What is being compared
The two subjects
Both are one scene apart, not two architectures apart — which is why the diagram below is a single scene with each subject's claim on it marked, rather than two pictures to hold in your head at once.
-
Secure access service edge (SASE)
Marked "only in Secure access service edge (SASE)" in the delta below.
The security services plus the wide-area networking that carries branch and site traffic to them, presented and ideally governed as one architecture.
-
Security service edge (SSE)
Marked "only in Security service edge (SSE)" in the delta below.
The cloud-delivered security services alone: one policy decision and a service chain for user traffic, with the wide-area network left as it is.
Architecture diagram
One service edge, with and without the networking half
A remote user and a branch site both reach a cloud service edge, where identity and device context feed one policy decision and a chain of security services before permitted traffic continues to internet, SaaS or private destinations. The branch path, its edge device and the wide-area orchestrator belong to SASE alone; everything else is common to both.
Step 1 / 5Reach the service edge from both directions
Architectural planes
Annotations
The same architecture, read top to bottom
Shared and public networks
UNTRUSTEDNot controlled and not trusted; anyone may be present on itRemote worker · Managed laptop · Adversary
- PersonRemote workerWorking from a shared network
- DeviceManaged laptopAgent reports posture and tunnels out
- AdversaryAdversaryScanning circuits and replaying credentials
Reaches this boundary
- BlockedAdversary stopped at Security service chainRefused: unknown device, no posture signal
- BlockedAdversary stopped at Security service chainRefused where branch breakout is steered inward
Within this boundary
- RequestRemote worker to Managed laptopOpens an application on the laptop
Branch site network
MANAGEDOperated or administered on the enterprise's behalfBranch site · Branch edge device
- DeviceBranch siteStaff, printers and local systems
- API or gatewayBranch edge device
- Per-application path selection
- Link health measurement
- Overlay tunnel to the service edge
- Local breakout, steered by policy
Reaches this boundary
- ContextWide-area orchestrator to Branch edge devicePath policy for each application class
- AttackAdversary to Branch edge deviceProbe against the branch internet circuit
Within this boundary
- RequestBranch site to Branch edge deviceSite traffic reaches the branch edge
Enterprise control plane
INTERNALOperated by the enterprise itselfWide-area orchestrator · Identity provider · Security operations
- Security controlWide-area orchestratorPath policy and circuit health
- Identity providerIdentity providerSign-on, second factor and risk claims
- Security controlSecurity operationsDetection, investigation and tuning
Reaches this boundary
- ContextRemote worker to Identity providerSign-on and a second factor
- TelemetryBranch edge device to Wide-area orchestratorLoss, latency and jitter per circuit
- TelemetryService edge gateway to Security operationsSession and admission events
- TelemetryPolicy engine to Security operationsEvery verdict and the signals behind it
- TelemetrySecurity service chain to Security operationsInspection and data-handling events
Cloud service edge
MANAGEDOperated or administered on the enterprise's behalfService edge gateway
- API or gatewayService edge gatewayTunnel termination and request admission
Policy control
MANAGEDOperated or administered on the enterprise's behalfPolicy engine
- Policy decisionPolicy engineAllow, step up, isolate or block
Security services
MANAGEDOperated or administered on the enterprise's behalfSecurity service chain
- Security controlSecurity service chain
- Brokered private application access
- Secure web gateway
- Cloud application controls
- Firewall as a service
- Data loss prevention
- Security controlSecurity service chain
Reaches this boundary
- RequestManaged laptop to Service edge gatewayAgent tunnel to the nearest edge
- RequestBranch edge device to Service edge gatewayOverlay, steered per application
- ContextIdentity provider to Policy engineIdentity, group and risk claims
- ContextManaged laptop to Policy engineDevice posture at the time of the request
- AttackAdversary to Service edge gatewayStolen credential replayed at the edge
Within this boundary
- RequestService edge gateway to Policy engineRequest presented for a decision
- RequestPolicy engine to Security service chainVerdict and the controls it requires
Public internet
UNTRUSTEDNot controlled and not trusted; anyone may be present on itWeb destination
- ApplicationWeb destinationUnclassified internet site
Reaches this boundary
- RequestSecurity service chain to Web destinationPermitted web traffic, inspected
Sanctioned SaaS
THIRD PARTYOperated by another organisation under its own termsSaaS platform
- ApplicationSaaS platformSanctioned but not enterprise-operated
Reaches this boundary
- RequestSecurity service chain to SaaS platformPermitted SaaS traffic
Private applications
INTERNALOperated by the enterprise itselfPrivate application · Application records
- ApplicationPrivate applicationNo inbound exposure to the internet
- Data storeApplication records
Reaches this boundary
- RequestSecurity service chain to Private applicationBrokered session, no network route
Within this boundary
- RequestPrivate application to Application recordsApplication reads its own records
Legend
Participants
- PersonSomeone making a request, and the identity they assertCircle with a person mark
- DeviceAn endpoint the request is made from, with its own postureRounded rectangle with an inset panel
- ApplicationAn application that serves or holds dataRounded rectangle with a panel stacked behind it
- Identity providerThe authority that asserts who is askingHexagon marked ID
- API or gatewayA network endpoint that terminates or forwards connectionsCapsule marked API
- Data storeWhere data restsCylinder
- Security controlA protective control, and the control plane's own observationRounded octagon with a check mark
- AdversaryA hostile party attempting somethingDiamond marked with an exclamation mark
- Policy decisionWhere policy is evaluated and one verdict is reachedDiamond with a split-path mark on its trailing point
Relationships
- RequestTraffic asking to reach a destinationSolid line. Filled arrowhead.
- TelemetryEvents reported onward for observation — observation, not controlDotted line. Small open arrowhead.
- ContextA signal that informs a decision without carrying the request itselfDash-dot line. Open chevron.
- AttackAn attempt made by an adversaryShort-dashed heavy line. Barbed open arrowhead.
- BlockedAn attempt terminated by the control the line starts fromHeavy line that stops short of its target. Perpendicular bar, and deliberately no arrowhead: the attempt did not arrive.
Boundaries
- Internal boundaryOperated by the enterprise itselfSolid outline. Badge reads INTERNAL.
- Managed boundaryOperated or administered on the enterprise's behalfDashed outline with a subtle wash. Badge reads MANAGED.
- Third party boundaryOperated by another organisation under its own termsDash-dot outline. Badge reads THIRD PARTY.
- Untrusted boundaryNot controlled and not trusted; anyone may be present on itDotted outline with a hatched leading corner. Badge reads UNTRUSTED.
Text alternative
The request, step by step
Reach the service edge from both directions
A remote user tunnels out from a managed laptop, and a branch site reaches the same edge through a device that chooses a path per application. Only the first of those two paths is in scope for an SSE purchase.
- Request: Remote worker to Managed laptop, “Opens an application on the laptop”
- Request: Managed laptop to Service edge gateway, “Agent tunnel to the nearest edge”
- Request: Branch site to Branch edge device, “Site traffic reaches the branch edge”
- Request: Branch edge device to Service edge gateway, “Overlay, steered per application”
Emphasised at this step: Service edge gateway.
Assemble the context for one decision
The service combines the asserted identity with device posture, the requested application and the destination. Both categories do this the same way, from the same signals.
- Context: Remote worker to Identity provider, “Sign-on and a second factor”
- Context: Identity provider to Policy engine, “Identity, group and risk claims”
- Context: Managed laptop to Policy engine, “Device posture at the time of the request”
- Request: Service edge gateway to Policy engine, “Request presented for a decision”
Emphasised at this step: Policy engine.
Decide once, then enforce
One engine decides whether to allow, challenge, isolate or block, and names the services the traffic must pass. The chain is identical in both categories, which is why the delta is not a security feature.
- Request: Policy engine to Security service chain, “Verdict and the controls it requires”
Emphasised at this step: Security service chain.
Route permitted traffic to its destination
Allowed traffic continues to the internet, a sanctioned SaaS platform or a private application, which then reads its own records. The private session is brokered, so no route to the wider network is created.
- Request: Security service chain to Web destination, “Permitted web traffic, inspected”
- Request: Security service chain to SaaS platform, “Permitted SaaS traffic”
- Request: Security service chain to Private application, “Brokered session, no network route”
- Request: Private application to Application records, “Application reads its own records”
Emphasised at this step: Private application.
Steer the network and observe both planes
Circuit measurements drive path policy at each site, while admission, verdict and inspection events reach security operations. With the networking half absent, the path data comes from a different supplier and has to be correlated by hand.
- Context: Wide-area orchestrator to Branch edge device, “Path policy for each application class”
- Telemetry: Branch edge device to Wide-area orchestrator, “Loss, latency and jitter per circuit”
- Telemetry: Service edge gateway to Security operations, “Session and admission events”
- Telemetry: Policy engine to Security operations, “Every verdict and the signals behind it”
- Telemetry: Security service chain to Security operations, “Inspection and data-handling events”
Emphasised at this step: Security operations.
Attempts shown, and where each one stops
Adversary attempts:
- Attack: Adversary to Service edge gateway, “Stolen credential replayed at the edge”
- Attack: Adversary to Branch edge device, “Probe against the branch internet circuit”
- Terminated here by Security service chain — “Refused: unknown device, no posture signal”. The line stops short and takes a bar, not an arrowhead.
- Terminated here by Security service chain — “Refused where branch breakout is steered inward”. The line stops short and takes a bar, not an arrowhead.
Annotations
This line is the whole difference
Relationship — Request: Branch edge device to Service edge gateway, “Overlay, steered per application”
An SSE purchase does not include it. Something still has to carry site traffic to the enforcement point, and whatever that is remains your design, your contract and your failure domain. The security services on the far side are the same either way.
What an SSE purchase leaves unsolved
Boundary “Branch site network” (managed)
Circuits, routers, path selection and local internet breakout stay where they are. That is often the correct choice, but it should be a decision with an owner rather than a gap discovered after the security service is live.
A local circuit is an uninspected path
Relationship — Attack: Adversary to Branch edge device, “Probe against the branch internet circuit”
Traffic leaving a site directly reaches the internet without meeting the service chain at all. The networking half exists partly to pull that traffic inward so the same policy applies to it, rather than relying on a separate device at every site to be configured consistently.
The service chain is not the differentiator
Component “Security service chain”
Brokered private access, web inspection, cloud application controls, firewalling and data-loss prevention appear in both categories. Judging two shortlists by counting these services compares them on the axis where they agree.
One decision is the point of both
Component “Policy engine”
Identity, device posture, application, destination and current risk reaching one engine is what makes a verdict reviewable. Several tools each reaching their own verdict is the state both categories were defined to replace.
Convergence is a claim to test, not a label to read
Boundary “Cloud service edge” (managed)
Two products behind one invoice can still hold two policy models, two consoles and two identity integrations. Ask which single object holds the policy, whether one change updates both halves, and whether the events from each half share identifiers and clocks.
The delta
What each one adds, and what they share
Computed from the claims each subject makes on the one scene above, not drawn separately. Every box and every path in that diagram is claimed by all of these designs or by exactly one of them, and the bands below are that partition.
37 elements in the scene · 28 shared · 9 in SASE · none in SSE
Shared by both designs
12 boxes · 16 paths
Claimed by Secure access service edge (SASE) and Security service edge (SSE) alike. This is the larger part of the scene, and saying so is part of the answer: two things worth comparing usually agree about most of the architecture, and a comparison that hides the agreement makes the difference look bigger than it is.
- Remote worker Person — Working from a shared network Shared by both designs.
- Managed laptop Device — Agent reports posture and tunnels out Shared by both designs.
- Adversary Adversary — Scanning circuits and replaying credentials Shared by both designs.
- Service edge gateway API or gateway — Tunnel termination and request admission Shared by both designs.
- Policy engine Policy decision — Allow, step up, isolate or block Shared by both designs.
- Security service chain Security control Shared by both designs.
- Identity provider Identity provider — Sign-on, second factor and risk claims Shared by both designs.
- Security operations Security control — Detection, investigation and tuning Shared by both designs.
- Web destination Application — Unclassified internet site Shared by both designs.
- SaaS platform Application — Sanctioned but not enterprise-operated Shared by both designs.
- Private application Application — No inbound exposure to the internet Shared by both designs.
- Application records Data store Shared by both designs.
- Remote worker to Managed laptop Request — Opens an application on the laptop Shared by both designs.
- Managed laptop to Service edge gateway Request — Agent tunnel to the nearest edge Shared by both designs.
- Remote worker to Identity provider Context — Sign-on and a second factor Shared by both designs.
- Identity provider to Policy engine Context — Identity, group and risk claims Shared by both designs.
- Managed laptop to Policy engine Context — Device posture at the time of the request Shared by both designs.
- Service edge gateway to Policy engine Request — Request presented for a decision Shared by both designs.
- Policy engine to Security service chain Request — Verdict and the controls it requires Shared by both designs.
- Security service chain to Web destination Request — Permitted web traffic, inspected Shared by both designs.
- Security service chain to SaaS platform Request — Permitted SaaS traffic Shared by both designs.
- Security service chain to Private application Request — Brokered session, no network route Shared by both designs.
- Private application to Application records Request — Application reads its own records Shared by both designs.
- Service edge gateway to Security operations Telemetry — Session and admission events Shared by both designs.
- Policy engine to Security operations Telemetry — Every verdict and the signals behind it Shared by both designs.
- Security service chain to Security operations Telemetry — Inspection and data-handling events Shared by both designs.
- Adversary to Service edge gateway Attack — Stolen credential replayed at the edge Shared by both designs.
- Security service chain to Adversary Blocked — Refused: unknown device, no posture signal Shared by both designs.
Only in Secure access service edge (SASE)
3 boxes · 6 paths
Present when the scene is read as Secure access service edge (SASE), and absent when it is read as Security service edge (SSE). 3 boxes and 6 paths.
- Branch site Device — Staff, printers and local systems Only in Secure access service edge (SASE).
- Branch edge device API or gateway Only in Secure access service edge (SASE).
- Wide-area orchestrator Security control — Path policy and circuit health Only in Secure access service edge (SASE).
- Branch site to Branch edge device Request — Site traffic reaches the branch edge Only in Secure access service edge (SASE).
- Branch edge device to Service edge gateway Request — Overlay, steered per application Only in Secure access service edge (SASE).
- Wide-area orchestrator to Branch edge device Context — Path policy for each application class Only in Secure access service edge (SASE).
- Branch edge device to Wide-area orchestrator Telemetry — Loss, latency and jitter per circuit Only in Secure access service edge (SASE).
- Adversary to Branch edge device Attack — Probe against the branch internet circuit Only in Secure access service edge (SASE).
- Security service chain to Adversary Blocked — Refused where branch breakout is steered inward Only in Secure access service edge (SASE).
Only in Security service edge (SSE)
nothing
Nothing. Every box and every path Security service edge (SSE) claims is also claimed by Secure access service edge (SASE), so this band is empty — which is what a strict subset looks like, and is the finding rather than a gap. A box invented here to balance the layout would assert an architecture that does not exist.
Dimension by dimension
How each behaves, on each axis that matters
| Dimension | Secure access service edge (SASE) | Security service edge (SSE) |
|---|---|---|
| What the category includes | Cloud-delivered security services and software-defined wide-area networking, named as one architecture with one policy intent. | Cloud-delivered security services only. The transport that gets traffic to them is out of scope and stays whatever it already was. |
| How branch and site traffic is handled | An edge device at each site steers traffic per application and link health, and the overlay terminates at the same enforcement point. | Sites reach the service however they already do, so existing circuits, routers and any backhaul remain your design and your cost. |
| Who is served first | Users and sites together, which is what makes it a wide-area network replacement rather than only a remote-access replacement. | Users and their devices, wherever they are. Site-to-site and machine-to-machine traffic is usually served by something else. |
| Security capability of the inspection chain | The same set of services: brokered private access, web gateway, cloud application controls, firewall and data-loss prevention. | The same set of services. This is not where the two differ, which is why comparing them on features answers the wrong question. |
| Traffic the enforcement point never sees | Branch internet breakout can be pulled through the service, so a local circuit does not become an uninspected path by default. | Any site traffic that leaves through a local circuit without an agent or tunnel is outside the service and outside its telemetry. |
| Procurement and integration risk | The label does not prove convergence. Two acquired products behind one invoice can still hold two policy models and two consoles. | The scope is narrower and easier to judge, but you now own the seam between the security service and whoever provides the network. |
| Operational ownership | Forces the network and security teams into one policy conversation, which is a benefit only if someone can arbitrate between them. | Leaves the existing split intact. That is lower disruption and also a missed chance to remove contradictions between the two policies. |
| Evidence available to investigators | Path, decision and inspection events from one service, so a slow or refused session can be traced end to end. | Decision and inspection events only. Correlating them with wide-area path data means joining two vendors' telemetry by hand. |
In full
The argument, at length
Security service edge and secure access service edge are not competing designs. SSE is a proper subset: the cloud-delivered security services, without the wide-area networking that carries site traffic to them. Everything in an SSE service chain also appears in a SASE service chain, which is why a feature-level comparison finds almost nothing and misses the actual decision.
The decision is about scope, and it has a cost either way. Choosing SSE leaves wide-area connectivity exactly as it is: the circuits, the routers, the path selection and, most importantly, any local internet breakout at each site. None of that becomes wrong, but none of it becomes solved either. If sites already break out locally through a device that nobody has reviewed in three years, an SSE purchase does not touch that path, and the traffic on it never reaches the policy engine you just bought.
Choosing SASE takes the network on as well, which is a larger programme with a larger blast radius and a harder internal negotiation. It also removes a seam. When one service both carries the traffic and decides about it, a slow session and a refused session can be investigated in one place, and one policy statement governs both users and sites.
The claim to test is convergence. Neither term is a standard, and no specification says what a product must do to use either label, so the word on the datasheet distinguishes nothing. Two acquired products sold together can carry the SASE label while holding two policy models, two consoles and two sets of logs that do not share identifiers. That arrangement is an SSE purchase and a wide-area network purchase with a single invoice attached, and it should be evaluated as such. The useful questions are which single object holds the policy, whether one change takes effect on both halves, and whether the events from each half can be correlated without manual work.
Choosing
When each one is the right answer
Grouped by subject rather than interleaved, and every subject has cases of its own — required by the content model, because a comparison in which one side never wins is an argument dressed as a comparison.
Choose Secure access service edge (SASE) when
-
Many sites, ageing circuits and a refresh already due
When wide-area contracts and branch hardware are up for renewal anyway, converging the two halves avoids designing the network twice and puts one policy decision in front of both users and sites.
-
Branch internet breakout is already an uninspected path
If sites break out locally with only a local firewall, the networking half is what pulls that traffic back under the same policy, which an SSE purchase on its own does not do.
Choose Security service edge (SSE) when
-
The workforce is remote and the wide-area network is not the problem
When traffic is mostly user-to-SaaS and user-to-internet, and existing circuits are adequate, SSE solves the part that hurts without a network replacement programme attached to it.
-
Circuits are contracted and cannot be replaced yet
A long carrier commitment or a regulated site design can make the network half undeliverable for years. Buying the security half now is honest sequencing, provided branch breakout is still inspected somehow.
Sources
-
NIST SP 800-207
Defines the policy engine, policy administrator and policy enforcement point split that both an SSE service and a SASE service implement.
-
NIST SP 800-215
Guide to a Secure Enterprise Network Landscape
Covers the shift of enterprise network security services out of the data centre, which is the movement both categories name.
-
NIST SP 1800-35
Implementing a Zero Trust Architecture: High-Level Document
Reference builds assembled from commercial products, useful for judging whether two halves of a service are genuinely one architecture.
-
CISA v2.0
Treats network and application pillars as needing to advance together, which is the coordination claim a SASE purchase makes.