Cloud and networks

Security service edge (SSE)

SSE is the cloud-delivered security half of SASE, typically a secure web gateway, a cloud access security broker, ZTNA and data loss prevention under one policy, with the wide-area networking half deliberately excluded.

Architecture diagram

A request through a security service edge

Users on any network and traffic from a branch site reach a cloud security service edge, where one policy decision selects a web gateway, a cloud access security broker, a private-access broker and data loss prevention. The wide-area transport carrying branch traffic sits outside the service and outside its protection.

Step 1 / 5Reach the service edge from anywhere

Architectural planes

Annotations

The same architecture, read top to bottom

  1. Users on any network

    UNTRUSTEDNot controlled and not trusted; anyone may be present on it

    Remote worker · Managed laptop · Adversary

    • PersonRemote workerHome or public network, no site involved
    • DeviceManaged laptopAgent forwards traffic to the service edge
    • AdversaryAdversary

    Reaches this boundary

    • BlockedAdversary stopped at Secure web gatewayRefused: newly registered domain, no category
    • BlockedAdversary stopped at ZTNA brokerNo application named, so no session exists
    • BlockedManaged laptop stopped at Data loss preventionUpload of regulated data refused

    Within this boundary

    • RequestRemote worker to Managed laptopOpens a site, a SaaS tenant or an application
    • AttackAdversary to Managed laptopLink to a credential-harvesting lookalike
  2. Branch and office sites

    MANAGEDOperated or administered on the enterprise's behalf

    Branch site

    • DeviceBranch siteUsers and devices behind one router
  3. Wide-area transport, bought separately

    MANAGEDOperated or administered on the enterprise's behalf

    Wide-area transport

    • DeviceWide-area transportCircuits and steering SSE does not include

    Reaches this boundary

    • RequestBranch site to Wide-area transportSite traffic over transport bought elsewhere
    • AttackAdversary to Wide-area transportAttacks the transport the service does not run
  4. Cloud security service edge

    MANAGEDOperated or administered on the enterprise's behalf

    Service edge ingress

    • API or gatewayService edge ingressAgent, proxy or tunnel termination
    • Policy control
      MANAGEDOperated or administered on the enterprise's behalf

      Policy engine

      • Policy decisionPolicy engineAllow, step up, isolate or block
    • Security services
      MANAGEDOperated or administered on the enterprise's behalf

      Secure web gateway · Cloud access security broker · ZTNA broker · Data loss prevention

      • Security controlSecure web gateway
        • URL and content categorisation
        • TLS inspection where permitted
        • Malware and script analysis
        • Browser isolation for risky sites
      • Security controlCloud access security broker
        • Sanctioned tenant enforcement
        • Unsanctioned SaaS discovery
        • API posture checks on tenants
        • Session controls inside SaaS
      • API or gatewayZTNA brokerBrokered access to one application at a time
      • Security controlData loss prevention
        • Content classification
        • Upload and download inspection
        • Block, mask, or allow and log

    Reaches this boundary

    • RequestManaged laptop to Service edge ingressTraffic forwarded to the nearest edge
    • RequestWide-area transport to Service edge ingressArrives at the security service edge
    • ContextIdentity provider to Policy engineIdentity, group and risk claims
    • ContextManaged laptop to Policy engineDevice posture from the local agent
    • AttackAdversary to ZTNA brokerProbes for exposed private applications

    Within this boundary

    • RequestService edge ingress to Policy engineRequest presented for one decision
    • RequestPolicy engine to Secure web gatewayWeb traffic, with the inspection it requires
    • RequestPolicy engine to Cloud access security brokerSaaS traffic, with the tenant rules that apply
    • RequestPolicy engine to ZTNA brokerPrivate application access, scoped to one app
    • RequestPolicy engine to Data loss preventionData rules for this user and destination
    • RequestData loss prevention to Cloud access security brokerInspected upload passed on
  5. Enterprise control plane

    INTERNALOperated by the enterprise itself

    Identity provider · Security operations

    • Identity providerIdentity providerSingle sign-on, MFA and risk claims
    • Security controlSecurity operationsOne event stream from four services

    Reaches this boundary

    • ContextRemote worker to Identity providerAuthentication and a second factor
    • TelemetryPolicy engine to Security operationsEvery verdict and the signals behind it
    • TelemetrySecure web gateway to Security operationsWeb inspection and isolation events
    • TelemetryCloud access security broker to Security operationsSaaS usage, including unsanctioned tenants
    • TelemetryData loss prevention to Security operationsData-handling decisions and their matches
  6. Public internet

    UNTRUSTEDNot controlled and not trusted; anyone may be present on it

    Web destination

    • ApplicationWeb destinationUnclassified internet site

    Reaches this boundary

    • RequestSecure web gateway to Web destinationPermitted web traffic, inspected
  7. Sanctioned SaaS

    THIRD PARTYOperated by another organisation under its own terms

    SaaS platform

    • ApplicationSaaS platformSanctioned but not enterprise-operated

    Reaches this boundary

    • RequestCloud access security broker to SaaS platformPermitted traffic to the sanctioned tenant
  8. Private applications

    INTERNALOperated by the enterprise itself

    Private application

    • ApplicationPrivate applicationReached by broker, with no inbound exposure

    Reaches this boundary

    • RequestZTNA broker to Private applicationBrokered session, no network route

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
  • 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

  1. Reach the service edge from anywhere

    A remote laptop forwards its traffic directly, and branch traffic travels over transport the organisation buys and operates separately before it arrives. Both paths end at the same enforcement point, which is what makes one policy possible.

    • Request: Remote worker to Managed laptop, “Opens a site, a SaaS tenant or an application”
    • Request: Managed laptop to Service edge ingress, “Traffic forwarded to the nearest edge”
    • Request: Branch site to Wide-area transport, “Site traffic over transport bought elsewhere”
    • Request: Wide-area transport to Service edge ingress, “Arrives at the security service edge”

    Emphasised at this step: Service edge ingress.

  2. Assemble identity and device context

    The service combines the asserted identity and its risk claims with device posture from the local agent and the destination being requested. Traffic arriving from a branch without an agent carries less context, and the policy has to say what that means.

    • Context: Remote worker to Identity provider, “Authentication and a second factor”
    • Context: Identity provider to Policy engine, “Identity, group and risk claims”
    • Context: Managed laptop to Policy engine, “Device posture from the local agent”
    • Request: Service edge ingress to Policy engine, “Request presented for one decision”

    Emphasised at this step: Identity provider.

  3. One decision selects the services

    A single policy engine decides the verdict and which of the constituent services the request must pass through. This step is the entire argument for buying the services together, and the one to verify in a trial rather than assume.

    • Request: Policy engine to Secure web gateway, “Web traffic, with the inspection it requires”
    • Request: Policy engine to Cloud access security broker, “SaaS traffic, with the tenant rules that apply”
    • Request: Policy engine to ZTNA broker, “Private application access, scoped to one app”
    • Request: Policy engine to Data loss prevention, “Data rules for this user and destination”

    Emphasised at this step: Policy engine.

  4. Each service enforces its own part

    The web gateway inspects internet traffic, data controls inspect the content of an upload before the cloud access security broker passes it to a sanctioned tenant, and the private-access broker exposes one application rather than the network holding it.

    • Request: Secure web gateway to Web destination, “Permitted web traffic, inspected”
    • Request: Data loss prevention to Cloud access security broker, “Inspected upload passed on”
    • Request: Cloud access security broker to SaaS platform, “Permitted traffic to the sanctioned tenant”
    • Request: ZTNA broker to Private application, “Brokered session, no network route”

    Emphasised at this step: Data loss prevention.

  5. Collect the evidence in one place

    Verdicts, web inspection, SaaS usage and data-handling decisions reach security operations as one stream. Correlating them on identity and session is what turns four separate tools' worth of events into an account of what happened.

    • Telemetry: Policy engine to Security operations, “Every verdict and the signals behind it”
    • Telemetry: Secure web gateway to Security operations, “Web inspection and isolation events”
    • Telemetry: Cloud access security broker to Security operations, “SaaS usage, including unsanctioned tenants”
    • Telemetry: Data loss prevention to Security operations, “Data-handling decisions and their matches”

    Emphasised at this step: Security operations.

Attempts shown, and where each one stops

  • Adversary attempts:

    • Attack: Adversary to Managed laptop, “Link to a credential-harvesting lookalike”
    • Attack: Adversary to ZTNA broker, “Probes for exposed private applications”
    • Attack: Adversary to Wide-area transport, “Attacks the transport the service does not run”
    • Terminated here by Secure web gateway — “Refused: newly registered domain, no category”. The line stops short and takes a bar, not an arrowhead.
    • Terminated here by ZTNA broker — “No application named, so no session exists”. The line stops short and takes a bar, not an arrowhead.
  • Actions a control refused, with no adversary involved:

    • Blocked: Managed laptop stopped at Data loss prevention, “Upload of regulated data refused”. The line stops short of its target: the action did not complete.

Annotations

  1. The boundary that defines the term

    Boundary “Wide-area transport, bought separately” (managed)

    This zone is the half SSE leaves out. The circuits, their failover and the steering of traffic across them stay with whoever owned them before, and adding the networking half to what sits beside it is what makes the architecture SASE instead.

  2. An attempt with nothing stopping it here

    Relationship — Attack: Adversary to Wide-area transport, “Attacks the transport the service does not run”

    This path ends at no control, which is deliberate. Availability attacks, misconfigured routers and circuit failures land on infrastructure the security service neither operates nor observes, and no amount of inspection at the edge changes that.

  3. Not every refusal stops an attacker

    Relationship — Blocked: Managed laptop stopped at Data loss prevention, “Upload of regulated data refused”

    The same terminator appears whether the party stopped is an adversary or an employee doing their job. Most data loss prevention decisions are the second kind, which is why the explanation given to the user matters as much as the rule itself.

  4. What consolidation is supposed to buy

    Component “Policy engine”

    One engine reaching one verdict over identity, device, destination and data sensitivity is the difference between an SSE service and four products sold together. If each service keeps its own policy model, the gaps between them survive the purchase.

  5. Where two of these services overlap

    Component “Cloud access security broker”

    Inline SaaS controls duplicate much of what a web gateway does. The part that does not overlap is the tenant-side view reached through the platform's own interfaces, which sees sharing and configuration no traffic inspection can observe.

  6. ZTNA arrives as one service among four

    Component “ZTNA broker”

    The broker inside an SSE bundle is the same mechanism sold on its own, and inherits the same limits, including that its precision depends on how narrowly each application is defined. Buying it as part of a suite does not narrow the definitions for you.

  7. The set applied is an outcome

    Boundary “Security services” (managed)

    Which services a request passes through follows from the verdict, so a web request, a SaaS upload and a private application session traverse different paths from the same ingress point. The zone lists what is available, not what every request receives.

The short version

Four things to take away

  • It is a scope boundary, not a capability

    SSE names which services are bought together. It adds nothing that a web gateway, a cloud access security broker, a broker for private applications and data loss prevention do not already do individually.

  • One policy across the services is the claim to test

    The value is a single decision and a single set of logs governing web, SaaS, private applications and data movement. Four consoles behind one invoice deliver the gaps of four products.

  • Wide-area connectivity remains your problem

    SSE excludes SD-WAN by definition. Circuits, path selection, branch failover and application steering are unchanged the day the service goes live, and still need an owner and a budget.

  • The label does not fix the contents

    No standard says which services an SSE offering must include, so coverage of non-web protocols, unmanaged devices, inspection limits and API-based SaaS controls varies widely between products carrying the same name.

Mechanics

How it works

Security service edge names a purchase boundary. Take the SASE architecture, remove the wide-area networking half, and what remains is a set of cloud-delivered security services with one policy in front of them: web filtering, control over SaaS usage, brokered access to private applications, and inspection of the data moving through all three. Nothing in that set is new. What the term asserts is that they are bought and operated as one thing.

The split exists for two reasons, and it is worth separating them. Architecturally, security services and network transport have different lifecycles and different owners: a security team can move enforcement into a cloud service without touching a WAN contract that runs for another three years. Commercially, the two halves come from different heritages — proxy and cloud-security vendors on one side, routing and appliance vendors on the other — and each sells the half it already had. The category made the partial purchase describable, which is useful, and also made it look complete, which it is not.

That incompleteness is specific rather than vague. After an SSE deployment, branch circuits are the same circuits, path selection is the same path selection, and a degraded link on a Monday morning is exactly as much of a problem as it was the previous week. Traffic from a site still has to reach the service edge over transport the security service does not operate. Teams that treat SSE as the whole answer tend to discover this the first time connectivity, rather than policy, is what breaks.

Where SSE genuinely earns its place is consistency. A remote worker and a person at a desk in the office should meet the same rules, and the reason they historically did not is that one path went through an appliance and the other did not. Moving enforcement to a service every path reaches removes the difference, and one event stream across the four services answers questions that four consoles could not. Both of those depend on a single policy model rather than a single supplier, and only one of those two is visible on an invoice.

Because no standard defines the bundle, the contents are the thing to inspect. Two offerings using this name may differ on non-web protocol coverage, on unmanaged devices, on whether firewall-as-a-service is included, on how much of your TLS traffic you are actually able to decrypt, and on whether SaaS control extends to the platform’s own interfaces or stops at inline traffic. Read the service against the enforcement model in NIST SP 800-207 and the network-security services described in SP 800-215, and treat the acronym as a shorthand for the scope of the conversation rather than as a claim about coverage.

One concrete situation

A regional office retires its proxy appliance

Cutover weekend, one of eleven sites

  1. Context

    The site's ageing web proxy and its separate cloud-app monitoring tool each hold part of the policy, and neither covers the third of the workforce that now works from home.

  2. Decision

    The team moves web, SaaS and private application access to one SSE service so a single policy follows the user, and keeps the existing WAN contract because SSE does not replace it.

  3. Enforcement

    Traffic from laptops and from the site now reaches the service edge, where one verdict selects web filtering, SaaS controls, brokered private access and data inspection for each request.

  4. Observation

    Events from all four services land in one place, which immediately exposes uninspected paths the old split tooling had been hiding rather than creating.

Scope, in both directions

What it covers, and what it does not

Both lists are authored and required. A control's limits are part of what it is, so the second panel is not a caveat appended to the first — it is the other half of the answer.

What it protects

  • Web and internet traffic from any location or network
  • Sanctioned and unsanctioned SaaS usage
  • Access to private applications without inbound exposure
  • Data moving through the channels the service can observe
  • Policy consistency across those services and locations

What it does not solve alone

  • Wide-area connectivity, circuits, path selection and branch failover
  • Traffic and protocols the service cannot see or is not permitted to decrypt
  • East-west traffic between systems inside a site or a cloud network
  • Insecure application code and flawed business logic
  • Identity lifecycle and privileged access governance
  • Endpoint prevention and response on the device itself
  • Data already copied outside the observed channels

Frequently confused

Adjacent terms, and how they relate

Terms adjacent to Security service edge (SSE): each term's primary job and its relationship to this one.
Term Primary job Relationship
Secure access service edge (SASE) Converged networking and cloud-delivered security SSE plus the wide-area networking half. The difference between the two terms is scope of purchase, not a difference in how security is enforced.
Software-defined WAN (SD-WAN) Application-aware wide-area connectivity The half SSE leaves out. Buying SSE alone is a deliberate choice to keep or separately replace the connectivity layer.
Zero trust network access (ZTNA) Brokered access to individual private applications One of the services inside SSE, and usually the one that displaces remote-access VPN.
Cloud access security broker (CASB) Visibility and control over SaaS usage Another constituent service. Its inline controls overlap a web gateway's, while its API-based controls reach data a proxy never sees.
Secure web gateway (SWG) Inspecting and filtering outbound web traffic The oldest service in the bundle, and the one whose usefulness depends most on how much TLS traffic you are actually permitted to decrypt.
Firewall as a service Cloud-delivered network filtering beyond web ports Included in some SSE offerings and not others, which is one of the concrete places where two products sharing the label differ.

Before procurement

What to settle first

Deliberately unnumbered: these are the questions to answer, not a sequence to work through in order.

  • Decide the WAN question first, in writing

    Name who owns connectivity after the SSE purchase, what happens at a branch when a circuit degrades, and whether a later SD-WAN or SASE step is planned or explicitly ruled out.

  • Check the bundle against your own traffic

    Compare the offering's coverage of non-web protocols, unmanaged devices, server-initiated flows and your specific SaaS platforms with what you actually run, not with the category description.

  • Establish the inspection limits before migration

    Confirm how TLS decryption, certificate pinning, privacy constraints and bypass lists behave, and record every bypass as an uninspected path with a named owner and a review date.

  • Unify the policy, not only the vendor

    Agree the authoritative identity source, one definition of device trust, and one exception process across all the services, or the consolidation buys you a shared invoice and nothing else.

  • Migrate one service at a time with rollback

    Move web filtering, SaaS control, private access and data inspection in separate stages, each with a defined rollback, because a single cutover makes every failure ambiguous.

  • Prove the logs are usable as evidence

    Check event detail, retention, export, clock accuracy and whether records from the four services correlate on identity and session before you need them in an investigation.

Sources