{"schema_version":"1.0","service":"Publicasta","type":"article","id":1177,"slug":"aws_privatelink_tunnel_endpoints_cidr_access_governance","title":"AWS PrivateLink Tunnel Endpoints Turn CIDR Sharing into a Network Governance Decision","excerpt":"AWS has introduced tunnel endpoints that let one VPC privately reach an entire CIDR range in another. The operational question is no longer only how to connect services, but how to review access when new resources appear inside a shared network segment.","language":"en","default_language":"en","canonical_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=en","image":{"url":"https://publicasta.com/storage/projects/17/pages/1177/2026/11/e8f96757-e8e9-48f0-8737-2e5156f4283c.webp","alt":"Editorial visualization of two private cloud network segments connected through a secure tunnel and central governance gateway."},"publisher":{"id":17,"slug":"it_today_news","name":"IT Today","url":"https://publicasta.com/it_today_news"},"author":{"name":"Anton R"},"published_at":"2026-11-09T13:51:25+00:00","updated_at":"2026-11-09T13:51:25+00:00","content_markdown":"## What AWS changed with tunnel endpoints\n\n ![Editorial visualization of two private cloud network segments connected through a secure tunnel and central governance gateway.](https://publicasta.com/storage/projects/17/pages/1177/2026/11/e8f96757-e8e9-48f0-8737-2e5156f4283c.webp)\n\n AWS’s November 5 announcement introduces tunnel endpoints, a new PrivateLink-backed VPC endpoint type for private access to network segments in another VPC. The change extends the resource-sharing model beyond a single named destination. A provider can define a network segment as a list of CIDR ranges, and a consumer can use a tunnel endpoint to reach resources located within those ranges. The central object being shared is therefore not necessarily a particular database, instance, or application address; it can be a logical slice of the provider’s network.\n\n That distinction matters because a CIDR range describes an address space rather than a fixed inventory of resources. A consumer application can reach resources inside the shared segment without the provider having to enumerate each one individually. AWS positions this as useful when resources are ephemeral or not known in advance. The provider can expose a segment that already contains several destinations, while the consumer receives a path to the segment through its own VPC. The convenience is architectural, but it also changes what reviewers must understand about the permission: access follows the range and its allowed ports, not only the resources that happen to occupy it when the configuration is created.\n\n The provider side starts with a CIDR resource configuration. A resource configuration can represent an individual resource, a group, or a list of CIDR ranges; for this feature, the relevant type is the CIDR resource configuration. The provider associates that configuration with a resource gateway. The gateway is the point of ingress into the provider VPC and is represented by a set of elastic network interfaces. When a consumer in another VPC or account accesses the shared configuration, traffic enters through that gateway before reaching a destination in the declared segment. The resource gateway can be associated with multiple resource configurations, so its placement and surrounding controls are part of the provider’s network design rather than an incidental implementation detail.\n\n Cross-account access uses AWS Resource Access Manager. The provider shares the resource configuration through an AWS RAM resource share, allowing specific principals in another account to connect to it. The consumer then creates a tunnel endpoint in its own VPC. This preserves the familiar PrivateLink direction of the relationship: the consumer creates the endpoint and initiates the connection, while the provider defines what is exposed and which principals may use the share. The result is private connectivity between VPCs over the AWS network, but the boundary is now expressed as a provider-approved set of network ranges rather than as one named resource endpoint.\n\n The new tunnel endpoint should be distinguished from the existing resource endpoint model. A resource endpoint is used for resource configurations representing types such as an ARN, a single resource, or a group. A tunnel endpoint is used for a CIDR resource configuration and provides a tunnel into the listed network segments. That difference determines the review question. With a named resource, the configuration identifies the destination directly. With a tunnel endpoint, the configuration identifies the address space in which destinations may be reached. The provider is intentionally choosing a broader abstraction, even when the initial range appears small.\n\n The feature also covers both IPv4 and IPv6 network segments. A shared network segment can contain IPv4 ranges, IPv6 ranges, or both, and tunnel endpoints support IPv4, IPv6, or dual-stack addresses subject to the endpoint subnets’ address types. The outer packet used to reach the endpoint and the inner packet used to reach the remote destination may use different IP types. This makes the capability applicable to mixed address environments, but it does not remove the need to define the intended ranges precisely. The IP family and range are still part of the exposure decision.\n\n Port controls remain a provider-side constraint. A resource configuration can define the ports on which it will accept requests, and client access on other ports is not allowed. For a CIDR resource configuration, AWS documents the protocol setting as `TCP_UDP`: TCP is used for application traffic, while UDP is supported only for DNS queries. The tunnel itself uses GENEVE encapsulation, which is covered in the networking mechanics section; at this stage, the important point is that the consumer is not receiving an unrestricted, ordinary route to the provider VPC. It is receiving a PrivateLink tunnel governed by the shared CIDRs and configured ports.\n\n The practical shift is therefore simple to state but significant to operate: PrivateLink can now represent access to a network segment, not just access to an individually identified resource. Providers gain a way to publish reachable ranges for workloads that change over time, and consumers gain a private path to those ranges through their own VPCs. Every such path should be treated as a deliberate network-sharing decision, because the object approved through AWS RAM defines the address space that the consumer is allowed to approach.\n\n ## Why CIDR-level sharing changes the access review\n\n The important change is not simply that a consumer VPC can reach a provider VPC through AWS PrivateLink. It is that the permission can now be expressed as access to a CIDR range rather than to one specifically named resource. That difference changes the object being reviewed. A resource-specific approval describes a known destination; a CIDR-level approval describes a portion of the provider’s network and can therefore cover resources that are added after the original connection was approved.\n\n This makes the provider’s address plan part of the access-control decision. A range may contain several resources today, but its security meaning is also determined by what the provider may deploy there later. If a new service, appliance, or other network resource receives an address inside the shared segment, it can fall within the same approved boundary. The review therefore has to ask not only which destinations are reachable now, but which destinations could become reachable without another consumer-side endpoint decision.\n\n The same question applies beyond resources deployed directly in the immediate segment. The shared range can include reachable connected networks, so the effective scope may be larger than a simple inventory of current addresses suggests. A team reviewing the connection should map the provider’s intended use of the range, its connected paths, and the way future deployments are expected to use that space. Without that context, a rule that appears constrained by a relatively small network block may still grant access to a changing collection of destinations.\n\n That is why a narrow-looking permission can expand over time. The expansion does not require the consumer to change its endpoint configuration. It can result from the provider adding resources, changing how the segment is used, or making additional networks reachable through it. In governance terms, the original approval is not a permanent statement about a fixed set of systems. It is an approval attached to an address scope whose contents may change.\n\n The practical consequence is that providers should approve the narrowest CIDR ranges that support the intended connection. A broad range may be convenient during initial setup, but convenience makes later review harder because the approval covers more possible destinations. The provider should be able to explain why each included portion of the range is needed and which future-resource exposure is acceptable. If the use case concerns one segment or a small group of destinations, the approved range should reflect that boundary instead of inheriting a larger aggregate block.\n\n Ports deserve the same discipline. CIDR scope answers where traffic may go; port scope helps define what kind of connection is permitted. Approving only the ports required by the service reduces the consequences if another resource later appears inside the same address range. A review that considers the range but treats the port set as an afterthought misses part of the exposure. Both dimensions should be recorded together: the provider-side addresses covered by the permission and the ports that consumers may use there.\n\n The broadest case is `0.0.0.0/0`. Using that range effectively removes the IPv4 address boundary from the approval, so it should be treated as an exceptional design choice rather than a default shortcut. It can encompass resources across the provider’s IPv4 space and makes future changes especially difficult to evaluate from the original approval alone. If the required destination is known, a more specific range provides a clearer governance boundary. If a broad range is unavoidable, the provider should document the rationale, the permitted ports, the expected reachable paths, and the conditions that would trigger a new review.\n\n This model also changes ownership. The provider controls the meaning and future contents of the shared ranges, while the consumer controls whether and how it initiates connections through its endpoint. That division means a consumer-side security review cannot substitute for provider-side network governance. Consumers need to understand the approved scope before using it, but providers must maintain an inventory of what the scope contains and reassess it when the network evolves.\n\n A useful approval record should therefore describe more than an endpoint identifier. It should state the exact CIDR range, the allowed ports, the intended resource or service purpose, the connected networks that may be reachable, and the owner responsible for reviewing later changes. When a provider deploys a new resource or modifies the connectivity of the shared segment, the question is whether that change remains inside the original security intent. If it does not, the PrivateLink relationship should return to review even if no endpoint setting changes.\n\n CIDR-level PrivateLink access can simplify connectivity, but it also turns address management into a continuing governance concern. The safest interpretation is to treat every approved range as a living boundary: precise enough to limit exposure today, documented enough to evaluate future deployments, and narrow enough that a later network change does not silently broaden access beyond what the original review intended.\n\n ## The networking mechanics teams must understand\n\n A tunnel endpoint is not a conventional interface endpoint with a different label. The consumer application must create an inner Layer 3 IP packet addressed to a destination in the provider’s shared CIDR range, then encapsulate that packet in GENEVE before sending it to the endpoint. The outer traffic uses UDP port 6081, and the GENEVE VNI must be set to `0`. For application traffic, the inner protocol must be TCP; UDP inside the tunnel is supported only for DNS queries. This distinction matters during implementation and testing: a security group or network policy that permits the application’s TCP port but blocks UDP 6081 between the compute environment and the endpoint will prevent the tunnel from working at all.\n\n The endpoint removes the GENEVE header and checks the inner destination against the CIDR ranges shared through the provider’s resource configuration. If the destination is outside those ranges, the endpoint does not forward it. The provider-side resource gateway then receives the traffic and performs source translation. The destination resource therefore sees the connection as coming from an address associated with the resource gateway, rather than from the consumer workload’s original inner address. That makes the connection feasible across the PrivateLink boundary and allows overlapping consumer and provider VPC CIDRs, but it changes what the provider can infer from ordinary application logs: the target does not see the original consumer workload IP. Teams that need attribution must correlate endpoint, gateway, and application telemetry rather than assuming the resource’s source address identifies the caller.\n\n The direction of connection initiation is another architectural constraint. The consumer VPC can open a TCP connection into the shared network segment, and the provider network can return traffic through the tunnel, but the provider cannot initiate a new connection toward the consumer VPC. This is a one-way access model at the session-initiation level, not a general bidirectional network link. Designs that depend on callbacks, unsolicited provider connections, or provider-side probing need a separate pattern or an application flow that keeps the consumer as the initiator.\n\n The tunnel also has its own routing context. The provider’s addresses exist inside the encapsulated packet, while the outer packet is routed to the tunnel endpoint’s address in the consumer VPC. In practice, AWS recommends isolating the GENEVE device and its routes so that traffic meant for the shared ranges enters the tunnel without accidentally capturing unrelated workload traffic. A dedicated routing context, such as a network namespace on Linux, can help separate the tunnel’s default route from the instance’s normal traffic. The important troubleshooting question is therefore twofold: does the inner destination use the intended tunnel route, and can the outer UDP packet reach the correct endpoint address?\n\n Overlapping CIDRs are supported because the consumer does not route the provider’s inner destination through ordinary VPC routing. The consumer routes the outer encapsulated packet locally to the tunnel endpoint, which decapsulates it before delivery through the resource gateway. This removes the usual requirement for unique address space between the two VPCs, but it does not remove the need to identify the intended destination precisely. If identical addresses exist in the consumer and provider environments, the application’s route and tunnel configuration must make clear which traffic is being encapsulated; otherwise, local routing can win before the packet ever reaches the endpoint.\n\n DNS requires similar care. Tunnel endpoints have a regional DNS name that resolves to the endpoint IP addresses in the configured subnets, but private DNS names are not supported for the tunnel endpoint itself. For name resolution through the remote VPC, the consumer sends the DNS query inside GENEVE to `169.254.168.253`. The query is resolved in the context of the provider VPC, using that VPC’s DHCP option set and respecting its Route 53 private hosted zones or custom resolvers. A CIDR resource configuration has no domain name of its own, so consumers resolve the names already published by the provider. A failed application lookup may therefore indicate an incorrectly configured tunnel nameserver or remote DNS policy, not a missing application route.\n\n Availability Zone placement is part of the data path, not merely a resilience preference. The tunnel endpoint’s subnets must be in Availability Zones that overlap with the Availability Zones used by the provider’s resource gateway. An endpoint can appear provisioned while still being unusable for an association if there is no overlapping zone. For production deployments, AWS recommends configuring the endpoint in at least two Availability Zones. Workloads should normally target the endpoint address in their own Availability Zone, while the failure design must account for what happens if that zone is impaired. Validate the actual zone alignment across accounts and check endpoint association status before diagnosing higher-layer connectivity problems.\n\n ## A practical rollout and governance checklist\n\n Adopting tunnel endpoints should begin with an inventory, not an endpoint creation. The provider team should list every CIDR that may be published through a resource configuration, the resources currently inside each range, and the networks that can become reachable through connected routing. That last part matters because a range is an address boundary, not a permanent catalog of today’s services. A later deployment can inherit access as soon as its address falls inside an approved range, while a connected network may extend the practical reach beyond what a reviewer sees in the initial diagram. Record the intended consumers, protocols, ports, and business purpose for each range, then make the provider and consumer owners explicit. The provider owns the exposure decision; the consumer owns its use of the endpoint and the traffic it sends through it.\n\n Start the first implementation with the narrowest useful range and the smallest set of ports. A trial should demonstrate a concrete application path rather than prove that a broad network can be reached. Avoid treating `0.0.0.0/0` as a convenient default: it removes the address-based boundary that makes a CIDR-level policy reviewable. If IPv6 is in scope, apply the same discipline to the approved IPv6 range. Keep the resource configuration, the endpoint association, and the approval record tied together so that an operator can answer three questions later: which consumer was approved, which addresses were included, and which traffic was expected.\n\n Before production approval, review the complete path on both sides. Confirm that the consumer has the required routes to the tunnel endpoint and that the provider has the routing needed to reach the target segment. Check security groups, network ACLs, host firewalls, and service-level authorization independently; PrivateLink connectivity does not replace those controls. The connection is consumer initiated, so the test plan should begin from the consumer VPC and verify both successful and denied flows. Test name resolution from the consumer environment as well. Remote-VPC DNS resolution can affect whether an application reaches the intended address, and a path that works by IP alone may not represent the application’s real operating mode.\n\n Availability deserves its own gate. Place the endpoint and the provider-side resource gateway so that the design has overlapping Availability Zones, and validate the placement in every participating VPC. AWS recommends deploying across at least two Availability Zones for resilience. A single-zone test can therefore conceal a production dependency that appears only when an application, route, or endpoint is placed elsewhere. Include zone loss and endpoint replacement in the runbook, and record which components must be recreated or re-associated if the topology changes.\n\n Cost and observability should be part of the approval rather than an afterthought. Establish a baseline for endpoint operation and data processing, then identify which team will watch those measures after launch. Monitor accepted and rejected connection attempts, unusual destination addresses, traffic volume, and changes to the provider’s resource configurations. The operational signal to prioritize is not merely “the endpoint is healthy,” but “the endpoint is carrying the destinations and volume that were approved.” Keep the first rollout small enough that an unexpected flow can be traced to a consumer, route, port, and resource configuration without guesswork.\n\n Finally, treat every provider-network change as an access review event. Adding a resource, enlarging a CIDR, attaching a connected network, changing routes, or changing the ports exposed through a resource configuration can alter the meaning of an existing approval. The provider should reassess the reachable set and notify consumers when that set changes; consumers should reassess whether their workloads still need the endpoint and whether their own controls remain least-privilege. This cadence turns tunnel endpoints from a one-time connectivity decision into controlled network governance: approve a defined segment, verify the path and safeguards, watch what is actually reachable, and reapprove when the provider network evolves.","available_translations":[{"language":"ar","title":"نقاط نهاية الأنفاق في AWS PrivateLink تحوّل مشاركة CIDR إلى قرار لحوكمة الشبكة","html_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=ar","markdown_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=ar","json_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=ar","api_url":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=ar"},{"language":"de","title":"AWS-PrivateLink-Tunnel-Endpunkte machen die gemeinsame CIDR-Nutzung zu einer Frage der Netzwerk-Governance","html_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=de","markdown_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=de","json_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=de","api_url":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=de"},{"language":"en","title":"AWS PrivateLink Tunnel Endpoints Turn CIDR Sharing into a Network Governance Decision","html_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=en","markdown_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=en","json_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=en","api_url":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=en"},{"language":"es","title":"Los endpoints de túnel de AWS PrivateLink convierten el uso compartido de CIDR en una decisión de gobernanza de red","html_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=es","markdown_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=es","json_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=es","api_url":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=es"},{"language":"fr","title":"Les points de terminaison de tunnel AWS PrivateLink transforment le partage de CIDR en décision de gouvernance réseau","html_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=fr","markdown_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=fr","json_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=fr","api_url":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=fr"},{"language":"pl","title":"Punkty końcowe tuneli AWS PrivateLink zmieniają współdzielenie CIDR w decyzję z zakresu zarządzania siecią","html_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=pl","markdown_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=pl","json_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=pl","api_url":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=pl"},{"language":"ru","title":"Туннельные конечные точки AWS PrivateLink превращают совместное использование CIDR в задачу сетевого управления","html_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=ru","markdown_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=ru","json_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=ru","api_url":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=ru"},{"language":"zh","title":"AWS PrivateLink 隧道端点：CIDR 共享已成为网络治理决策","html_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=zh","markdown_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=zh","json_url":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=zh","api_url":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=zh"}],"_links":{"self":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=en","api":"https://publicasta.com/api/public/v1/channels/it_today_news/articles/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=en","html":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=en","canonical":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance?lang=en","markdown":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.md?lang=en","json":"https://publicasta.com/it_today_news/aws_privatelink_tunnel_endpoints_cidr_access_governance.json?lang=en","channel":"https://publicasta.com/api/public/v1/channels/it_today_news","channel_articles":"https://publicasta.com/api/public/v1/channels/it_today_news/articles","search":"https://publicasta.com/api/public/v1/search","documentation":"https://publicasta.com/api-docs#reading-publicasta","openapi":"https://publicasta.com/api-docs/openapi.json","llms":"https://publicasta.com/llms.txt"}}