Join our Newsletter — 33% off our NHI Course
Home› Guides› Site-to-Site VPN Security Guide: Branch, Partner and Cloud…
Guide Network Security

Site-to-Site VPN Security Guide: Branch, Partner and Cloud Tunnels Done Safely

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 14 November 2026 14 min read
On this page

Site-to-site VPN tunnels join your network to branches, partners and cloud providers, and most of them outlive the projects, contracts and people that created them. The protocol settings are usually reviewed once; the harder questions come later: who owns this tunnel, what may cross it, which agreement covers it, where the pre-shared key lives and when it should be switched off. This guide covers the governance side: ownership and inventory, partner interconnection agreements, cloud VPN gateway configuration, monitoring and retirement. It is written for network security engineers, cloud platform teams and third-party risk owners. IPsec algorithm choices are covered in the IPsec Security Guide and are only referenced here.

Key takeaways

  • Every tunnel needs a named owner and an end date. A tunnel with no business owner, no review date and no agreement is a standing trust relationship nobody is watching.
  • Partner tunnels are third-party access. Cover them with an interconnection security agreement or equivalent contract terms, as NIST SP 800-47 Rev. 1 describes, and review them with the supplier.
  • Treat the tunnel as untrusted on arrival. Narrow traffic selectors and firewall policy on the tunnel interface limit what a compromised partner or branch can reach.
  • Cloud gateways are configured by API and need guardrails. Restrict who can create VPN connections, store keys in a secrets manager and set explicit crypto policies rather than relying on interoperability defaults.
  • Retirement is a control, not housekeeping. Remove the tunnel, the key or certificate, the routes and the firewall rules together when the relationship ends.

Types of site-to-site tunnel and their risks

Site-to-site VPNs differ less in technology than in who sits at the far end. The governance burden follows the trust relationship.

Tunnel typeFar end controlled byMain risksPrimary control
Branch or siteYouFlat routing from small sites with weaker physical security; unmanaged edge devicesSegmentation of branch traffic and managed, patched gateways
Partner or supplierAnother organisationCompromise of the partner becomes your compromise (ATT&CK T1199 Trusted Relationship); forgotten tunnels after contract endInterconnection agreement, narrow selectors, inner firewall policy, review cycle
Cloud provider gatewayYou, through the provider's control planeAnyone with cloud permissions can create a tunnel to an arbitrary peer; keys visible in configuration downloadsIAM restrictions, organisation policy, secrets manager for keys
Cloud to cloudYou, across two providersInconsistent crypto and logging between platforms; routes that bridge environmentsOne baseline applied to both ends; see the Multicloud Networking Security guide
Managed service providerA provider operating your devicesProvider staff and tooling gain a persistent path inContract terms, logging you can read, access reviews

A tunnel that joins two networks without filtering is the condition ATT&CK describes as T1599 Network Boundary Bridging. Treat the decrypted side as a separate zone. See the Network Segmentation Guide.

Ownership and tunnel inventory

You cannot govern tunnels you have not listed. Build an inventory from three sources and reconcile them: VPN gateway configurations (on-premises devices and cloud APIs), firewall rules referencing tunnel interfaces, and contracts or tickets that requested connectivity. NIST SP 800-53 CM-8 System Component Inventory and CSF 2.0 ID.AM-03, which asks for maintained representations of internal and external network data flows, both apply.

Inventory fieldWhy it matters
Tunnel ID, gateway and peer public IPTies log entries and alerts to a record
Business owner and technical ownerSomeone accountable for reviews and incidents
Counterparty and contract or agreement referenceLinks the tunnel to the legal basis for it
Permitted subnets, ports and directionThe approved data flow to test against
Authentication method and credential locationPre-shared key in a vault, or certificate issuer and expiry
Crypto profile and IKE versionShows which tunnels sit on a legacy profile
Review date and end dateTriggers recertification and retirement
Data classification crossing the tunnelSets monitoring and agreement depth

Keep this in the asset inventory or CMDB rather than a spreadsheet owned by one engineer. See the Asset Inventory and CMDB Security Guide.

Partner tunnels and interconnection agreements

NIST SP 800-47 Rev. 1, Managing the Security of Information Exchanges (July 2021), focuses on protecting information before, during and after an exchange rather than on any connection technology. It names the agreement types organisations use: interconnection security agreements, memoranda of understanding or agreement, information exchange agreements, service level agreements, nondisclosure agreements and user agreements. NIST SP 800-53 CA-3 Information Exchange requires those agreements to be approved and managed. For a partner VPN, agree in writing:

  • Purpose and data: the business reason, the systems and subnets on each side, and the data classification crossing the tunnel.
  • Technical baseline: IKE version, crypto profile, authentication method (certificate or a unique random pre-shared key per peer), rekey and lifetime settings, and who changes them.
  • Key handling: how keys or certificates are exchanged (never in plain email), rotated, and revoked when staff with access leave either organisation.
  • Security obligations: patching of the partner's gateway, notification of incidents affecting the tunnel within an agreed time, and the right to suspend the tunnel during an incident.
  • Logging and contacts: what each side logs, retention, and named 24x7 contacts for outages and incidents.
  • Review and termination: an annual review date, and the steps and timeline for teardown at contract end.

The contract clauses themselves are in the Third-Party Security Contract Clauses Template. CSF 2.0 GV.SC-07 and GV.SC-10 place supplier risk and post-relationship activities in the supply chain programme, so register partner tunnels with your third-party risk team, not just the network team.

Cloud VPN gateway configuration

Each major cloud provider offers a managed site-to-site VPN gateway configured through its control plane. Platforms are listed alphabetically; the order is not a ranking. Each section covers the customer-side settings that matter for governance, from the provider's documentation as of November 2026.

Amazon Web Services: AWS Site-to-Site VPN

  • Resilience: each connection has two tunnels with unique public IP addresses in different Availability Zones; configure both on the customer gateway so traffic fails over.
  • Authentication: a pre-shared key is the default, which you can supply or let AWS generate. AWS documents storing the key in AWS Secrets Manager rather than in the VPN service, or using a private certificate from a subordinate CA in AWS Private CA instead of a key.
  • Crypto policy: AWS states that tunnel endpoints evaluate customer gateway proposals starting with the lowest configured value, regardless of proposal order, so restrict accepted options per tunnel with modify-vpn-connection-options. AWS also documents generating recommended configurations that exclude legacy options such as IKEv1, and a GetActiveVpnTunnelStatus API that reports the algorithms and Diffie-Hellman groups in use on active tunnels.
  • Logging: tunnel activity logs and BGP logs can be published to Amazon CloudWatch Logs, including IKE phase 1 and phase 2 state, DPD status and messages such as pre-shared key mismatches and unacceptable traffic selectors.
  • Availability: Large Bandwidth Tunnels (up to 5 Gbps) apply to connections attached to a Transit Gateway or Cloud WAN and are not offered in every Region.

See the AWS Security Guide.

Google Cloud: Cloud VPN

  • Resilience: Google describes HA VPN as the recommended method: two interfaces, each with its own external IP address, dynamic BGP routing through Cloud Router, and a 99.99% SLA for most topologies when both interfaces are used.
  • Authentication: tunnels authenticate with a pre-shared key that must match on both ends; Google recommends generating a strong key.
  • Crypto policy: Cloud VPN supports IKEv1 and IKEv2, and cipher selection is available with IKEv2. Google advises choosing the strongest cipher both peers support, since its proposal order is not sorted by strength.
  • Guardrails: Google recommends restricting which IP addresses can be set as peer VPN gateways to block unauthorised tunnels, and keeping Cloud VPN and Cloud Router resources in a separate project to simplify IAM.
  • Logging: predefined Cloud VPN dashboards show tunnel health across a project, with tunnel and gateway logs for diagnosis. External gateway addresses accept only ESP, UDP 500 and UDP 4500.

See the Google Cloud Security Guide.

Microsoft Azure: Azure VPN Gateway

  • Resilience and type: RouteBased gateways support IKEv1 and IKEv2 and dead peer detection; Microsoft recommends IKEv2 for in-place rekeys, and the protocol type cannot be changed after a connection is created, so choose it at build time.
  • Authentication: site-to-site connections authenticate with a pre-shared key.
  • Crypto policy: apply a custom IPsec/IKE policy per connection (all SKUs except Basic). Once set, the gateway uses only that policy as initiator and responder. Microsoft's default proposals are chosen for interoperability, and it advises stronger groups such as DH Group 14, Group 24 or ECP256/384 where compliance requires them. Microsoft also notes ESP NULL encryption is possible on some gateways and should be used only where encryption is applied elsewhere.
  • Traffic selectors: with UsePolicyBasedTrafficSelectors, Microsoft states the gateway accepts whatever selector the remote gateway proposes, so define every on-premises to virtual network prefix pair on the peer device instead of any-to-any.
  • Logging: resource logs TunnelDiagnosticLog, IKEDiagnosticLog, RouteDiagnosticLog and GatewayDiagnosticLog go to the AzureDiagnostics table, with metrics such as TunnelIngressBytes and TunnelIngressPacketDropCount.

See the Microsoft Azure Security Guide.

Guardrails common to every platform

  • Restrict creation rights: only the network platform team should hold permission to create VPN gateways, customer or local gateways and connections; enforce it with IAM and organisation-level policy, and alert on any creation event.
  • Keep keys out of configuration files: downloaded device configurations can contain the pre-shared key. Store keys in a secrets manager, delete local copies and rotate after any engineer with access leaves.
  • Codify tunnels: manage gateways and connections through infrastructure as code with peer review, so the repository is a second inventory and drift is detectable.

Where identity fits

A site-to-site tunnel authenticates two gateways, not people. The pre-shared key or gateway certificate is a non-human credential that grants network-level access to everything the selectors allow. Give each peer pair a unique credential, record its owner and rotation date in the inventory, and prefer certificates from a managed CA with defined validity and revocation where both peers support them. The same applies to the cloud identities that can change tunnels: a compromised cloud administrator role can add a tunnel to an attacker-controlled peer without touching your on-premises firewall (ATT&CK T1078.004 Cloud Accounts). See the Machine Identity, PKI and Certificate Lifecycle Guide.

Implementation: bringing existing tunnels under control

  1. Discover: export every tunnel from on-premises gateways and cloud APIs, and match each to a firewall rule set, routing entry and requester. Unmatched tunnels go on a review list, not straight to deletion.
  2. Assign owners: give each tunnel a business and technical owner within a fixed window; tunnels still unowned after it move to a scheduled shutdown with notice to the counterparty.
  3. Paper the partners: put an interconnection agreement or contract schedule in place for every partner tunnel, prioritising those carrying sensitive data.
  4. Baseline: apply the crypto profile from the IPsec guide, move keys to a vault, narrow selectors and add inner firewall policy. Change one tunnel at a time in an agreed window with the peer.
  5. Test: confirm only the approved flows pass, using a test from each side; record results against the inventory entry. See the Network Segmentation Testing Guide.
  6. Roll back safely: keep the previous configuration for each tunnel and restore it on that tunnel only if the peer cannot establish, never by loosening a global policy.
  7. Operate: recertify each tunnel at its review date, and run retirement as below when the end date or contract end arrives.

Tunnel retirement

Tunnels are created by projects and rarely removed by them. Tie retirement to contract end, site closure and cloud account decommissioning, and complete every step:

  • Confirm with the owner and counterparty that no traffic still depends on the tunnel; check the last weeks of byte counts.
  • Disable first, delete later: shut the tunnel and wait an agreed period before deletion, so a forgotten dependency surfaces without data loss.
  • Remove everything with it: the gateway connection, the customer or peer gateway object, routes and BGP sessions, firewall rules and NAT entries, the secret in the vault and any certificate (revoke it).
  • Close the paperwork: mark the agreement terminated and the inventory entry retired, with the date and approver.

CSF 2.0 GV.SC-10 calls for supply chain plans to cover activities after a partnership ends; tunnel teardown is one of them.

Decision guidance

  • Site-to-site VPN or private connectivity: dedicated circuits to a cloud provider suit high, steady throughput; VPN suits branches, partners and backup paths. Many organisations run VPN as encrypted backup to a private circuit.
  • Tunnel or application access for partners: if a partner needs a handful of applications rather than network reach, application-level access through ZTNA or an API gateway avoids a network bridge. See the ZTNA Guide.
  • Pre-shared key or certificate: certificates scale better and revoke cleanly; a unique random key per peer, held in a vault, is acceptable for a small number of peers where both sides lack PKI.
  • Small versus large organisations: a small organisation with a few tunnels can keep the inventory in the CMDB and review annually; large estates need automated discovery from cloud APIs, infrastructure as code and quarterly recertification.

Detection and measurement

Feed gateway logs to the SIEM and alert on new tunnels, peer IP changes, repeated negotiation failures and traffic to subnets outside the approved selectors. Detection content for VPN and firewall logs is in the VPN and Firewall Log Detection Guide. Track these metrics:

  • Inventory coverage: percentage of discovered tunnels with an owner, agreement and end date.
  • Overdue recertifications: tunnels past their review date.
  • Legacy profile count: tunnels on IKEv1 or a time-limited legacy crypto profile.
  • Keys outside a vault: pre-shared keys not held in a secrets manager.
  • Idle tunnels: tunnels with no traffic for 90 days, as retirement candidates.
  • Time to retire: days from contract end to tunnel deletion.

Framework mapping

Control areaNIST SP 800-53 Rev. 5NIST CSF 2.0ATT&CK
Partner agreements and information exchangeCA-3, SA-9GV.SC-07T1199
Tunnel inventory and data flowsCM-8ID.AM-03None
Boundary and inner filteringSC-7PR.IR-01T1599, M1030
Transmission protection and keysSC-8, SC-12PR.AA-01None
Cloud control-plane accessAC-2PR.AA-05T1078.004
Monitoring and reviewAU-6DE.CM-03T1133
Teardown at relationship endPS-7, SA-9GV.SC-10None

In ISO/IEC 27001:2022 Annex A the topic sits under 5.20 Addressing information security within supplier agreements, 8.20 Networks security and 8.21 Security of network services.

Common failure patterns

  • The orphaned partner tunnel: the contract ended years ago, the tunnel still negotiates every eight hours and nobody knows who the peer contact is.
  • Any-to-any selectors with no inner filtering: the partner's whole network can reach yours, so their compromise becomes yours.
  • One key for many peers: the same pre-shared key reused across branches or partners, exchanged by email and never rotated.
  • Shadow cloud tunnels: a project team with broad cloud permissions creates a VPN to a vendor's network outside change control.
  • Defaults left in place: interoperability-first proposal lists accepted because nobody set an explicit policy on the cloud side.
  • Partial teardown: the tunnel is deleted but routes, firewall rules and the stored key remain, ready to be reused.

Practitioner checklist

  • Inventory every site-to-site tunnel with owner, counterparty, permitted flows, credential location, review and end dates.
  • Put an interconnection agreement or contract schedule in place for each partner tunnel.
  • Apply narrow traffic selectors and firewall policy on every tunnel interface.
  • Store pre-shared keys in a secrets manager, unique per peer, or use gateway certificates from a managed CA.
  • Set an explicit crypto policy on each cloud gateway connection and restrict accepted proposals.
  • Limit who can create cloud VPN resources and alert on creation and peer changes.
  • Send tunnel, IKE and routing logs to the SIEM and track the six metrics above.
  • Recertify tunnels on schedule and run full teardown at contract or site end.

Standards and references

Related NHI Mgmt Group resources: IPsec Security Guide · VPN Security Guide · VPN Protocols Compared · VPN and Firewall Log Detection Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 14 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.