Fortinet FortiDDoS: Mitigation Policy Design

In FortiDDoS service-protection planning, distributed denial-of-service defense is difficult because the control must distinguish malicious traffic from legitimate demand while the system is already under stress. FortiDDoS Service Protection Policies provide a framework for applying protection profiles to traffic associated with protected subnets and services. The engineering challenge is deciding what normal looks like, which protocols require specialized limits, when to move from detection to prevention, and how to avoid blocking real customers during an unusual but legitimate spike.

That makes a FortiDDoS policy an availability control with security consequences. It should be designed with the same rigor as network security fundamentals: documented scope, explicit assumptions, measurable baselines, change control, and tested recovery.

DDoS telemetry should feed Fortinet security operations rather than remain isolated at the mitigation device. Mitigation devices can identify traffic patterns and drops, but analysts still need upstream intelligence, application context, and business impact to decide whether the event is volumetric abuse, protocol exploitation, application-layer exhaustion, or an internal capacity problem.

Define the protected service before tuning thresholds

A Service Protection Policy should map to a business service or subnet whose normal traffic can be described. Mixing unrelated applications behind one policy makes baselining harder because a spike in one service changes the apparent normal for another.

Document expected protocols, ports, client populations, geographic patterns, peak periods, failover behavior, and legitimate burst scenarios. A public DNS service and an internal API may both use UDP or TCP, but their acceptable traffic shapes differ dramatically.

Start with accurate routing and protection-subnet configuration. A perfect profile applied to the wrong traffic path provides no protection, while a profile that unexpectedly covers shared infrastructure can cause collateral impact.

Map each protected subnet to an application owner and escalation contact. During an attack, security engineers need to know whether unusual traffic corresponds to a legitimate event, a failing dependency, or an actual hostile campaign. Ownership data should be available without searching separate asset systems.

Document bypass and maintenance conditions as well. Planned load tests can resemble attacks, and uncoordinated testing may cause the DDoS control to block the very traffic engineers are trying to measure.

Use protocol-specific profiles where the attack mechanics differ

FortiDDoS supports multiple Service Protection Policy profile types, including IP, ICMP, TCP, HTTP, SSL/TLS, NTP, DNS, DTLS, and QUIC. These exist because different protocols fail differently under abuse. TCP SYN pressure, DNS amplification, HTTP request floods, and TLS handshake exhaustion require different signals.

FortiDDoS-F documentation recommends associating all supported protection-profile types with each Service Protection Policy; omitting a profile outright can weaken mitigation. Configure the relevant protocol settings, thresholds, and detection or prevention mode for the traffic observed on the protected path. Review the deployed FortiDDoS-F version and maintain an owner for every profile and exception. Broader firewall capabilities still matter—stateful policy, application awareness, threat prevention—but a dedicated DDoS control focuses on rate, behavior, and service exhaustion patterns that may need protection before conventional security inspection can help.

Protocol profiles should be reviewed against the actual service stack. A TLS-terminating load balancer may absorb SSL/TLS behavior before traffic reaches the protected backend, while DNS services may use both UDP and TCP. Protect what the FortiDDoS appliance can actually observe on its path.

QUIC and newer protocols deserve explicit review because traffic characteristics can differ from traditional TCP-based applications. Enabling a profile without understanding adoption can create blind spots or unnecessary enforcement.

Baseline legitimate peaks, not daily averages

Attack detection depends on the difference between expected and abnormal behavior. Daily averages are often misleading because production traffic has sharp peaks during backups, software launches, financial deadlines, streaming events, marketing campaigns, or regional business hours.

Collect percentile distributions and peak windows for each important metric. Include failover tests and disaster-recovery events in the baseline because traffic may concentrate on fewer sites during an outage. Capacity and protection thresholds should survive those legitimate modes. DDoS attack signals can appear in different orders: rising connection attempts, resource exhaustion, protocol anomalies, and service latency may lead or lag one another depending on the attack.

Baselines should account for growth. A service that gains customers can outgrow thresholds that were safe six months earlier. Capacity planning and protection tuning should share traffic forecasts so legitimate success does not become self-inflicted denial of service.

Use clean historical windows when possible. If an organization learns thresholds during an active or recurring attack, the resulting baseline may normalize malicious traffic and make later detection less sensitive.

Separate detection tuning from prevention tuning

A profile that detects unusual activity does not automatically have safe prevention thresholds. Detection can afford to be more sensitive because an analyst can interpret context. Prevention must consider the cost of a false positive on real users.

Introduce blocking or rate enforcement progressively. Observe detection-only behavior, compare alerts to application metrics, then enable prevention on the patterns with high confidence. Keep rollback steps and known-good thresholds ready during the first production changes.

Shared profiles require special care because changes can affect every Service Protection Policy that references them. Before editing a widely assigned profile, enumerate all dependent services and decide whether a clone is safer for testing.

Test changes with controlled traffic where feasible. Generate representative connection rates and protocol behavior, observe detection and drop counters, and confirm the application remains healthy. Configuration review alone cannot reveal every interaction with real traffic.

Rollback should be specific. Know which profile or threshold to restore, who can make the change, and how to verify recovery. ‘Disable DDoS protection’ is not a safe emergency plan.

Maintain a record of mitigation debt: services protected by temporary thresholds, manual exceptions, or profiles that nobody has reviewed recently. Those conditions can become dangerous during the next traffic surge because operators will not know which settings reflect current business demand.

After every incident, review temporary threshold changes and restore the approved baseline deliberately; confirm from service metrics that the rollback returned both protection and legitimate availability to the intended state.

Application-layer attacks require business context

HTTP and TLS floods can resemble legitimate high demand. Bot traffic may complete handshakes and request real content, making pure packet rate less informative. Coordinate DDoS policy with application owners who understand expensive URLs, session behavior, authentication, cacheability, and normal client retry patterns.

Protect the scarce resource rather than an arbitrary packet count. If one API call triggers a costly database operation, request rate and concurrency may matter more than bandwidth. If TLS negotiation is the bottleneck, handshake characteristics matter. If DNS recursion is abused, query type and response amplification become relevant.

Application observability should be available during mitigation so the security team can see whether drops are improving service health or merely shifting the bottleneck.

Application teams can help identify requests that are cheap for the client but expensive for the server, such as search, export, authentication, image transformation, or cache-bypass paths. These economic asymmetries are common targets for application-layer exhaustion.

Pair mitigation data with origin and CDN metrics. If the edge is dropping traffic but the backend remains saturated, retries, alternate paths, or unprotected endpoints may be bypassing the intended control.

Runbooks should include upstream and downstream coordination

Some attacks exceed local circuit capacity before on-premises or inline controls can process them. The runbook should identify ISP contacts, upstream scrubbing options, cloud protections, BGP or routing procedures, and authority for emergency changes.

Downstream systems also need coordination. Load balancers, firewalls, application gateways, and servers may interpret the mitigated traffic differently. A change that reduces one layer’s load can still produce connection churn or retry storms elsewhere.

Use the broader Fortinet environment for consistent telemetry and response where appropriate, but keep ownership explicit. DDoS containment may involve network, security, application, cloud, and provider teams simultaneously.

Pre-negotiate communication templates for major incidents. Network providers, application owners, executives, and customers need different levels of technical detail. Clear status language reduces pressure for improvised changes while engineers are tuning mitigation under load.

Exercise upstream escalation at least annually. Contact lists and scrubbing procedures that have never been tested are likely to fail when routes, contracts, or provider portals have changed. Record the exact evidence the provider requires before it will activate mitigation.

Measure service survival, not just dropped packets

Packet drops are a means, not the outcome. Measure application availability, latency, error rate, protected bandwidth, connection success, mitigation time, false-positive complaints, and the difference between detected and blocked traffic. These tell whether the policy kept the service usable.

After every event, compare the attack characteristics with the configured profiles. Update baselines only when the traffic represents a durable business change; otherwise the attack itself can distort future expectations.

Well-designed FortiDDoS policies make the organization more predictable under stress. The goal is to enter an attack with known thresholds, known rollback points, known escalation paths, and enough evidence to protect service availability without guessing in real time.

Post-incident reviews should compare attack source distribution, protocol mix, duration, peak rate, mitigation actions, false positives, and business impact. Capture what analysts had to discover manually during the event and add that information to the runbook.

Regular exercises should include both attack and configuration-failure scenarios. A service can become unavailable because an overly aggressive protection change blocks users even when no attacker is present.

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!