Fortinet FortiSASE: Secure Internet Access

In FortiSASE deployments, Secure Internet Access is the part of Fortinet’s cloud-delivered security model that protects user traffic headed toward the public internet. Current Fortinet architecture guidance separates Secure Internet Access (SIA), Secure Private Access (SPA), and Secure SaaS Access (SSA), which is a useful boundary because each path solves a different connectivity problem. SIA is about steering internet-bound traffic through FortiSASE points of presence so security controls can inspect and enforce policy before traffic reaches external services. That makes SIA both an access-control layer and a telemetry source for Fortinet security operations.

The design should not begin with a list of security profiles. It should begin with the traffic path. Every protected user must have a predictable way to reach FortiSASE, and the organization must understand what happens when that steering method, identity signal, endpoint agent, branch tunnel, or point of presence is unavailable. A SASE architecture succeeds when networking and security are operated as one service path rather than as separate appliances connected by assumptions.

Choose the steering method before tuning security policy

Fortinet documents several SIA steering patterns, including endpoint-agent steering and IPsec-based connectivity from sites. Some designs can also use agentless or explicit-proxy behavior for web traffic; Fortinet documents those proxy SIA steering paths as web-only (HTTP and HTTPS), not equivalent to all-protocol endpoint-agent or routed IPsec steering. These options are not interchangeable. Endpoint steering can follow a mobile user across networks, while site tunnels are well suited to branch locations where many devices share an egress path. Explicit proxy models can be efficient for selected web applications but do not automatically represent every protocol a user may generate.

The first design question is therefore scope: which traffic must be protected, from which device populations, and under which mobility conditions? A branch with managed Windows laptops may tolerate a different steering model than a warehouse with scanners and embedded devices. If different methods coexist, operations must be able to identify which steering path a failing user is on. Otherwise a DNS issue, agent enrollment problem, tunnel failure, and policy block can all look like “FortiSASE is down” even though they occur in different layers.

This is one of the practical consequences of what SASE changes: the service edge becomes part of the forwarding architecture. Routing, DNS, identity, device posture, inspection, and egress selection have to be tested together.

Steering scope also needs an explicit bypass model. Some destinations may require direct access because of local printers, emergency services, captive portals, software bootstrap traffic, or technical constraints. Those exceptions should be narrow and observable. A broad “do not steer when this process runs” rule can create a shadow internet path outside inspection. During pilot testing, capture which flows bypass FortiSASE and verify that the reason remains valid after application updates. If the endpoint agent makes steering decisions locally, configuration drift and agent health become part of the security boundary; if a branch tunnel provides steering, route ownership and tunnel failover become the equivalent dependency.

Inspection policy should reflect the destination and the user context

SIA can apply controls such as intrusion prevention, application control, web and DNS filtering, antimalware inspection, sandboxing, and botnet or command-and-control protection. Stacking every feature at its strictest setting is not automatically safer. It can create brittle behavior, excess false positives, or performance costs that cause teams to introduce broad bypasses later. A stronger model defines which controls are mandatory for categories of traffic and where exceptions require explicit ownership.

TLS inspection deserves separate planning because it changes the trust relationship between endpoint and destination. Certificate deployment, applications that use pinning, privacy-sensitive categories, and regulated traffic can all affect whether decryption is technically possible or organizationally appropriate. The policy should specify where decryption is required, where it is prohibited, and what compensating telemetry remains available when content cannot be inspected.

Application and web controls also need change discipline. A SaaS application can shift domains, APIs, and content-delivery dependencies without the organization changing its own policy. Teams should test important business applications with realistic user workflows, not only with a browser opening the home page.

Policy evaluation should be tested with both expected and adversarial traffic. It is not enough to confirm that approved business SaaS works. Validate malware categories, newly registered domains, anonymizers, risky file downloads, command-and-control indicators, and applications that tunnel over common ports. Security profiles should produce evidence that explains why a session was blocked or allowed. This matters during incident response because analysts need to distinguish a policy decision from a connectivity defect. If the verdict cannot be reconstructed from logs, operators will often create an exception simply to restore service, weakening the control without understanding the original event.

Identity must survive the trip through the cloud edge

Internet access decisions become more useful when policy can distinguish a named user, group, device class, or risk state rather than only an IP address. That requires reliable identity integration and a clear understanding of what FortiSASE knows at the moment a connection is evaluated. Shared devices, stale sessions, emergency accounts, contractors, and users who change group membership can expose hidden assumptions in otherwise clean policy designs.

Operations should verify identity in both allowed and denied cases. If an analyst cannot tell which identity FortiSASE associated with a transaction, troubleshooting turns into guesswork. The same is true for device posture. A “noncompliant” outcome is only actionable when the team can see which posture condition failed, when it was last assessed, and how the user can recover.

Identity freshness is another production concern. A user can change groups, lose employment status, move from a corporate device to an unmanaged device, or keep a long-lived browser session while the underlying risk has changed. The access architecture should define when identity and posture are reevaluated and how quickly changes reach policy. Short reevaluation intervals increase responsiveness but may create more dependencies on identity services. Long intervals reduce load but increase the time stale authorization remains valid. The correct choice depends on the sensitivity of the application and the organization’s ability to operate the identity path reliably.

Branch tunnels need routing and failure-domain design

When a branch reaches FortiSASE through IPsec, the tunnel is only one component of the forwarding decision. Route advertisement or static routing, local breakout rules, DNS reachability, MTU behavior, NAT, and asymmetric paths all influence whether traffic actually traverses the intended edge. A tunnel can be administratively up while application flows still bypass it or return through a different path.

Redundancy should be tested as a traffic event, not only as a control-plane state. Teams need to know how quickly new sessions move to a surviving path, whether existing sessions are reset, whether DNS answers change, and whether the secondary path receives the same policy. A failover that preserves connectivity but loses identity or inspection is not an equivalent service.

Branches also need a clear local-survivability decision. If FortiSASE or WAN connectivity is unavailable, should internet access fail closed, fail open through a local circuit, or preserve only a restricted emergency path? The answer should differ by site function. A call center handling sensitive data may accept a hard outage to preserve inspection, while a retail location may need limited payment or operational traffic to continue. Whatever choice is made, the failover mode must be visible to operations. Silent fallback to uninspected internet access can keep the business online while creating a security condition no one realizes exists.

Egress design affects application trust and allowlisting

Many business services still use source IP as an allowlisting signal. Moving internet access into a cloud service changes where those sessions appear to originate. The team must inventory applications that depend on fixed egress addresses, document the required FortiSASE egress behavior, and coordinate changes before migration. This matters for SaaS administration portals, partner APIs, payment systems, and legacy services that were built around office public IP ranges.

Egress also affects geolocation and data-residency expectations. A user may physically sit in one country while the security service sends traffic through a different regional point of presence. Organizations with location-sensitive applications should validate the actual path and not assume nearest-PoP behavior always produces the required business outcome.

Troubleshooting should follow the packet path

A useful SIA runbook starts with steering: did the endpoint or branch send this flow toward FortiSASE? Next confirm tunnel or agent health, identity, DNS behavior, policy match, security-profile verdict, and egress. Only after those layers are known should the team escalate to a generic service issue. This sequence prevents a policy denial from being investigated like a routing outage.

Logging must capture enough evidence to explain the decision. Analysts should be able to associate user, device, destination, application, rule, security action, and time. When a user reports intermittent failure, compare successful and failed sessions rather than inspecting only the failure in isolation. Differences in identity, destination IP, application classification, or point of presence often reveal the actual boundary.

Capacity planning should include the security features that actually run, not just circuit bandwidth. TLS decryption, malware scanning, sandbox submission, and complex application control can change processing demand and latency. Test large downloads, interactive SaaS, video, and high-connection-rate applications during representative load. Also observe user experience when traffic is sent to a secondary point of presence. An architecture can meet bandwidth targets while still producing unacceptable session setup delay or jitter. SIA should therefore have user-experience objectives alongside security objectives, with clear thresholds for latency, loss, and steering failures.

Operate SIA as a production connectivity service

Mature SIA operations track steering success, tunnel availability, agent health, authentication failures, policy-denial trends, malware or intrusion events, latency by region, and egress changes. Those measures should be reviewed by networking and security together because neither team owns the whole user experience. A rise in inspection blocks may be a security success, a false-positive problem, or the first signal of compromised endpoints.

FortiSASE SIA is most reliable when it is treated as a routed security service with explicit failure domains. The surrounding Fortinet ecosystem can supply broader security and operations tooling, but the internet-access path still has to stand on its own: steering must be deterministic, identity must be trustworthy, inspection must be intentional, and failover must preserve the security properties users rely on.

Leave a Reply

How It Works

img
Step 1. Choose Exam
on ExamLabs
Download IT Exams Questions & Answers
img
Step 2. Open Exam with
Avanset Exam Simulator
Press here to download VCE Exam Simulator that simulates real exam environment
img
Step 3. Study
& Pass
IT Exams Anywhere, Anytime!