Tuning an IPS Policy: Weeding Out False Positives and Moving to Blocking

How do you move an IPS policy from monitoring mode to blocking? Narrowing the signature profile, narrow exceptions for false positives and verifying that the IPS is working.
Technical summary
Technology
IPS, IDS, NGFW, Suricata, signature profile, suricata-update
Topic
A controlled move of the IPS from monitoring mode to blocking mode
Level
Intermediate–advanced
Prerequisites
  • Network zones and a critical asset inventory
  • A log manager or SIEM where IPS logs are collected
  • A change and maintenance window process
In this article8 sections

On most firewalls with an IPS license enabled, the picture is the same: either the profile has sat in monitoring-only mode for years, or the default profile was attached to all traffic and, after the first wrongful block, effectively switched off with broad exceptions. The license is paid for, but there’s no protection. Below we explain how to tune an IPS policy in a live environment: from narrowing signature coverage to exception management and a phased move to blocking mode.

IDS and IPS: placement and operating mode differences

An IDS inspects a copy of traffic (SPAN, TAP or a virtual mirror port) and only raises alarms; if the IDS fails, traffic isn’t affected. An IPS, on the other hand, sits in the traffic path (inline): it can drop packets, reset sessions and temporarily block a source. This difference has two practical consequences. First, every IPS false positive is an outage. Second, the IPS can be a bottleneck and a point of failure; whether traffic is allowed through (fail-open) or cut (fail-closed) in a hardware failure should be a conscious decision. Making that decision during the failure is too late.

Feature IDS IPS
You are here Mirror/TAP, out of the traffic path Inline, in the traffic path
Action Alarm Alarm, packet drop, session reset
Impact of a false positive Analyst workload Outage
Latency impact None Yes, depends on capacity
Typical use Visibility, east-west monitoring North-south and inter-zone protection

On NGFWs, IPS usually works as a profile attached to a security policy. On the same device you can monitor in one policy and block in another. The phased migration plan relies on this flexibility.

Choosing the signature profile: coverage by asset and service

The number one source of false positives and performance loss is attaching a single “all signatures” profile to every policy. Running signatures for Windows client vulnerabilities on traffic to a web server running on Linux both burns CPU and produces meaningless alarms. To narrow coverage, split the profile along these axes:

  • Direction and target: Server-side signatures for traffic coming to servers published to the internet; client-side signatures for users’ internet egress.
  • Operating system and application: The platforms the assets in the DMZ actually use (for example Linux, nginx, a particular database).
  • Severity: Initially, high and critical signatures are candidates for blocking; low and informational ones can stay in monitoring mode.
  • Zone: Traffic passing from the user network to the server network needs a separate profile to see lateral movement between segments. The zone design itself is the subject of network security and segmentation work.

In practice, three to five profiles are enough for most organizations: published servers, internet egress, between internal zones and, if needed, OT or the guest network. As the number of profiles increases, so does the operational load. Any extras go unmaintained.

Collecting a baseline in monitoring mode

Before moving to blocking, run each profile in alarm-only mode for at least two to four weeks. The goal isn’t to see “zero alarms” but to understand what the alarms look like. Information to collect during this period:

  • Alarm count per signature and the top 20 most triggered signatures
  • Source and destination distribution for each signature: does it come from a single system, or spread across the network?
  • The time pattern of alarms: do they coincide with backup, scanning or batch job hours?
  • Device resource usage: CPU, session count and latency values

Review alarms not one by one but grouped in the log manager by signature ID, source and destination. Turning this data into detection rules in a SIEM is the subject of the firewall log analysis guide.

False positive analysis and exception management

Common sources of false positives are: authorized vulnerability scanners, protocol queries from monitoring systems, old or not fully standard-compliant application protocols, large binary data streams in backup and replication traffic, and signatures matching with incomplete context because encrypted traffic isn’t decrypted.

Before closing an alarm as a false positive, answer four questions: Which vulnerability or behavior does the signature target? Is the target system running software vulnerable to it? Is it expected for the source system to produce this traffic? Does a packet capture show what the signature actually matched? An alarm closed without answers stays closed at the next real attack too.

If an exception is needed, write it in its narrowest form. The order is: a specific signature + a specific source/destination, then a specific signature + a specific subnet, and only last, disabling the signature entirely. Disabling the whole profile for a server isn’t an exception. In open-source Suricata, this approach is managed with suppress lines in the threshold.config file, and signature-level changes with the disable.conf, drop.conf and modify.conf files of the suricata-update tool:

# /etc/suricata/threshold.config
# Example SID: suppress matches only from the authorized scanner
suppress gen_id 1, sig_id 9000001, track by_src, ip 10.20.30.15

# /etc/suricata/drop.conf  (after the baseline; alert -> drop via suricata-update)
group:emerging-exploit.rules

# Update the rules, test the configuration
sudo suricata-update
sudo suricata -T -c /etc/suricata/suricata.yaml

On commercial NGFWs, the same logic is applied through per-signature action changes and IP exception (exempt) fields. Whatever the platform, every exception should live with a record:

Field Example content
Signature ID and name Vendor signature ID and description
Scope Source 10.20.30.15, destination DMZ web pool
Justification and evidence Authorized scanner; packet capture attached
Approver and owner Security team / vulnerability management owner
Review date In three months; recheck on signature updates

A phased plan for moving to blocking mode

The migration order is set to balance risk and visibility:

  1. Internet-published services, critical and high signatures. The area with the highest likelihood of attack and the most predictable traffic profile.
  2. Internet egress, client-side critical signatures. Exploit kit, known malware communication and command-and-control signatures come at this stage.
  3. Traffic between internal zones. The area with the highest false positive risk; it’s started last, with the narrowest signature set.
  4. Medium severity. Added selectively based on baseline data after the first three stages have settled.

At each stage, the change is made in a maintenance window, blocking logs are watched closely for the first 48 hours, and the rollback step (switching the profile to monitoring mode) is written in advance. Give the support team a clear instruction too: when a “the application isn’t working” complaint comes in, the first place to look is the IPS blocking log. Without this information, the team spends half a day searching on the server.

Performance impact and hardware capacity

On vendor datasheets, firewall throughput, IPS throughput and “threat protection” throughput are separate values; with IPS on, real capacity drops noticeably below firewall throughput. These values are measured with a traffic mix chosen by the vendor and may not represent your traffic. When evaluating capacity, look at:

  • Profile coverage: A narrowed signature set means higher real capacity on the same hardware.
  • TLS decryption: When enabled together with IPS, resource consumption rises significantly.
  • Peak hour behavior: Not the average but CPU and session values during the busiest period are what count.
  • Fail-open/fail-closed: The IPS engine’s behavior on a software or hardware failure is verified in a test environment.

How to read these values at the device selection stage is covered in the enterprise firewall selection article.

Verifying that the IPS really works

“IPS on” showing in the dashboard doesn’t mean you’re protected from an enterprise cybersecurity perspective. For regular verification:

  • Harmless trigger test: Like the quick-start test in the Suricata documentation, a harmless request that triggers a known signature (curl http://testmynids.org/uid/index.html) should give the expected result in monitoring and blocking mode. This request triggers a signature in the Suricata/ET rule sets; it may not give the same result on commercial NGFWs. In that case, use the vendor’s recommended test method or a harmless test file with a known signature (for example EICAR).
  • Update check: The signature database version and last update time should be connected to the monitoring system; updates stopping should raise an alarm. This is the common silent failure: the license expires, signatures aren’t updated for months, and nobody notices.
  • Coverage check: It should be reported periodically that every security policy is bound to the correct IPS profile; forgetting the profile on newly added rules is common.
  • Independent testing: An authorized penetration test or attack simulation measures the IPS’s response to real attack techniques. Before the test, decide consciously whether the test sources will be added to the exception list.

Handling the IPS policy from profile design to operation can be planned as part of firewall installation and rule management work. For Suricata rule management details, the Suricata User Guide is the primary source.

Frequently asked questions

Because you can’t know in advance which signatures will produce false positives in your traffic. Blocking turned on before a baseline is collected cuts off legitimate applications and pushes the team to effectively switch off the IPS with broad exceptions.

How long should it stay in monitoring mode?

Two to four weeks per profile is a common start; month-end jobs, backup windows and periodic scans should fall within this period. The real criterion isn’t the duration but having explained each of the most frequently triggered signatures.

What’s the difference between disabling a signature entirely and defining an exception?

Disabling a signature removes that detection for the whole network. An exception suppresses the signature only for a specific source or destination, and detection continues for other traffic. The narrowest exception is always the preference.

Can an IPS see attacks in encrypted traffic?

Without TLS decryption it sees only connection metadata and some handshake information; it can’t see attack content at the application layer. For published servers, IPS and TLS decryption should be planned together.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Cybersecurity

Firewall, VPN, endpoint security and threat prevention. We reduce the attack surface and measure and report the risks.

Knowledge Center

Other articles