Join our Newsletter — 33% off our NHI Course
Home› Guides› SASE vs SD-WAN: When Transport Alone Stops Being…
Guide Network Security

SASE vs SD-WAN: When Transport Alone Stops Being Enough

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 14 November 2026 17 min read
On this page

SD-WAN and SASE are often sold as stages of the same journey, but they answer different questions: SD-WAN decides how traffic gets between sites, while SASE decides whether a given user, device or workload should be allowed to reach a given resource at all. Treating SD-WAN as a security product leaves branches with encrypted tunnels and very little inspection, and treating SASE as a WAN replacement can leave branch routing, resilience and local devices unmanaged. This guide compares the two as a decision: what each one protects against, how each fails, where branch security gaps appear, how identity-aware access changes the picture, and when SD-WAN alone is still the right answer. It is for network and security architects choosing between staying on SD-WAN, adding a separate security service edge, or moving to a single-vendor SASE platform. Component detail lives in the SASE and SSE Guide and the SD-WAN Security Guide; this page is about choosing between them.

Key takeaways

  • SD-WAN is transport with policy, not a security stack. It classifies application flows and picks a path, and it encrypts tunnels between your own edges, but it does not by itself inspect web traffic, govern SaaS use or authenticate users to applications.
  • SASE adds the security stack and moves it to the identity. Policy follows the user and device to a cloud point of presence rather than following the circuit to a branch firewall.
  • The gap usually opens at local internet breakout. The moment a branch sends traffic straight to the internet, the inspection that used to happen in the data centre has to happen somewhere else, or it stops happening.
  • Both concentrate control. SD-WAN concentrates it in orchestrators and controllers you operate; SASE concentrates it in a provider's cloud and your admin tenant. Either way, management plane identity is the highest-value target.
  • SD-WAN alone can still be right. Small site counts, traffic that mostly stays on private paths, and a strong existing inspection point can justify deferring SASE, provided the gaps are written down and owned.

The short answer

Choose on the problem, not the label. If your main problem is WAN cost, link resilience and application performance between known sites, SD-WAN solves it and SASE is optional. If your main problem is that users, SaaS and internet traffic no longer pass through a place you can inspect, you need security service edge functions (secure web gateway, cloud access security broker, zero trust network access, firewall as a service), and SD-WAN will not provide them on its own. If you have both problems, the choice is between a single-vendor SASE platform, where one provider delivers the SD-WAN edge and the security cloud under one policy model, and a dual-vendor design, where your existing SD-WAN steers traffic into a separate security service edge.

NIST SP 800-215, the Guide to a Secure Enterprise Network Landscape (final, November 2022), frames this shift directly: it reviews the limits of traditional network access approaches and describes the evolving WAN infrastructure as a source of comprehensive security services, naming SASE and ZTNA as examples. The Mplify Alliance (formerly MEF) standardised both sides separately: MEF 70.x defines SD-WAN service attributes, and MEF 117 (October 2022, since superseded by MEF 117.1) defines a SASE service framework whose attributes cover security functions, policies and connectivity services for secure access by users, devices and applications.

Transport versus security stack

The clearest way to separate them is by layer of responsibility. SD-WAN owns the path; SASE owns the decision about the session.

QuestionSD-WANSASE
Primary jobSteer application flows across multiple underlays (broadband, LTE or 5G, MPLS) with path selection and failoverApply security policy to users, devices and sites wherever they connect from
Unit of policyApplication flow, site and tunnelUser and group identity, device posture, application and data
Where enforcement runsOn branch edges and hubs you operateMostly in provider points of presence, with an optional SD-WAN edge or client agent
EncryptionTunnels between your own edges; underlay treated as untrustedTunnels or agent connections from sites and users to the provider cloud, then onward sessions to applications
Web and SaaS inspectionNot inherent; some edges add local firewall or URL filteringCore function: secure web gateway, CASB, DLP, TLS inspection
Private application accessNetwork-level reachability between sitesPer-application brokered access (ZTNA) without broad network reachability
Remote usersOut of scope unless a separate VPN is addedIn scope through an agent or browser-based access
Standards referenceMEF 70.1 and 70.2; MEF 88 application flow securityMEF 117 and 117.1

The Mplify SD-WAN standards describe the edge as classifying packets into application flows, applying policy and choosing the tunnel each flow uses; MEF 70.1 (November 2021) added infrastructure for application-based security, and MEF 88 covers application flow security for SD-WAN services, with MEF 131 defining certification tests against it. That is real security value: segmentation by application, encryption across untrusted links and a consistent policy at every edge. It is still network security for known sites, not access control for identities.

Threat model for each option

SD-WAN on its own

What it defends well: eavesdropping and tampering on the underlay (ATT&CK T1040 Network Sniffing, T1557 Adversary-in-the-Middle) through encrypted overlays, inconsistent branch configuration through central templates, and lateral spread between site segments through VPN or segment separation on the fabric.

What it leaves open: malicious or risky web destinations reached from a local breakout, unsanctioned SaaS use and data leaving to cloud storage, remote users who never touch the fabric, and broad network reachability once a device is inside a segment. An attacker who phishes a branch user lands on a trusted network segment, and SD-WAN policy rarely knows who that user is.

Concentrated risk: the orchestrator and controllers. In February 2026 NSA, CISA and their Australian, Canadian, New Zealand and UK partners issued a joint advisory, Exploitation of SD-WAN Appliances, describing actors who exploited an authentication bypass in a widely deployed SD-WAN controller, added a rogue control-plane peer and escalated to root for long-term persistence. The mitigations read as management plane hygiene: keep control components behind a firewall, isolate management interfaces, forward logs to a remote syslog server, use pairwise keying for control and data planes, and keep session timeouts short. This is ATT&CK T1190 Exploit Public-Facing Application against the one system that can reconfigure every branch.

SASE

What it defends well: web-borne threats through inline inspection, data loss to SaaS and personal cloud storage through CASB and DLP, and exposure of private applications through per-application brokering instead of network-level VPN access (ATT&CK T1133 External Remote Services is the technique that VPN-heavy designs leave exposed). It also applies the same policy to a user in a branch, at home or travelling.

What it leaves open: traffic that never reaches the service (devices without an agent, IoT and OT at the branch, east-west traffic inside a site), and anything the provider is configured not to inspect, such as TLS bypass categories.

Concentrated risk: your admin tenant and the provider. Whoever holds administrator rights in the SASE console can change policy for every user, so admin accounts, API tokens and identity provider federation become the crown jewels. You also inherit the provider's availability: if a point of presence fails and fail-open is configured, traffic leaves uninspected; if fail-closed, users lose access.

Architecture in words: three common designs

  1. SD-WAN with central inspection. Branch edges build tunnels to regional hubs or the data centre, where a firewall stack inspects internet-bound traffic before it leaves. Security stays where it was; the cost is backhaul latency for SaaS and a hub that remains a single inspection point.
  2. Dual-vendor SASE. Branch SD-WAN edges keep doing path selection, and send internet and SaaS traffic through IPsec or GRE tunnels to a separate security service edge provider. Remote users connect to the same provider with an agent. Two consoles and two policy models must stay aligned, but each part can be chosen on its own merits and replaced separately.
  3. Single-vendor SASE. One provider supplies the branch edge, the overlay and the security cloud under one policy and one console. Fewer integration seams and one identity integration, at the cost of deeper dependence on one provider's coverage, roadmap and outage behaviour.

CISA's TIC 3.0 Branch Office Use Case (version 2.0, April 2023) describes the same patterns from a policy angle: branch traffic to cloud services, to the web and to the agency campus may go direct, through the campus, or through a CASB or other security as a service provider, and agencies may place capabilities at the branch, the CASB or the cloud provider so long as commensurate protections are maintained. It names SD-WAN techniques as a way to add resilient paths and requires policy enforcement parity across all of them. That parity requirement is the practical test for any design on this list.

Branch security gaps to close before you decide

GapHow it appears with SD-WAN aloneHow SASE addresses itWhat still needs local control
Local internet breakoutSaaS and web traffic leaves the branch without central inspectionBreakout traffic is tunnelled to the nearest point of presence for inspectionBypass lists and fail-open behaviour must be reviewed
Guest and IoT devicesPlaced on segments, often with broad internet accessSite tunnel inspection covers their internet traffic, not their local trafficLocal segmentation and network access control
Remote and travelling usersNot covered without a separate VPNAgent-based access with the same policyEndpoint posture and patching
East-west traffic in a siteSegment rules on the edge, if configuredUsually not visible to the cloud serviceEdge firewalling and microsegmentation
DNSLocal resolvers or ISP DNSProtective DNS and DNS filtering in the serviceBlocking direct DNS egress so clients cannot bypass
Edge device exposureEdges and controllers reachable from the internetFewer inbound exposures if the provider brokers accessEdge hardening, update cadence and management isolation

The joint Five Eyes guidance Security Considerations for Edge Devices (February 2025, published by Canada's Cyber Centre with ASD's ACSC, NCSC-NZ, NCSC-UK, CISA and NSA) applies to SD-WAN edges as it does to firewalls and VPN gateways: no management interfaces exposed to the internet, management through an out-of-band network from dedicated workstations, phishing-resistant MFA for all administrative access, and centralised off-device logging with alerts on administrative logons and configuration changes. Moving to SASE does not remove the branch edge; it changes what the edge does. See the Edge Device Security Guide.

Identity-aware access: the real dividing line

The deepest difference is that SD-WAN policy is written against networks and applications, while SASE policy is written against identities. The UK NCSC's zero trust architecture principles include knowing your user, service and device identities, assessing device and service health, using policies to authorise requests, and not trusting any network, including your own. SD-WAN can satisfy "do not trust the network" for the transport. It cannot by itself satisfy the identity and device health principles, because the edge sees an IP address and an application signature, not a user and a posture.

  • User identity: SASE integrates with the identity provider so policy can say "finance group, managed device, may reach the payments application" rather than "VLAN 30 may reach 10.20.0.0/16". See the ZTNA Guide.
  • Device posture: agent or identity provider signals (managed, encrypted, compliant) can gate access per session, which SD-WAN segment membership cannot.
  • Non-human identities: both models depend on machine identities: SD-WAN edge certificates and serial allowlists, controller peer identities, SASE connector credentials, API tokens used by automation and the federation trust with the identity provider. Inventory them with owners, rotate tokens, and alert when any of them is used from a new source. The 2026 advisory's rogue peer technique is an attack on exactly this layer.
  • Administrators: separate network administration from security policy administration, use named accounts with phishing-resistant MFA, and log every policy change to a system the administrators cannot alter.

Identity-aware policy is only as good as the identity data. A SASE rollout often exposes stale groups, shared accounts and unmanaged devices that the network-based model was hiding. Budget time to clean them up before switching policies from monitor to enforce.

Decision criteria

SignalPoints towards SD-WAN alone (for now)Points towards dual-vendor SASEPoints towards single-vendor SASE
Traffic patternMostly site to data centre on private pathsHeavy SaaS and internet use, existing SD-WAN investmentHeavy SaaS and internet use, WAN due for refresh
Remote workforceSmall, already on a well-managed access serviceLarge, served by a separate security service edgeLarge, and you want one policy for branch and remote
Existing inspectionStrong central stack with capacity for backhaulCentral stack ageing or overloadedCentral stack being retired
Team structureNetwork team owns everythingSeparate network and security teams with their own toolsConverged team, or appetite to converge
Supplier risk tolerancePrefer to keep inspection in housePrefer to avoid single-provider dependenceAccept deeper dependence for simpler operations
Regulatory constraintsData must not be inspected outside your estateNeed to choose inspection regions independently of WANProvider's regions and data handling meet requirements

Small organisations: with a handful of sites and a cloud-first workforce, a security service edge for users plus simple branch connectivity often beats a full SD-WAN fabric. Large organisations: dual-vendor is common during transition because SD-WAN contracts and security tooling rarely expire together; plan the end state, then decide whether convergence is worth a second migration.

Worked scenario

A retailer runs 140 stores on an SD-WAN fabric with MPLS and broadband underlays, backhauling all internet traffic to two data centres for inspection. Store systems now use cloud point of sale and SaaS stock management, and head office staff work hybrid through a legacy VPN. Problems: backhaul latency at peak trading, VPN concentrators near end of support, and no visibility of SaaS use from stores.

  1. Classify traffic. Payment and store systems traffic stays on the SD-WAN fabric to the data centre and payment processors; SaaS, web and staff internet traffic is the candidate for breakout.
  2. Choose the model. The SD-WAN contract has three years left and performs well, so the retailer chooses dual-vendor: keep SD-WAN for transport and segmentation, add a security service edge for breakout traffic and remote users.
  3. Close the store gaps. Guest Wi-Fi and IoT stay on separate segments with egress only through the security service tunnel; direct DNS egress is blocked at the edge.
  4. Move remote users first. Head office users move from VPN to per-application access, which retires the concentrators and tests identity provider integration on a population with good support.
  5. Pilot store breakout. Ten stores send SaaS and web traffic to the service in monitor mode for two weeks, comparing blocked categories and latency against backhaul. Rollback is a template change that restores the default route to the hubs.
  6. Govern both planes. SD-WAN and SASE admin roles are split, API tokens are inventoried, and both consoles forward logs to the SIEM with alerts on policy and peer changes.

Migration paths and rollback

  1. Inventory first: sites, underlays, applications, current inspection points and every bypass. Map data flows, as NSA's network and environment pillar guidance (March 2024) recommends, including any that are still unencrypted.
  2. Decide the end state: SD-WAN with central inspection, dual-vendor or single-vendor, and write down which gaps remain owned locally.
  3. Integrate identity: connect the identity provider, clean groups and decide device posture signals before any enforcement.
  4. Monitor mode: steer traffic through the security service with logging only, and compare outcomes with the existing stack.
  5. Pilot rings: remote users, then low-risk sites, then high-traffic sites; keep the old path live and documented as the rollback route.
  6. Enforce and retire: switch to blocking per ring, then decommission backhaul inspection only after parity is shown across all paths.

Detailed sequencing, parallel running and cutover controls are in the SASE Migration guide. If you are still choosing a WAN underlay, see SD-WAN vs MPLS.

Detection and measurement

  • Management plane events: admin logons, policy changes, new control-plane peers, new or revoked edge certificates, API token use from new addresses (both SD-WAN orchestrator and SASE tenant).
  • Path changes: sites failing over to an uninspected path, tunnels to the security service going down, fail-open activations.
  • Bypass growth: changes to TLS inspection exceptions and steering bypass lists.
  • Access decisions: denied and step-up events by user and device posture, to confirm identity policy is enforcing as designed.
MetricWhat good looks like
Share of internet-bound traffic inspected, by siteClose to 100 per cent, with documented exceptions
Sites with policy parity across all underlaysAll sites, verified after every template change
Management plane changes reconciled to change recordsEvery change matched within one working day
Machine credentials with a named owner and rotation dateAll edge, connector and API credentials
Remote access sessions brokered per application rather than network VPNRising towards the target set in the migration plan

Common failure patterns

  • Calling SD-WAN "secure" because it encrypts. Encryption protects the underlay; it does nothing about what a user downloads at a breakout point.
  • Turning on local breakout before inspection is ready. Performance improves on day one and visibility disappears the same day.
  • Uncontrolled bypass lists. Exceptions added for performance or broken applications accumulate until much of the traffic is uninspected.
  • Exposed management planes. Orchestrators, controllers and SASE consoles reachable from the internet with password-only admin accounts, the pattern the 2026 joint advisory and the edge device guidance warn against.
  • Two consoles, two truths. In dual-vendor designs, segment and policy changes on one side are not mirrored on the other, so parity quietly drifts.
  • Forgetting non-agent devices. Printers, cameras and store systems never see the SASE agent, so their access depends entirely on branch segmentation.

Framework mapping

FrameworkReferenceRelevance
NIST CSF 2.0PR.IR-01, PR.DS-02, PR.AA-03, DE.CM-01Network protection, data in transit, authentication of users, services and hardware, network monitoring
NIST SP 800-53 Rev. 5SC-7, SC-8, AC-4, AC-17, IA-3, CM-7, SI-4Boundary protection, transmission protection, flow enforcement, remote access, device authentication, least functionality, monitoring
CIS Controls v8.1Control 12 (Network Infrastructure Management), Control 13 (Network Monitoring and Defense)Edge and fabric management, inspection and monitoring
ISO/IEC 27001:2022 Annex AA.8.20 Networks security, A.8.21 Security of network services, A.8.22 Segregation of networks, A.8.23 Web filteringNetwork controls, provider services, segmentation and web filtering
MITRE ATT&CKT1040, T1557, T1190, T1133, T1599Underlay interception, exposed edges and controllers, remote services, boundary bridging

Practitioner checklist

  • Write down which problem you are solving: transport, inspection, remote access or all three.
  • Map every site's traffic by destination and record where it is inspected today.
  • Do not enable local breakout at any site until inspection for that traffic is live.
  • Choose single-vendor or dual-vendor on contract timing, team structure and supplier risk, and record the reasoning.
  • Remove internet exposure of SD-WAN and SASE management interfaces and require phishing-resistant MFA for admins.
  • Inventory edge certificates, connector credentials and API tokens with owners and rotation dates.
  • Run every change in monitor mode and pilot rings with a documented rollback route.
  • Review bypass and TLS inspection exception lists monthly.
  • Forward both planes' logs off-device and alert on peer, policy and admin changes.

Standards and references

Related NHI Mgmt Group resources: SASE and SSE Guide · SD-WAN Security Guide · SASE Migration · SD-WAN vs MPLS · CASB vs SWG

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.