What Is SASE? Secure Access Service Edge Explained
What is SASE? SASE (Secure Access Service Edge, pronounced "sassy") is an architecture that delivers networking and network security together as a cloud service, so that every user, branch office, device and application connects through the same set of security controls no matter where they are. Instead of routing traffic back to a corporate data center to be inspected, SASE inspects it at a nearby cloud point of presence and applies one identity-based policy. This guide explains what SASE is, which technologies it combines, how it differs from VPNs, SD-WAN and SSE, and how to evaluate and roll it out without a rip-and-replace.
- What Is SASE? A Plain-English Definition
- Why SASE Exists: The Problem With Hub-and-Spoke
- The Core Components of SASE
- How SASE Works: Architecture and Request Flow
- SASE vs SSE vs SD-WAN vs VPN vs Zero Trust
- Benefits and Trade-Offs
- Deployment Models: Single-Vendor, Dual-Vendor, Managed
- Common SASE Use Cases
- How to Evaluate a SASE Provider
- Migration Roadmap & Readiness Checklist
- Frequently Asked Questions
SASE stands for Secure Access Service Edge. Gartner introduced the term in 2019 to describe a shift it saw in how enterprises connect and protect people and applications: networking and security functions that used to live in separate boxes in a data center were converging into a single cloud-delivered service located close to the user.
The idea is easiest to see through an analogy. A traditional network is like a single airport security checkpoint at the city center: everyone, wherever they live, must travel to it to be screened, and only then continue to their destination. SASE places screening at many local airports, applies the same rules everywhere, and lets travelers go directly to where they are headed once they have been checked.
For roughly two decades, enterprise networks followed a hub-and-spoke design. Branch offices connected to a central data center over private links such as MPLS, remote workers connected over a VPN, and all traffic passed through the data center's firewalls, proxies and filtering tools before reaching the internet or any application. This made sense when applications lived in the data center and employees sat in offices.
Three shifts broke the model. Applications moved to SaaS and public cloud, so the data center stopped being the destination. People started working from anywhere, so the office stopped being the place users connect from. And attackers shifted toward stolen credentials and legitimate-looking sessions, so trusting anything just because it came from "inside" the network became a liability.
Firewall · proxy · VPN concentrator · DLP
One identity-based policy · inspect once, enforce everywhere
SASE bundles several technologies that were previously bought and run separately. Gartner's core set is SD-WAN, secure web gateway, CASB, zero trust network access and firewall as a service. Most platforms add data loss prevention, remote browser isolation and digital experience monitoring.
| Component | Full Name | What It Does | Side |
|---|---|---|---|
| SD-WAN | Software-Defined Wide Area Network | Connects branches and sites over any mix of links (broadband, 5G, MPLS) and steers each application over the best path based on policy and performance | Networking |
| SWG | Secure Web Gateway | Filters and inspects web traffic: blocks malicious sites, enforces acceptable use, scans downloads, and can inspect encrypted (TLS) traffic | Security |
| CASB | Cloud Access Security Broker | Gives visibility and control over SaaS and cloud service use: discovers shadow IT, enforces data policies, and detects risky configurations or behavior | Security |
| ZTNA | Zero Trust Network Access | Grants a verified user and device access to a specific private application, not to the whole network; replaces broad VPN access | Security |
| FWaaS | Firewall as a Service | Delivers firewall functions (port and protocol control, intrusion prevention, application awareness) from the cloud for all locations and users | Security |
| DLP | Data Loss Prevention | Detects and blocks sensitive data leaving through web, SaaS, email or uploads, using one policy across channels | Common add-on |
| RBI | Remote Browser Isolation | Runs risky or uncategorized web content in an isolated environment and sends only a safe rendering to the user | Common add-on |
| DEM | Digital Experience Monitoring | Measures end-to-end application performance from the user's device to help separate network, security and app problems | Common add-on |
Zero trust network access is the component that most visibly changes the user experience and the risk profile, so it is worth separating from the rest.
| Aspect | Traditional VPN | ZTNA (in SASE) |
|---|---|---|
| Access granted | Network-level: the user lands on a network segment | Application-level: only the specific app the policy allows |
| Trust model | Authenticate once, then broad implicit trust | Continuous verification of user, device and context |
| Lateral movement | Easier if the account or device is compromised | Reduced, because other apps are not reachable or visible |
| Where inspection happens | At a central concentrator, if at all | At the nearest cloud PoP, on every session |
| Exposure | Public VPN gateways that attackers scan for | Apps connect outbound to the service, with no inbound listener to attack |
Every SASE design has three layers: the edges that generate traffic, the cloud fabric that inspects and routes it, and the destinations users are trying to reach. A single policy engine sits across the fabric, making decisions from identity and context.
Path selection
Optimized backbone
Consider an employee on hotel Wi-Fi opening an internal finance application. In a SASE design, several things happen in milliseconds:
| Step | What Happens | Component |
|---|---|---|
| 1 | The device's agent (or the branch's SD-WAN appliance) sends the traffic to the nearest SASE point of presence | SD-WAN / client agent |
| 2 | The user is authenticated through the identity provider, with multi-factor authentication | Identity integration |
| 3 | The device's posture is checked: managed, disk encrypted, OS patched, endpoint protection running | ZTNA / posture check |
| 4 | Policy decides whether this user, on this device, from this location, may reach this specific app; if yes, a connection to that app only is brokered | ZTNA + policy engine |
| 5 | The session's traffic is inspected for threats and sensitive data, and the session is logged for audit | FWaaS, SWG, DLP |
SASE is surrounded by overlapping terms, which causes much of the confusion in buying discussions. They sit at different levels: some are technologies, some are architectures, and one is a security philosophy.
| Term | What It Is | Relationship to SASE |
|---|---|---|
| SASE | Converged networking and security delivered from the cloud | The full architecture |
| SSE (Security Service Edge) | The security half of SASE: SWG, CASB and ZTNA, often with DLP and more | SASE minus the SD-WAN networking half. A common first step for organizations with an SD-WAN already in place. |
| SD-WAN | Software-defined connectivity and traffic steering between sites | The networking half of SASE |
| VPN | Encrypted tunnel that gives a remote user access to a private network | A legacy remote-access approach that ZTNA replaces |
| Zero trust | A security principle: never trust implicitly, verify explicitly, grant least privilege | A philosophy. SASE, and ZTNA in particular, is one way to implement it for access. |
| Next-gen firewall (NGFW) | Appliance-based traffic inspection with application awareness | FWaaS is the cloud-delivered equivalent; NGFWs often remain for data centers |
SASE is not automatically the right move for every organization. The benefits are real, but so are the costs and constraints, and a balanced assessment helps you decide how far and how fast to go.
Better performance for cloud apps: traffic exits near the user instead of detouring through a data center.
Reduced lateral movement: per-application access limits what a stolen credential can reach.
Operational simplification: fewer appliances to patch, fewer consoles, one place for logs.
Scalability: new sites and users are onboarded by configuration rather than hardware shipments.
Migration effort: policies, exceptions and legacy applications must be mapped and tested.
TLS inspection complexity: decrypting traffic needs certificates, exemptions and privacy review.
Organizational change: networking and security teams must share ownership of one platform.
Legacy and latency-sensitive apps: some protocols and on-premises workloads still need local controls.
One of the biggest decisions is whether to buy the whole stack from one provider or combine two. Gartner has described single-vendor SASE as converged networking and security delivered by one vendor, and treats multi-vendor integrations of unrelated point products separately.
| Model | Description | Best Fit | Watch For |
|---|---|---|---|
| Single-vendor SASE | SD-WAN and security service edge from one provider, one console, one policy model | Midsized enterprises, joint network and security teams, organizations without an entrenched SD-WAN | Uneven strength across components; lock-in; whether the platform is truly unified or acquired pieces joined by a portal |
| Dual-vendor SASE | SD-WAN from one vendor and SSE from another, with explicit integration | Larger enterprises with separate networking and security teams or existing SD-WAN investments | Integration depth; two support relationships; policy duplication |
| Managed SASE | A service provider or MSSP operates the platform for you | Lean IT teams or distributed organizations without 24/7 operations capacity | Change-request turnaround; visibility into provider actions; contract exit terms |
| DIY multi-vendor | Assembling separate point products into a SASE-like design | Rarely recommended as a starting point | Inconsistent policy, more integration work, and weaker visibility |
| Use Case | Problem Today | How SASE Helps |
|---|---|---|
| Replace the VPN | Broad network access, scanned-for gateways, poor remote experience | ZTNA gives per-application access with device checks and no exposed inbound listener |
| Branch and retail connectivity | Expensive private circuits and a firewall appliance at every site | SD-WAN over commodity links with cloud-delivered security at each site |
| Secure SaaS adoption | Shadow IT, unmanaged sharing, no data visibility | CASB discovery and inline controls; DLP policy for uploads and sharing |
| Mergers and acquisitions | Merging networks and security stacks takes months | Onboard the acquired company's users and sites to the same fabric without flattening networks |
| Third-party and contractor access | Contractors on VPN with more access than needed | Time-bound, per-app access from unmanaged devices, using browser isolation where needed |
| IoT and OT segments | Devices that cannot run agents and are hard to segment | Site-level enforcement and device identification at the edge |
| Controlling AI tool use | Staff pasting sensitive data into unapproved AI services | Web and SaaS policies that allow sanctioned AI tools, block others, and apply DLP to prompts and uploads. See AI Governance and Security Policy. |
Marketing pages make every platform sound unified. These criteria separate genuine convergence from a stack of products behind a common login.
| Criterion | Questions to Ask |
|---|---|
| Architecture | Is it one codebase and one policy engine, or acquired products behind a portal? Is traffic inspected once with all engines, or passed between separate services? |
| Global coverage | Where are the PoPs relative to your users and key applications? Does the provider own the compute, or lease it? What are the documented latency figures in your regions? |
| Security depth | What does TLS inspection cover and exempt? How are threat intelligence and malware detection tested? What DLP and CASB inline controls exist? |
| Identity and device integration | Which identity providers and endpoint tools integrate natively? Can policy use device posture and risk signals continuously? |
| Networking | How does SD-WAN handle voice, video and real-time apps? Does it support your existing circuits and data center connectivity? |
| Visibility and operations | Is there a single log and analytics view? Can logs stream to your SIEM? What is the change workflow and API coverage? |
| Resilience | What are the uptime commitments and the failure behavior (fail-open vs fail-closed) per service? How are outages communicated? |
| Commercials and exit | How is licensing counted (users, bandwidth, sites)? What features cost extra? What are data export and exit terms? |
Providers tend to come from one of two directions, and their strengths often reflect where they started. The lists below are examples to start research, not rankings or endorsements. Offerings and acquisitions change frequently, so verify current capabilities and recent analyst evaluations before shortlisting.
Most organizations adopt SASE in stages, starting with the problem that hurts most, rather than replacing everything at once. A typical sequence follows.
The following is a vendor-neutral sketch of the kind of rules a SASE policy engine evaluates. Syntax differs by product, but the logic of identity, device posture, application and data in a single place is the point.
# sase-policy-sketch.yaml (illustrative, not tied to any vendor) rules: - name: finance-app-access action: allow when: user_group: finance device: {managed: true, disk_encrypted: true, os_patched_within_days: 30} app: internal-finance # private app, reached via ZTNA mfa: phishing_resistant inspect: [threat_protection, dlp] log: full - name: contractor-limited-access action: allow when: user_group: contractors device: {managed: false} app: project-portal session: {isolation: browser_isolated, download: block, expires_hours: 8} - name: sanctioned-ai-tools action: allow when: {app_category: generative_ai, app: [company-approved-assistant]} inspect: [dlp] # block restricted data in prompts and uploads - name: unsanctioned-ai-tools action: block when: {app_category: generative_ai} message: "Use the approved assistant. See the AI acceptable-use policy." - name: default-private-apps action: deny # deny by default for anything not explicitly allowed when: {app_type: private}
- ☐You have an application inventory that distinguishes SaaS, private and legacy applications
- ☐Your identity provider supports strong MFA and can supply user and group context to the platform
- ☐Endpoint management can report device posture (encryption, patch level, protection status)
- ☐Networking and security teams share a project owner and success metrics
- ☐A TLS inspection plan covers certificates, exemptions (health, finance, pinned apps) and privacy review
- ☐You have mapped existing firewall and VPN rules and flagged ones to retire rather than migrate
- ☐Failure behavior (fail-open or fail-closed) and a break-glass access method are defined and tested
- ☐Logs flow to your SIEM, and alerting and incident procedures cover the new platform
- ☐Contract end dates for MPLS, firewalls and VPN are known, so retirement can be timed to avoid paying twice