Pod) KubePod() *v1.Pod {
+	zero := int64(0)
+	return &v1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      p.Name,
+			Labels:    p.LabelSelector(),
+			Namespace: p.Namespace,
+		},
+		Spec: v1.PodSpec{
+			TerminationGracePeriodSeconds: &zero,
+			Containers:                    p.ContainerSpecs(),
+		},
+	}
+}
+
+// QualifiedServiceAddress returns the address that can be used to hit a service from
+// any namespace in the cluster
+func (p *Pod) QualifiedServiceAddress(dnsDomain string) string {
+	return fmt.Sprintf("%s.%s.svc.%s", p.ServiceName(), p.Namespace, dnsDomain)
+}
+
+// ServiceName returns the unqualified service name
+func (p *Pod) ServiceName() string {
+	return fmt.Sprintf("s-%s-%s", p.Namespace, p.Name)
+}
+
+// Service returns a kube service spec
+func (p *Pod) Service() *v1.Service {
+	service := &v1.Service{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      p.ServiceName(),
+			Namespace: p.Namespace,
+		},
+		Spec: v1.ServiceSpec{
+			Selector: p.LabelSelector(),
+		},
+	}
+	for _, container := range p.Containers {
+		service.Spec.Ports = append(service.Spec.Ports, v1.ServicePort{
+			Name:     fmt.Sprintf("service-port-%s-%d", strings.ToLower(string(container.Protocol)), container.Port),
+			Protocol: container.Protocol,
+			Port:     container.Port,
+		})
+	}
+	return service
+}
+
+// Container is the abstract representation of what matters to network policy tests for
+// a container; i.e. it ignores kube implementation details
+type Container struct {
+	Port     int32
+	Protocol v1.Protocol
+}
+
+// Name returns the container name
+func (c *Container) Name() string {
+	return fmt.Sprintf("cont-%d-%s", c.Port, strings.ToLower(string(c.Protocol)))
+}
+
+// PortName returns the container port name
+func (c *Container) PortName() string {
+	return fmt.Sprintf("serve-%d-%s", c.Port, strings.ToLower(string(c.Protocol)))
+}
+
+// Spec returns the kube container spec
+func (c *Container) Spec() v1.Container {
+	var (
+		// agnHostImage is the image URI of AgnHost
+		agnHostImage = imageutils.GetE2EImage(imageutils.Agnhost)
+		cmd          []string
+	)
+
+	switch c.Protocol {
+	case v1.ProtocolTCP:
+		cmd = []string{"/agnhost", "serve-hostname", "--tcp", "--http=false", "--port", fmt.Sprintf("%d", c.Port)}
+	case v1.ProtocolUDP:
+		cmd = []string{"/agnhost", "serve-hostname", "--udp", "--http=false", "--port", fmt.Sprintf("%d", c.Port)}
+	case v1.ProtocolSCTP:
+		cmd = []string{"/agnhost", "netexec", "--sctp-port", fmt.Sprintf("%d", c.Port)}
+	default:
+		framework.Failf("invalid protocol %v", c.Protocol)
+	}
+	return v1.Container{
+		Name:            c.Name(),
+		ImagePullPolicy: v1.PullIfNotPresent,
+		Image:           agnHostImage,
+		Command:         cmd,
+		SecurityContext: &v1.SecurityContext{},
+		Ports: []v1.ContainerPort{
+			{
+				ContainerPort: c.Port,
+				Name:          c.PortName(),
+				Protocol:      c.Protocol,
+			},
+		},
+	}
+}
diff --git a/test/e2e/network/netpol/network_policy.go b/test/e2e/network/netpol/network_policy.go
new file mode 100644
index 00000000000..be6d36b0235
--- /dev/null
+++ b/test/e2e/network/netpol/network_policy.go
@@ -0,0 +1,246 @@
+/*
+Copyright 2016 The Kubernetes 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 netpol
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"k8s.io/apimachinery/pkg/util/wait"
+
+	v1 "k8s.io/api/core/v1"
+	networkingv1 "k8s.io/api/networking/v1"
+	"k8s.io/kubernetes/test/e2e/network"
+
+	"github.com/onsi/ginkgo"
+
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/kubernetes/test/e2e/framework"
+)
+
+const (
+	addSCTPContainers = false
+	isVerbose         = true
+
+	// useFixedNamespaces is useful when working on these tests: instead of creating new pods and
+	//   new namespaces for each test run, it creates a fixed set of namespaces and pods, and then
+	//   reuses them for each test case.
+	// The result: tests run much faster.  However, this should only be used as a convenience for
+	//   working on the tests during development.  It should not be enabled in production.
+	useFixedNamespaces = false
+
+	// See https://github.com/kubernetes/kubernetes/issues/95879
+	// The semantics of the effect of network policies on loopback calls may be undefined: should
+	//   they always be ALLOWED; how do Services affect this?
+	//   Calico, Cillium, Antrea seem to do different things.
+	// Since different CNIs have different results, that causes tests including loopback to fail
+	//   on some CNIs.  So let's just ignore loopback calls for the purposes of deciding test pass/fail.
+	ignoreLoopback = true
+)
+
+/*
+You might be wondering, why are there multiple namespaces used for each test case?
+
+These tests are based on "truth tables" that compare the expected and actual connectivity of each pair of pods.
+Since network policies live in namespaces, and peers can be selected by namespace,
+howing the connectivity of pods in other namespaces is key information to show whether a network policy is working as intended or not.
+
+We use 3 namespaces each with 3 pods, and probe all combinations ( 9 pods x 9 pods = 81 data points ) -- including cross-namespace calls.
+
+Here's an example of a test run, showing the expected and actual connectivity, along with the differences.  Note how the
+visual representation as a truth table greatly aids in understanding what a network policy is intended to do in theory
+and what is happening in practice:
+
+		Oct 19 10:34:16.907: INFO: expected:
+
+		-	x/a	x/b	x/c	y/a	y/b	y/c	z/a	z/b	z/c
+		x/a	X	.	.	.	.	.	.	.	.
+		x/b	X	.	.	.	.	.	.	.	.
+		x/c	X	.	.	.	.	.	.	.	.
+		y/a	.	.	.	.	.	.	.	.	.
+		y/b	.	.	.	.	.	.	.	.	.
+		y/c	.	.	.	.	.	.	.	.	.
+		z/a	X	.	.	.	.	.	.	.	.
+		z/b	X	.	.	.	.	.	.	.	.
+		z/c	X	.	.	.	.	.	.	.	.
+
+		Oct 19 10:34:16.907: INFO: observed:
+
+		-	x/a	x/b	x/c	y/a	y/b	y/c	z/a	z/b	z/c
+		x/a	X	.	.	.	.	.	.	.	.
+		x/b	X	.	.	.	.	.	.	.	.
+		x/c	X	.	.	.	.	.	.	.	.
+		y/a	.	.	.	.	.	.	.	.	.
+		y/b	.	.	.	.	.	.	.	.	.
+		y/c	.	.	.	.	.	.	.	.	.
+		z/a	X	.	.	.	.	.	.	.	.
+		z/b	X	.	.	.	.	.	.	.	.
+		z/c	X	.	.	.	.	.	.	.	.
+
+		Oct 19 10:34:16.907: INFO: comparison:
+
+		-	x/a	x/b	x/c	y/a	y/b	y/c	z/a	z/b	z/c
+		x/a	.	.	.	.	.	.	.	.	.
+		x/b	.	.	.	.	.	.	.	.	.
+		x/c	.	.	.	.	.	.	.	.	.
+		y/a	.	.	.	.	.	.	.	.	.
+		y/b	.	.	.	.	.	.	.	.	.
+		y/c	.	.	.	.	.	.	.	.	.
+		z/a	.	.	.	.	.	.	.	.	.
+		z/b	.	.	.	.	.	.	.	.	.
+		z/c	.	.	.	.	.	.	.	.	.
+*/
+var _ = network.SIGDescribe("Netpol [LinuxOnly]", func() {
+	f := framework.NewDefaultFramework("netpol")
+
+	ginkgo.Context("NetworkPolicy between server and client", func() {
+		ginkgo.BeforeEach(func() {
+			if useFixedNamespaces {
+				_ = initializeResources(f)
+
+				_, _, _, model, k8s := getK8SModel(f)
+				framework.ExpectNoError(k8s.CleanNetworkPolicies(model.NamespaceNames), "unable to clean network policies")
+				err := wait.Poll(waitInterval, waitTimeout, func() (done bool, err error) {
+					for _, ns := range model.NamespaceNames {
+						netpols, err := k8s.ClientSet.NetworkingV1().NetworkPolicies(ns).List(context.TODO(), metav1.ListOptions{})
+						framework.ExpectNoError(err, "get network policies from ns %s", ns)
+						if len(netpols.Items) > 0 {
+							return false, nil
+						}
+					}
+					return true, nil
+				})
+				framework.ExpectNoError(err, "unable to wait for network policy deletion")
+			} else {
+				framework.Logf("Using %v as the default dns domain for this cluster... ", framework.TestContext.ClusterDNSDomain)
+				framework.ExpectNoError(initializeResources(f), "unable to initialize resources")
+			}
+		})
+
+		ginkgo.AfterEach(func() {
+			if !useFixedNamespaces {
+				_, _, _, model, k8s := getK8SModel(f)
+				framework.ExpectNoError(k8s.deleteNamespaces(model.NamespaceNames), "unable to clean up netpol namespaces")
+			}
+		})
+
+		ginkgo.It("should support a 'default-deny-ingress' policy [Feature:NetworkPolicy]", func() {
+			nsX, _, _, model, k8s := getK8SModel(f)
+			policy := GetDenyIngress("deny-ingress")
+			CreatePolicy(k8s, policy, nsX)
+
+			reachability := NewReachability(model.AllPods(), true)
+			reachability.ExpectPeer(&Peer{}, &Peer{Namespace: nsX}, false)
+
+			ValidateOrFail(k8s, model, &TestCase{FromPort: 81, ToPort: 80, Protocol: v1.ProtocolTCP, Reachability: reachability})
+		})
+
+		ginkgo.It("should support a 'default-deny-all' policy [Feature:NetworkPolicy]", func() {
+			np := &networkingv1.NetworkPolicy{}
+			policy := `
+			{
+				"kind": "NetworkPolicy",
+				"apiVersion": "networking.k8s.io/v1",
+				"metadata": {
+				   "name": "deny-all-tcp-allow-dns"
+				},
+				"spec": {
+				   "podSelector": {
+					  "matchLabels": {}
+				   },
+				   "ingress": [],
+				   "egress": [{
+						"ports": [
+							{
+								"protocol": "UDP",
+								"port": 53
+							}
+						]
+					}],
+				   "policyTypes": [
+					"Ingress",
+					"Egress"
+				   ]
+				}
+			 }
+			 `
+			err := json.Unmarshal([]byte(policy), np)
+			framework.ExpectNoError(err, "unmarshal network policy")
+
+			nsX, _, _, model, k8s := getK8SModel(f)
+			CreatePolicy(k8s, np, nsX)
+
+			reachability := NewReachability(model.AllPods(), true)
+			reachability.ExpectPeer(&Peer{}, &Peer{Namespace: nsX}, false)
+			reachability.ExpectPeer(&Peer{Namespace: nsX}, &Peer{}, false)
+
+			ValidateOrFail(k8s, model, &TestCase{FromPort: 81, ToPort: 80, Protocol: v1.ProtocolTCP, Reachability: reachability})
+		})
+	})
+})
+
+// getNamespaces returns the canonical set of namespaces used by this test, taking a root ns as input.  This allows this test to run in parallel.
+func getNamespaces(rootNs string) (string, string, string, []string) {
+	if useFixedNamespaces {
+		rootNs = ""
+	} else {
+		rootNs = rootNs + "-"
+	}
+	nsX := fmt.Sprintf("%sx", rootNs)
+	nsY := fmt.Sprintf("%sy", rootNs)
+	nsZ := fmt.Sprintf("%sz", rootNs)
+	return nsX, nsY, nsZ, []string{nsX, nsY, nsZ}
+}
+
+// defaultModel creates a new "model" pod system under namespaces (x,y,z) which has pods a, b, and c.  Thus resulting in the
+// truth table matrix that is identical for all tests, comprising 81 total connections between 9 pods (x/a, x/b, x/c, ..., z/c).
+func defaultModel(namespaces []string, dnsDomain string) *Model {
+	protocols := []v1.Protocol{v1.ProtocolTCP, v1.ProtocolUDP}
+	if addSCTPContainers {
+		protocols = append(protocols, v1.ProtocolSCTP)
+	}
+	return NewModel(namespaces, []string{"a", "b", "c"}, []int32{80, 81}, protocols, dnsDomain)
+}
+
+// getK8sModel uses the e2e framework to create all necessary namespace resources, and returns the default probing model used
+// in the scaffold of this test.
+func getK8SModel(f *framework.Framework) (string, string, string, *Model, *Scenario) {
+	k8s := NewScenario(f)
+	rootNs := f.Namespace.GetName()
+	nsX, nsY, nsZ, namespaces := getNamespaces(rootNs)
+
+	model := defaultModel(namespaces, framework.TestContext.ClusterDNSDomain)
+
+	return nsX, nsY, nsZ, model, k8s
+}
+
+// initializeResources generates a model and then waits for the model to be up-and-running (i.e. all pods are ready and running in their namespaces).
+func initializeResources(f *framework.Framework) error {
+	_, _, _, model, k8s := getK8SModel(f)
+
+	framework.Logf("initializing cluster: ensuring namespaces, deployments, and pods exist and are ready")
+
+	err := k8s.InitializeCluster(model)
+	if err != nil {
+		return err
+	}
+
+	framework.Logf("finished initializing cluster state")
+
+	return k8s.waitForHTTPServers(model)
+}
diff --git a/test/e2e/network/netpol/policies.go b/test/e2e/network/netpol/policies.go
new file mode 100644
index 00000000000..425f3e7a474
--- /dev/null
+++ b/test/e2e/network/netpol/policies.go
@@ -0,0 +1,429 @@
+/*
+Copyright 2020 The Kubernetes 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 netpol
+
+import (
+	"fmt"
+
+	"k8s.io/apimachinery/pkg/util/intstr"
+
+	v1 "k8s.io/api/core/v1"
+	networkingv1 "k8s.io/api/networking/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+)
+
+// GetDenyIngress returns a default deny policy named 'name'.
+func GetDenyIngress(name string) *networkingv1.NetworkPolicy {
+	return &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{},
+			Ingress:     []networkingv1.NetworkPolicyIngressRule{},
+		},
+	}
+}
+
+// GetRandomIngressPolicies returns "num" random policies that allow a unique:n label, i.e.
+// unique:1, unique:2, and so on.  Used for creating a 'background' set of policies.
+func GetRandomIngressPolicies(num int) []*networkingv1.NetworkPolicy {
+	policies := []*networkingv1.NetworkPolicy{}
+
+	for i := 0; i < num; i++ {
+		policy := &networkingv1.NetworkPolicy{
+			ObjectMeta: metav1.ObjectMeta{
+				Name: fmt.Sprintf("allow-all-%v", i),
+			},
+			Spec: networkingv1.NetworkPolicySpec{
+				PodSelector: metav1.LabelSelector{
+					MatchLabels: map[string]string{
+						"unique": fmt.Sprintf("%v", i),
+					},
+				},
+				Ingress: []networkingv1.NetworkPolicyIngressRule{{}},
+			},
+		}
+		policies = append(policies, policy)
+	}
+	return policies
+}
+
+// GetAllowIngress allows all ingress
+func GetAllowIngress(name string) *networkingv1.NetworkPolicy {
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: map[string]string{},
+			},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{
+				{},
+			},
+		},
+	}
+	return policy
+}
+
+// GetAllowIngressByPort allows ingress by port
+func GetAllowIngressByPort(name string, port *intstr.IntOrString) *networkingv1.NetworkPolicy {
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: map[string]string{},
+			},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{
+				{
+					Ports: []networkingv1.NetworkPolicyPort{
+						{Port: port},
+					},
+				},
+			},
+		},
+	}
+	return policy
+}
+
+// GetAllowEgressByPort allows egress by port
+func GetAllowEgressByPort(name string, port *intstr.IntOrString) *networkingv1.NetworkPolicy {
+	protocolUDP := v1.ProtocolUDP
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: map[string]string{},
+			},
+			Egress: []networkingv1.NetworkPolicyEgressRule{
+				{
+					Ports: []networkingv1.NetworkPolicyPort{
+						{Port: port},
+						{
+							Protocol: &protocolUDP,
+							Port:     &intstr.IntOrString{Type: intstr.Int, IntVal: 53},
+						},
+					},
+				},
+			},
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeEgress},
+		},
+	}
+	return policy
+}
+
+// GetDenyAll denies ingress traffic, AS WELL as egress traffic.
+// - BOTH policy types must be specified
+// - The Egress rule must (like the ingress default rule) be a array with 0 values.
+func GetDenyAll(name string) *networkingv1.NetworkPolicy {
+	policy := GetDenyIngress(name)
+	policy.Spec.PolicyTypes = []networkingv1.PolicyType{networkingv1.PolicyTypeEgress, networkingv1.PolicyTypeIngress}
+	policy.Spec.Egress = []networkingv1.NetworkPolicyEgressRule{}
+	policy.Spec.Ingress = []networkingv1.NetworkPolicyIngressRule{}
+	return policy
+}
+
+// GetAllowIngressByPod allows ingress by pod labels
+func GetAllowIngressByPod(name string, targetLabels map[string]string, peerPodSelector *metav1.LabelSelector) *networkingv1.NetworkPolicy {
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: targetLabels,
+			},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{{
+				From: []networkingv1.NetworkPolicyPeer{{
+					PodSelector: peerPodSelector,
+				}},
+			}},
+		},
+	}
+	return policy
+}
+
+// GetDenyIngressForTarget denies all ingress for target
+func GetDenyIngressForTarget(targetSelector metav1.LabelSelector) *networkingv1.NetworkPolicy {
+	return &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "deny-ingress-via-label-selector",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: targetSelector,
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeIngress},
+			Ingress:     []networkingv1.NetworkPolicyIngressRule{},
+		},
+	}
+}
+
+// GetAllowIngressByNamespace allows ingress for namespace
+func GetAllowIngressByNamespace(name string, targetLabels map[string]string, peerNamespaceSelector *metav1.LabelSelector) *networkingv1.NetworkPolicy {
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: targetLabels,
+			},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{{
+				From: []networkingv1.NetworkPolicyPeer{{
+					NamespaceSelector: peerNamespaceSelector,
+				}},
+			}},
+		},
+	}
+	return policy
+}
+
+// GetAllowIngressByNamespaceAndPort allows ingress for namespace AND port
+func GetAllowIngressByNamespaceAndPort(name string, targetLabels map[string]string, peerNamespaceSelector *metav1.LabelSelector, port *intstr.IntOrString) *networkingv1.NetworkPolicy {
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: targetLabels,
+			},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{{
+				From: []networkingv1.NetworkPolicyPeer{{
+					NamespaceSelector: peerNamespaceSelector,
+				}},
+				Ports: []networkingv1.NetworkPolicyPort{
+					{Port: port},
+				},
+			}},
+		},
+	}
+	return policy
+}
+
+// GetAllowIngressByNamespaceOrPod allows ingress for pods with matching namespace OR pod labels
+func GetAllowIngressByNamespaceOrPod(name string, targetLabels map[string]string, peerNamespaceSelector *metav1.LabelSelector, peerPodSelector *metav1.LabelSelector) *networkingv1.NetworkPolicy {
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: targetLabels,
+			},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{{
+				From: []networkingv1.NetworkPolicyPeer{
+					{
+						NamespaceSelector: peerNamespaceSelector,
+					},
+					{
+						PodSelector: peerPodSelector,
+					},
+				},
+			}},
+		},
+	}
+	return policy
+}
+
+// GetAllowIngressByNamespaceAndPod allows ingress for pods with matching namespace AND pod labels
+func GetAllowIngressByNamespaceAndPod(name string, targetLabels map[string]string, peerNamespaceSelector *metav1.LabelSelector, peerPodSelector *metav1.LabelSelector) *networkingv1.NetworkPolicy {
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: targetLabels,
+			},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{{
+				From: []networkingv1.NetworkPolicyPeer{{
+					NamespaceSelector: peerNamespaceSelector,
+					PodSelector:       peerPodSelector,
+				}},
+			}},
+		},
+	}
+	return policy
+}
+
+// GetAllowEgressByNamespaceAndPod allows egress for pods with matching namespace AND pod labels
+func GetAllowEgressByNamespaceAndPod(name string, targetLabels map[string]string, peerNamespaceSelector *metav1.LabelSelector, peerPodSelector *metav1.LabelSelector) *networkingv1.NetworkPolicy {
+	protocolUDP := v1.ProtocolUDP
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: targetLabels,
+			},
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeEgress},
+			Egress: []networkingv1.NetworkPolicyEgressRule{
+				{
+					To: []networkingv1.NetworkPolicyPeer{
+						{
+							NamespaceSelector: peerNamespaceSelector,
+							PodSelector:       peerPodSelector,
+						},
+					},
+				},
+				{
+					Ports: []networkingv1.NetworkPolicyPort{
+						{
+							Protocol: &protocolUDP,
+							Port:     &intstr.IntOrString{Type: intstr.Int, IntVal: 53},
+						},
+					},
+				},
+			},
+		},
+	}
+	return policy
+}
+
+// GetAllowEgress allows all egress
+func GetAllowEgress() *networkingv1.NetworkPolicy {
+	return &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "allow-egress",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: map[string]string{},
+			},
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeEgress},
+			Egress:      []networkingv1.NetworkPolicyEgressRule{{}},
+		},
+	}
+}
+
+// GetAllowEgressForTarget allows all egress for a target
+func GetAllowEgressForTarget(targetSelector metav1.LabelSelector) *networkingv1.NetworkPolicy {
+	return &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "allow-egress-for-target",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: targetSelector,
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeEgress},
+			Egress:      []networkingv1.NetworkPolicyEgressRule{{}},
+		},
+	}
+}
+
+// GetAllowEgressByCIDR creates an egress netpol with an ipblock
+func GetAllowEgressByCIDR(podname string, podserverCIDR string) *networkingv1.NetworkPolicy {
+	protocolUDP := v1.ProtocolUDP
+	return &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "allow-client-a-via-cidr-egress-rule",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: map[string]string{
+					"pod": podname,
+				},
+			},
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeEgress},
+			// Allow traffic to only one CIDR block.
+			Egress: []networkingv1.NetworkPolicyEgressRule{
+				{
+					To: []networkingv1.NetworkPolicyPeer{
+						{
+							IPBlock: &networkingv1.IPBlock{
+								CIDR: podserverCIDR,
+							},
+						},
+					},
+				},
+				{
+					Ports: []networkingv1.NetworkPolicyPort{
+						{
+							Protocol: &protocolUDP,
+							Port:     &intstr.IntOrString{Type: intstr.Int, IntVal: 53},
+						},
+					},
+				},
+			},
+		},
+	}
+}
+
+// GetAllowEgressByCIDRExcept creates an egress netpol with an ipblock and except
+func GetAllowEgressByCIDRExcept(podname string, podserverCIDR string, except []string) *networkingv1.NetworkPolicy {
+	protocolUDP := v1.ProtocolUDP
+	return &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "allow-client-a-via-cidr-egress-rule",
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: map[string]string{
+					"pod": podname,
+				},
+			},
+			PolicyTypes: []networkingv1.PolicyType{networkingv1.PolicyTypeEgress},
+			// Allow traffic to only one CIDR block.
+			Egress: []networkingv1.NetworkPolicyEgressRule{
+				{
+					To: []networkingv1.NetworkPolicyPeer{
+						{
+							IPBlock: &networkingv1.IPBlock{
+								CIDR:   podserverCIDR,
+								Except: except,
+							},
+						},
+					},
+				},
+				{
+					Ports: []networkingv1.NetworkPolicyPort{
+						{
+							Protocol: &protocolUDP,
+							Port:     &intstr.IntOrString{Type: intstr.Int, IntVal: 53},
+						},
+					},
+				},
+			},
+		},
+	}
+}
+
+// GetAllowIngressOnProtocolByPort is a base network policy template which distinguishes between the types of v1.Protocol available in v1 core
+func GetAllowIngressOnProtocolByPort(name string, protocol v1.Protocol, targetLabels map[string]string, portNum *intstr.IntOrString) *networkingv1.NetworkPolicy {
+	policy := &networkingv1.NetworkPolicy{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: name,
+		},
+		Spec: networkingv1.NetworkPolicySpec{
+			PodSelector: metav1.LabelSelector{
+				MatchLabels: targetLabels,
+			},
+			Ingress: []networkingv1.NetworkPolicyIngressRule{{
+				Ports: []networkingv1.NetworkPolicyPort{{
+					Port:     portNum,
+					Protocol: &protocol,
+				}},
+			}},
+		},
+	}
+	return policy
+}
diff --git a/test/e2e/network/netpol/probe.go b/test/e2e/network/netpol/probe.go
new file mode 100644
index 00000000000..5b4a46e8219
--- /dev/null
+++ b/test/e2e/network/netpol/probe.go
@@ -0,0 +1,117 @@
+/*
+Copyright 2020 The Kubernetes 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 netpol
+
+import (
+	"github.com/onsi/ginkgo"
+	v1 "k8s.io/api/core/v1"
+	"k8s.io/kubernetes/test/e2e/framework"
+)
+
+// ProbeJob packages the data for the input of a pod->pod connectivity probe
+type ProbeJob struct {
+	PodFrom        *Pod
+	PodTo          *Pod
+	FromPort       int
+	ToPort         int
+	ToPodDNSDomain string
+	Protocol       v1.Protocol
+}
+
+// ProbeJobResults packages the data for the results of a pod->pod connectivity probe
+type ProbeJobResults struct {
+	Job         *ProbeJob
+	IsConnected bool
+	Err         error
+	Command     string
+}
+
+// ProbePodToPodConnectivity runs a series of probes in kube, and records the results in `testCase.Reachability`
+func ProbePodToPodConnectivity(k8s *Scenario, model *Model, testCase *TestCase) {
+	k8s.ClearCache()
+	numberOfWorkers := 30
+	allPods := model.AllPods()
+	size := len(allPods) * len(allPods)
+	jobs := make(chan *ProbeJob, size)
+	results := make(chan *ProbeJobResults, size)
+	for i := 0; i < numberOfWorkers; i++ {
+		go probeWorker(k8s, jobs, results)
+	}
+	for _, podFrom := range allPods {
+		for _, podTo := range allPods {
+			jobs <- &ProbeJob{
+				PodFrom:        podFrom,
+				PodTo:          podTo,
+				FromPort:       testCase.FromPort,
+				ToPort:         testCase.ToPort,
+				ToPodDNSDomain: model.DNSDomain,
+				Protocol:       testCase.Protocol,
+			}
+		}
+	}
+	close(jobs)
+
+	for i := 0; i < size; i++ {
+		result := <-results
+		job := result.Job
+		if result.Err != nil {
+			framework.Logf("unable to perform probe %s -> %s: %v", job.PodFrom.PodString(), job.PodTo.PodString(), result.Err)
+		}
+		testCase.Reachability.Observe(job.PodFrom.PodString(), job.PodTo.PodString(), result.IsConnected)
+		expected := testCase.Reachability.Expected.Get(job.PodFrom.PodString().String(), job.PodTo.PodString().String())
+		if result.IsConnected != expected {
+			framework.Logf("Validation of %s -> %s FAILED !!!", job.PodFrom.PodString(), job.PodTo.PodString())
+			framework.Logf("error %v ", result.Err)
+			if expected {
+				framework.Logf("Expected allowed pod connection was instead BLOCKED --- run '%v'", result.Command)
+			} else {
+				framework.Logf("Expected blocked pod connection was instead ALLOWED --- run '%v'", result.Command)
+			}
+		}
+	}
+}
+
+// probeWorker continues polling a pod connectivity status, until the incoming "jobs" channel is closed, and writes results back out to the "results" channel.
+// it only writes pass/fail status to a channel and has no failure side effects, this is by design since we do not want to fail inside a goroutine.
+func probeWorker(k8s *Scenario, jobs <-chan *ProbeJob, results chan<- *ProbeJobResults) {
+	defer ginkgo.GinkgoRecover()
+	for job := range jobs {
+		podFrom := job.PodFrom
+		containerFrom, err := podFrom.FindContainer(int32(job.FromPort), job.Protocol)
+		// 1) sanity check that the pod container is found before we run the real test.
+		if err != nil {
+			result := &ProbeJobResults{
+				Job:         job,
+				IsConnected: false,
+				Err:         err,
+				Command:     "(skipped, pod unavailable)",
+			}
+			results <- result
+		} else {
+			// 2) real test runs here...
+			connected, command, err := k8s.Probe(podFrom.Namespace, podFrom.Name, containerFrom.Name(), job.PodTo.QualifiedServiceAddress(job.ToPodDNSDomain), job.Protocol, job.ToPort)
+			result := &ProbeJobResults{
+				Job:         job,
+				IsConnected: connected,
+				Err:         err,
+				Command:     command,
+			}
+			results <- result
+		}
+	}
+
+}
diff --git a/test/e2e/network/netpol/reachability.go b/test/e2e/network/netpol/reachability.go
new file mode 100644
index 00000000000..1c4090acc72
--- /dev/null
+++ b/test/e2e/network/netpol/reachability.go
@@ -0,0 +1,188 @@
+/*
+Copyright 2020 The Kubernetes 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 netpol
+
+import (
+	"fmt"
+	v1 "k8s.io/api/core/v1"
+	"k8s.io/kubernetes/test/e2e/framework"
+	"strings"
+)
+
+// TestCase describes the data for a netpol test
+type TestCase struct {
+	FromPort     int
+	ToPort       int
+	Protocol     v1.Protocol
+	Reachability *Reachability
+}
+
+// PodString represents a namespace 'x' + pod 'a' as "x/a".
+type PodString string
+
+// NewPodString instantiates a PodString from the given namespace and name.
+func NewPodString(namespace string, podName string) PodString {
+	return PodString(fmt.Sprintf("%s/%s", namespace, podName))
+}
+
+// String converts back to a string
+func (pod PodString) String() string {
+	return string(pod)
+}
+
+func (pod PodString) split() (string, string) {
+	pieces := strings.Split(string(pod), "/")
+	if len(pieces) != 2 {
+		framework.Failf("expected ns/pod, found %+v", pieces)
+	}
+	return pieces[0], pieces[1]
+}
+
+// Namespace extracts the namespace
+func (pod PodString) Namespace() string {
+	ns, _ := pod.split()
+	return ns
+}
+
+// PodName extracts the pod name
+func (pod PodString) PodName() string {
+	_, podName := pod.split()
+	return podName
+}
+
+// Peer is used for matching pods by either or both of the pod's namespace and name.
+type Peer struct {
+	Namespace string
+	Pod       string
+}
+
+// Matches checks whether the Peer matches the PodString:
+// - an empty namespace means the namespace will always match
+// - otherwise, the namespace must match the PodString's namespace
+// - same goes for Pod: empty matches everything, otherwise must match exactly
+func (p *Peer) Matches(pod PodString) bool {
+	return (p.Namespace == "" || p.Namespace == pod.Namespace()) && (p.Pod == "" || p.Pod == pod.PodName())
+}
+
+// Reachability packages the data for a cluster-wide connectivity probe
+type Reachability struct {
+	Expected *TruthTable
+	Observed *TruthTable
+	Pods     []*Pod
+}
+
+// NewReachability instantiates a reachability
+func NewReachability(pods []*Pod, defaultExpectation bool) *Reachability {
+	var podNames []string
+	for _, pod := range pods {
+		podNames = append(podNames, pod.PodString().String())
+	}
+	r := &Reachability{
+		Expected: NewTruthTableFromItems(podNames, &defaultExpectation),
+		Observed: NewTruthTableFromItems(podNames, nil),
+		Pods:     pods,
+	}
+	return r
+}
+
+// AllowLoopback expects all communication from a pod to itself to be allowed.
+// In general, call it after setting up any other rules since loopback logic follows no policy.
+func (r *Reachability) AllowLoopback() {
+	for _, pod := range r.Pods {
+		podName := pod.PodString().String()
+		r.Expected.Set(podName, podName, true)
+	}
+}
+
+// Expect sets the expected value for a single observation
+func (r *Reachability) Expect(from PodString, to PodString, isConnected bool) {
+	r.Expected.Set(string(from), string(to), isConnected)
+}
+
+// ExpectAllIngress defines that any traffic going into the pod will be allowed/denied (true/false)
+func (r *Reachability) ExpectAllIngress(pod PodString, connected bool) {
+	r.Expected.SetAllTo(string(pod), connected)
+	if !connected {
+		framework.Logf("Denying all traffic *to* %s", pod)
+	}
+}
+
+// ExpectAllEgress defines that any traffic going out of the pod will be allowed/denied (true/false)
+func (r *Reachability) ExpectAllEgress(pod PodString, connected bool) {
+	r.Expected.SetAllFrom(string(pod), connected)
+	if !connected {
+		framework.Logf("Denying all traffic *from* %s", pod)
+	}
+}
+
+// ExpectPeer sets expected values using Peer matchers
+func (r *Reachability) ExpectPeer(from *Peer, to *Peer, connected bool) {
+	for _, fromPod := range r.Pods {
+		if from.Matches(fromPod.PodString()) {
+			for _, toPod := range r.Pods {
+				if to.Matches(toPod.PodString()) {
+					r.Expected.Set(string(fromPod.PodString()), string(toPod.PodString()), connected)
+				}
+			}
+		}
+	}
+}
+
+// Observe records a single connectivity observation
+func (r *Reachability) Observe(fromPod PodString, toPod PodString, isConnected bool) {
+	r.Observed.Set(string(fromPod), string(toPod), isConnected)
+}
+
+// Summary produces a useful summary of expected and observed data
+func (r *Reachability) Summary(ignoreLoopback bool) (trueObs int, falseObs int, ignoredObs int, comparison *TruthTable) {
+	comparison = r.Expected.Compare(r.Observed)
+	if !comparison.IsComplete() {
+		framework.Failf("observations not complete!")
+	}
+	falseObs, trueObs, ignoredObs = 0, 0, 0
+	for from, dict := range comparison.Values {
+		for to, val := range dict {
+			if ignoreLoopback && from == to {
+				// Never fail on loopback, because its not yet defined.
+				ignoredObs++
+			} else if val {
+				trueObs++
+			} else {
+				falseObs++
+			}
+		}
+	}
+	return
+}
+
+// PrintSummary prints the summary
+func (r *Reachability) PrintSummary(printExpected bool, printObserved bool, printComparison bool) {
+	right, wrong, ignored, comparison := r.Summary(ignoreLoopback)
+	if ignored > 0 {
+		framework.Logf("warning: the results of %d pod->pod cases have been ignored", ignored)
+	}
+	framework.Logf("reachability: correct:%v, incorrect:%v, result=%t\n\n", right, wrong, wrong == 0)
+	if printExpected {
+		framework.Logf("expected:\n\n%s\n\n\n", r.Expected.PrettyPrint(""))
+	}
+	if printObserved {
+		framework.Logf("observed:\n\n%s\n\n\n", r.Observed.PrettyPrint(""))
+	}
+	if printComparison {
+		framework.Logf("comparison:\n\n%s\n\n\n", comparison.PrettyPrint(""))
+	}
+}
diff --git a/test/e2e/network/netpol/test_helper.go b/test/e2e/network/netpol/test_helper.go
new file mode 100644
index 00000000000..75010ea8253
--- /dev/null
+++ b/test/e2e/network/netpol/test_helper.go
@@ -0,0 +1,164 @@
+/*
+Copyright 2020 The Kubernetes 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 netpol
+
+import (
+	"context"
+	"time"
+
+	"github.com/onsi/ginkgo"
+	networkingv1 "k8s.io/api/networking/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/util/wait"
+	"k8s.io/kubernetes/test/e2e/framework"
+	"sigs.k8s.io/yaml"
+)
+
+const (
+	waitInterval = 1 * time.Second
+	waitTimeout  = 30 * time.Second
+)
+
+// prettyPrint a networkPolicy
+func prettyPrint(policy *networkingv1.NetworkPolicy) string {
+	raw, err := yaml.Marshal(policy)
+	framework.ExpectNoError(err, "marshal network policy to yaml")
+	return string(raw)
+}
+
+// CreatePolicy creates a policy in the given namespace
+func CreatePolicy(k8s *Scenario, policy *networkingv1.NetworkPolicy, namespace string) {
+	if isVerbose {
+		framework.Logf("****************************************************************")
+		framework.Logf("Network Policy creating %s/%s \n%s", namespace, policy.Name, prettyPrint(policy))
+		framework.Logf("****************************************************************")
+	}
+
+	_, err := k8s.CreateNetworkPolicy(namespace, policy)
+	framework.ExpectNoError(err, "Unable to create netpol %s/%s", namespace, policy.Name)
+}
+
+// UpdatePolicy updates a networkpolicy
+func UpdatePolicy(k8s *Scenario, policy *networkingv1.NetworkPolicy, namespace string) {
+	if isVerbose {
+		framework.Logf("****************************************************************")
+		framework.Logf("Network Policy updating %s/%s \n%s", namespace, policy.Name, prettyPrint(policy))
+		framework.Logf("****************************************************************")
+	}
+
+	_, err := k8s.UpdateNetworkPolicy(namespace, policy)
+	framework.ExpectNoError(err, "Unable to update netpol %s/%s", namespace, policy.Name)
+}
+
+// ValidateOrFail validates connectivity
+func ValidateOrFail(k8s *Scenario, model *Model, testCase *TestCase) {
+	ginkgo.By("Validating reachability matrix...")
+
+	// 1st try
+	ginkgo.By("Validating reachability matrix... (FIRST TRY)")
+	ProbePodToPodConnectivity(k8s, model, testCase)
+	// 2nd try, in case first one failed
+	if _, wrong, _, _ := testCase.Reachability.Summary(ignoreLoopback); wrong != 0 {
+		framework.Logf("failed first probe %d wrong results ... retrying (SECOND TRY)", wrong)
+		ProbePodToPodConnectivity(k8s, model, testCase)
+	}
+
+	// at this point we know if we passed or failed, print final matrix and pass/fail the test.
+	if _, wrong, _, _ := testCase.Reachability.Summary(ignoreLoopback); wrong != 0 {
+		testCase.Reachability.PrintSummary(true, true, true)
+		framework.Failf("Had %d wrong results in reachability matrix", wrong)
+	}
+	if isVerbose {
+		testCase.Reachability.PrintSummary(true, true, true)
+	}
+	framework.Logf("VALIDATION SUCCESSFUL")
+}
+
+// UpdateNamespaceLabels sets the labels for a namespace
+func UpdateNamespaceLabels(k8s *Scenario, ns string, newNsLabel map[string]string) {
+	err := k8s.setNamespaceLabels(ns, newNsLabel)
+	framework.ExpectNoError(err, "Update namespace %s labels", ns)
+	err = wait.PollImmediate(waitInterval, waitTimeout, func() (done bool, err error) {
+		namespace, err := k8s.getNamespace(ns)
+		if err != nil {
+			return false, err
+		}
+		for key, expected := range newNsLabel {
+			if actual, ok := namespace.Labels[key]; !ok || (expected != actual) {
+				return false, nil
+			}
+		}
+		return true, nil
+	})
+	framework.ExpectNoError(err, "Unable to wait for ns %s to update labels", ns)
+}
+
+// AddPodLabels adds new labels to a deployment's template
+func AddPodLabels(k8s *Scenario, pod *Pod, newPodLabels map[string]string) {
+	kubePod, err := k8s.ClientSet.CoreV1().Pods(pod.Namespace).Get(context.TODO(), pod.Name, metav1.GetOptions{})
+	framework.ExpectNoError(err, "Unable to get pod %s/%s", pod.Namespace, pod.Name)
+	if kubePod.Labels == nil {
+		kubePod.Labels = map[string]string{}
+	}
+	for key, val := range newPodLabels {
+		kubePod.Labels[key] = val
+	}
+	_, err = k8s.ClientSet.CoreV1().Pods(pod.Namespace).Update(context.TODO(), kubePod, metav1.UpdateOptions{})
+	framework.ExpectNoError(err, "Unable to add pod %s/%s labels", pod.Namespace, pod.Name)
+
+	err = wait.PollImmediate(waitInterval, waitTimeout, func() (done bool, err error) {
+		waitForPod, err := k8s.GetPod(pod.Namespace, pod.Name)
+		if err != nil {
+			return false, err
+		}
+		for key, expected := range newPodLabels {
+			if actual, ok := waitForPod.Labels[key]; !ok || (expected != actual) {
+				return false, nil
+			}
+		}
+		return true, nil
+	})
+	framework.ExpectNoError(err, "Unable to wait for pod %s/%s to update labels", pod.Namespace, pod.Name)
+}
+
+// ResetNamespaceLabels resets the labels for a namespace
+func ResetNamespaceLabels(k8s *Scenario, ns string) {
+	UpdateNamespaceLabels(k8s, ns, (&Namespace{Name: ns}).LabelSelector())
+}
+
+// ResetPodLabels resets the labels for a deployment's template
+func ResetPodLabels(k8s *Scenario, pod *Pod) {
+	kubePod, err := k8s.ClientSet.CoreV1().Pods(pod.Namespace).Get(context.TODO(), pod.Name, metav1.GetOptions{})
+	framework.ExpectNoError(err, "Unable to get pod %s/%s", pod.Namespace, pod.Name)
+	kubePod.Labels = pod.LabelSelector()
+	_, err = k8s.ClientSet.CoreV1().Pods(pod.Namespace).Update(context.TODO(), kubePod, metav1.UpdateOptions{})
+	framework.ExpectNoError(err, "Unable to add pod %s/%s labels", pod.Namespace, pod.Name)
+
+	err = wait.PollImmediate(waitInterval, waitTimeout, func() (done bool, err error) {
+		waitForPod, err := k8s.GetPod(pod.Namespace, pod.Name)
+		if err != nil {
+			return false, nil
+		}
+		for key, expected := range pod.LabelSelector() {
+			if actual, ok := waitForPod.Labels[key]; !ok || (expected != actual) {
+				return false, nil
+			}
+		}
+		return true, nil
+	})
+	framework.ExpectNoError(err, "Unable to wait for pod %s/%s to update labels", pod.Namespace, pod.Name)
+}
diff --git a/test/e2e/network/netpol/truthtable.go b/test/e2e/network/netpol/truthtable.go
new file mode 100644
index 00000000000..b34b6a6bdb5
--- /dev/null
+++ b/test/e2e/network/netpol/truthtable.go
@@ -0,0 +1,171 @@
+/*
+Copyright 2020 The Kubernetes 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 netpol
+
+import (
+	"strings"
+
+	"k8s.io/kubernetes/test/e2e/framework"
+)
+
+// TruthTable takes in n items and maintains an n x n table of booleans for each ordered pair
+type TruthTable struct {
+	Froms  []string
+	Tos    []string
+	toSet  map[string]bool
+	Values map[string]map[string]bool
+}
+
+// NewTruthTableFromItems creates a new truth table with items
+func NewTruthTableFromItems(items []string, defaultValue *bool) *TruthTable {
+	return NewTruthTable(items, items, defaultValue)
+}
+
+// NewTruthTable creates a new truth table with froms and tos
+func NewTruthTable(froms []string, tos []string, defaultValue *bool) *TruthTable {
+	values := map[string]map[string]bool{}
+	for _, from := range froms {
+		values[from] = map[string]bool{}
+		for _, to := range tos {
+			if defaultValue != nil {
+				values[from][to] = *defaultValue
+			}
+		}
+	}
+	toSet := map[string]bool{}
+	for _, to := range tos {
+		toSet[to] = true
+	}
+	return &TruthTable{
+		Froms:  froms,
+		Tos:    tos,
+		toSet:  toSet,
+		Values: values,
+	}
+}
+
+// IsComplete returns true if there's a value set for every single pair of items, otherwise it returns false.
+func (tt *TruthTable) IsComplete() bool {
+	for _, from := range tt.Froms {
+		for _, to := range tt.Tos {
+			if _, ok := tt.Values[from][to]; !ok {
+				return false
+			}
+		}
+	}
+	return true
+}
+
+// Set sets the value for from->to
+func (tt *TruthTable) Set(from string, to string, value bool) {
+	dict, ok := tt.Values[from]
+	if !ok {
+		framework.Failf("from-key %s not found", from)
+	}
+	if _, ok := tt.toSet[to]; !ok {
+		framework.Failf("to-key %s not allowed", to)
+	}
+	dict[to] = value
+}
+
+// SetAllFrom sets all values where from = 'from'
+func (tt *TruthTable) SetAllFrom(from string, value bool) {
+	dict, ok := tt.Values[from]
+	if !ok {
+		framework.Failf("from-key %s not found", from)
+	}
+	for _, to := range tt.Tos {
+		dict[to] = value
+	}
+}
+
+// SetAllTo sets all values where to = 'to'
+func (tt *TruthTable) SetAllTo(to string, value bool) {
+	if _, ok := tt.toSet[to]; !ok {
+		framework.Failf("to-key %s not found", to)
+	}
+	for _, from := range tt.Froms {
+		tt.Values[from][to] = value
+	}
+}
+
+// Get gets the specified value
+func (tt *TruthTable) Get(from string, to string) bool {
+	dict, ok := tt.Values[from]
+	if !ok {
+		framework.Failf("from-key %s not found", from)
+	}
+	val, ok := dict[to]
+	if !ok {
+		framework.Failf("to-key %s not found in map (%+v)", to, dict)
+	}
+	return val
+}
+
+// Compare is used to check two truth tables for equality, returning its
+// result in the form of a third truth table.  Both tables are expected to
+// have identical items.
+func (tt *TruthTable) Compare(other *TruthTable) *TruthTable {
+	if len(tt.Froms) != len(other.Froms) || len(tt.Tos) != len(other.Tos) {
+		framework.Failf("cannot compare tables of different dimensions")
+	}
+	for i, fr := range tt.Froms {
+		if other.Froms[i] != fr {
+			framework.Failf("cannot compare: from keys at index %d do not match (%s vs %s)", i, other.Froms[i], fr)
+		}
+	}
+	for i, to := range tt.Tos {
+		if other.Tos[i] != to {
+			framework.Failf("cannot compare: to keys at index %d do not match (%s vs %s)", i, other.Tos[i], to)
+		}
+	}
+
+	values := map[string]map[string]bool{}
+	for from, dict := range tt.Values {
+		values[from] = map[string]bool{}
+		for to, val := range dict {
+			values[from][to] = val == other.Values[from][to]
+		}
+	}
+	return &TruthTable{
+		Froms:  tt.Froms,
+		Tos:    tt.Tos,
+		toSet:  tt.toSet,
+		Values: values,
+	}
+}
+
+// PrettyPrint produces a nice visual representation.
+func (tt *TruthTable) PrettyPrint(indent string) string {
+	header := indent + strings.Join(append([]string{"-\t"}, tt.Tos...), "\t")
+	lines := []string{header}
+	for _, from := range tt.Froms {
+		line := []string{from}
+		for _, to := range tt.Tos {
+			mark := "X"
+			val, ok := tt.Values[from][to]
+			if !ok {
+				mark = "?"
+			} else if val {
+				mark = "."
+			}
+			line = append(line, mark+"\t")
+		}
+		lines = append(lines, indent+strings.Join(line, "\t"))
+	}
+	return strings.Join(lines, "\n")
+}