UTMStack

Articles13 min read

Suricata SIEM Integration: Send EVE JSON Logs and Tune Alerts

By UTMStack Team

UTMStack blog: Suricata SIEM integration

Suricata watches network traffic and raises an alert when a packet matches a rule. On its own, it writes those alerts to a file on the sensor, and that is where many of them stay. Nobody reads a log file at three in the morning. The value comes when the alerts reach a SIEM (security information and event management system), where they sit next to firewall, server and login records and someone is told about the ones that matter. This guide shows how to connect the two, what to send, and how to keep the volume under control.

The short answer: Suricata writes everything to one JSON (JavaScript Object Notation) log called EVE, by default /var/log/suricata/eve.json. To send it to a SIEM, either set the eve-log output to filetype: syslog and forward that syslog facility with rsyslog, or leave the file in place and have a log shipper read it. Forward alert records first, then add dns, http, tls and flow if you have the storage. Raise rsyslog’s message size limit so long records are not cut. In Suricata, severity 1 is the most serious, not the least.

Is Suricata a SIEM?

No. Suricata is a network intrusion detection system (IDS). It can also block traffic when it runs inline as an intrusion prevention system (IPS). It inspects packets, matches them against rules and records what it saw. It does not collect logs from other products, keep them for years, or link a network alert to a failed login on a server. A SIEM does those jobs. The two work together: Suricata is one of the best sources a SIEM can have, because it sees traffic that servers and endpoints never log.

Suricata is open source under the GNU General Public License version 2. The current release line is 8.0, first published in July 2025. The 7.0 line reached the end of its life on July 7, 2026, so sensors still on 7.0 no longer get fixes. Our guide to intrusion detection systems covers how network and host sensors differ.

What Suricata writes: the EVE log

EVE is Suricata’s main output. Every line is one JSON record with an event_type field that says what kind of record it is.

Event typeWhat it recordsVolumeSend to the SIEM?
alertA packet or flow matched a ruleLow to mediumAlways
anomalyTraffic that breaks protocol rulesLowYes
dnsEach query and answerHighYes if storage allows; it is the best record of what a host looked up
httpEach web request with host, address and user agentHighYes if storage allows
tlsEach encrypted session with server name and certificate detailsHighYes if storage allows
fileinfoFiles seen in traffic, with their hashesMediumUseful for malware cases
ssh, smb, rdp, dhcp and other protocolsOne record per session or transactionVariesPick the ones that match your risks
flowA summary of each connection when it endsVery highOnly with a plan for the volume
statsCounters about the sensor itselfLowNo; it is health data, not security data

The volume column is a rule of thumb from how the records are produced, not a measurement. An alert is written only when a rule matches. A flow record is written for every connection on the network. Suricata’s own documentation warns that turning on every log lowers performance and uses more disk space.

You choose what is written in the types list under eve-log in suricata.yaml. That list is the first and cheapest filter you have.

Anatomy of a Suricata alert record

An illustration with made-up values, using the field names from Suricata's EVE format

"timestamp": "2026-11-15T03:12:44.120+0000"1 "flow_id": 18856290134972132 "event_type": "alert"3 "src_ip": "203.0.113.45", "src_port": 518224 "dest_ip": "10.20.4.17", "dest_port": 4455 "proto": "TCP", "app_proto": "smb"6 "alert": {"action": "allowed", "signature_id": 2000001, "signature": "Example SMB exploit attempt", "category": "Attempted Administrator Privilege Gain", "severity": 1}7
  1. When the sensor saw it. Keep the sensor's clock in sync.
  2. Ties this alert to the dns, http, tls and flow records of the same connection.
  3. The record type. Filter on this first.
  4. Where the traffic came from.
  5. Where it was going.
  6. Transport and application protocol.
  7. The rule that matched, whether the packet was allowed or blocked, and the severity. 1 is the most serious.
Real records carry more fields, such as rule metadata and details of the web request or file involved.

Two fields deserve attention. The flow_id lets you jump from an alert to everything else Suricata recorded about the same connection, which is often what turns a vague alert into a clear answer. And alert.action says allowed unless Suricata runs inline and a drop or reject rule matched. In plain detection mode, every alert reads allowed, because the sensor only watches. That does not mean the attack worked.

Two ways to send EVE logs to a SIEM

From sensor to SIEM

Both routes deliver the same JSON records; they differ in what can go wrong on the way

Suricata sensor
Route 1EVE to syslogSuricata hands each record to the local syslog service, which forwards it
Route 2EVE file plus a shipperSuricata writes eve.json; rsyslog, Fluent Bit or Vector reads the file and sends each line
SIEMParses the JSON, checks detection rules, links alerts with other log sources, keeps the records
UDP is the User Datagram Protocol, TCP the Transmission Control Protocol and TLS is Transport Layer Security. Route 1 is quicker to set up. Route 2 keeps a local copy and copes better with long records and outages.

Route 1: EVE over syslog

1. Tell Suricata to write EVE to syslog. In suricata.yaml:

outputs:
  - eve-log:
      enabled: yes
      filetype: syslog
      identity: "suricata"
      facility: local5
      level: Info
      types:
        - alert
        - anomaly
        - dns
        - http
        - tls

Then restart Suricata with sudo systemctl restart suricata.

2. Forward that facility. Add a file such as /etc/rsyslog.d/60-suricata.conf:

local5.*  @@siem.example.com:514

Two @ signs mean TCP. One means UDP. Restart rsyslog afterwards.

3. Raise the message size limit. This is the step people miss. rsyslog cuts any message longer than its maximum size, which is 8K by default. An alert that carries web request details and rule metadata can be longer than that. A JSON record with its end cut off cannot be parsed, so the SIEM stores a broken line or drops it. Set a higher limit at the top of /etc/rsyslog.conf, on the sensor and on any relay in between:

global(maxMessageSize="64k")

Prefer TCP over UDP for the same reason. The original syslog standard, RFC 3164, limits a packet to 1,024 bytes, and some receivers still enforce small limits on UDP.

Route 2: read the file with a log shipper

Leave filetype: regular so Suricata keeps writing eve.json. Then point a shipper at the file. The rsyslog file input module, Fluent Bit’s tail input and Vector’s file source all read a file line by line and send each line on. This route has three advantages. The file is a local buffer, so a short network outage loses nothing. Long records are less likely to be cut. And you keep a copy on the sensor for packet-level investigation.

The price is one more moving part, and a file that needs rotation. If you send flow records, check the disk on the sensor.

Check that it works

Suricata’s quick start guide gives a test. Run curl http://testmynids.org/uid/index.html from a monitored host. It triggers signature 2100498 from the Emerging Threats Open rules, and the alert should reach the SIEM within seconds. For an ongoing check, search the SIEM for event_type equal to alert and compare the count over an hour with grep -c '"event_type":"alert"' /var/log/suricata/eve.json on the sensor. If the SIEM has fewer, look for cut or dropped messages.

Suricata severity runs the other way

Each Suricata rule belongs to a class, and each class has a priority. That priority becomes alert.severity in the EVE record. Suricata’s documentation states it plainly: the highest priority is 1.

Suricata priorities, from most to least serious

Example classes from the classification.config file that ships with Suricata

Priority 1Act on theseA Network Trojan was detected. Attempted Administrator Privilege Gain. Potential Corporate Privacy Violation.
Priority 2ReviewAttempted Information Leak.
Priority 3ContextMisc activity. Detection of a Network Scan. Not Suspicious Traffic.
Priority 4BackgroundA TCP connection was detected.
The labels "act", "review", "context" and "background" are ours. Suricata only gives the numbers. A rule can also set its own priority.

Many SIEM products use the opposite scale, where a bigger number means a worse event. If the parser copies the number without turning it around, your “high severity” filter shows the least important alerts. After you connect a sensor, find one alert you know is serious, such as a trojan signature, and check which label the SIEM gave it.

Note one oddity in the default file. “Potential Corporate Privacy Violation” is priority 1. Policy rules of that class can fire often on normal traffic, which is one reason raw priority 1 counts look alarming on a new sensor.

Cut the noise before it reaches the SIEM

A new sensor with a full rule set is loud. Fix that at the sensor, where it is cheapest. Every alert you drop there is one the SIEM does not have to store, parse and show.

  • Choose your rule sources. sudo suricata-update fetches the Emerging Threats Open rule set by default, which is free. suricata-update list-sources shows others, and a paid Emerging Threats Pro set exists.
  • Disable rules that do not apply. List their signature IDs in /etc/suricata/disable.conf and run suricata-update again. Rules for software you do not run are the first candidates.
  • Suppress known-good sources. In threshold.config, a line such as suppress gen_id 1, sig_id 2002087, track by_src, ip 192.0.2.10 silences one rule for one address, for example your vulnerability scanner.
  • Limit repeats. The same file supports threshold and rate_filter entries that cap how often one rule alerts for one source.
  • Send fewer event types. Remove flow from the types list, or send it to cheaper storage, if nobody searches it.
  • Tell Suricata what is yours. Set HOME_NET correctly in suricata.yaml. Many rules only make sense when the sensor knows which side is inside.

What remains is work for the SIEM: grouping repeated alerts from one source, and raising priority when the target is an important server or when another log source agrees. Our guides to reducing SIEM false positives and SIEM correlation rules cover that side.

Detections worth building on Suricata data

Suricata alerts are a starting point. Combined with its protocol records and with other sources, they support stronger detections.

DetectionSuricata records usedWhy it works
Repeated priority 1 alerts from one sourcealertOne alert can be chance; several in minutes rarely are
Scan followed by a loginalert plus server login logsShows the scan found something
Queries to a brand-new or random-looking domain, many timesdnsCommon in tunneling and malware check-ins
Hacking tool or file-sync tool user agenthttpTools such as scanners and bulk copy utilities announce themselves
Internal host talking to many others on file sharing or remote desktop portsalert, flow, smb, rdpThe pattern of an intruder moving sideways
Unusual certificate on an outbound sessiontlsSelf-signed or mismatched names on traffic leaving the network
Known-bad address or domainAny record plus threat intelligenceCheap and precise when the feed is good

For the wider method, see our guide to network traffic analysis.

How UTMStack handles Suricata logs

We build UTMStack, so weigh our view accordingly. Here is how it works with Suricata, including what it does not do.

  • Collection. The UTMStack Collector is a small Linux service that runs on the UTMStack server or another Linux host in your network. It listens for Suricata on port 7019 over UDP or TCP, and over TLS once you load a certificate. The setup screen shows the eve-log block and the rsyslog line for your server, which is Route 1 above. See integrations.
  • Parsing. A dedicated parser removes the syslog header and reads the EVE JSON. It maps source and destination to common fields, adds the country of each address, and drops stats records so sensor counters do not fill your storage.
  • 35 built-in rules on Suricata data. Thirty-three cover network threats: port scans, DNS and ICMP (Internet Control Message Protocol) tunneling, Cobalt Strike traffic patterns, hacking tool and malware user agents, lookups of mining pools and Tor gateways, rclone data theft, movement between internal hosts, suspicious file downloads, certificate anomalies and more. Two raise an alert when one address triggers repeated Suricata alerts. Each maps to MITRE ATT&CK and is checked as each log arrives, before storage, alongside the rest of the 600+ detection rules.
  • Suricata inside pfSense. If Suricata runs as a pfSense package, its alerts arrive through the pfSense integration, which has its own rule for high-priority Snort and Suricata alerts.
  • Free to start. Syslog integrations, Suricata included, are part of the free Community edition, which installs on your own Linux server.

What it does not do:

  • It does not run Suricata for you. UTMStack receives and analyzes the records. Installing sensors, placing them on the network and managing rule sets stays with you.
  • It does not change Suricata’s verdicts. UTMStack response flows run through agents on Windows, Linux and macOS, where they can block an address on that host. They do not push drop rules into Suricata.
  • Check the severity label. The parser adds a severity label of low, medium or high from Suricata’s number. Confirm that it matches Suricata’s scale, where 1 is the most serious, on a known alert before you build dashboards or rely on the two rules that use the label. You can adjust the pipeline in the interface.
  • Several network rules depend on the text of Suricata’s own alerts. They work best with the Emerging Threats rules loaded. A few expect fields that a stock sensor may not send. Test the rules you plan to depend on with a real event.
  • Long records must arrive whole. A record cut by syslog is not parsed as JSON. Apply the message size setting above.
  • The Collector does not store logs. Records that arrive while the UTMStack server is unreachable are lost, so keep the Collector close to the server or on it.

UTMStack is priced per device, never by log volume, and each sensor counts as one device. That matters with a source as talkative as Suricata. Ask for a quote or see pricing.

Frequently asked questions

Where is the Suricata log file? By default in /var/log/suricata. The main one is eve.json. The directory can be changed in suricata.yaml.

Should I send flow records to the SIEM? Only if you will use them. They are the largest part of the output and record every connection. Start with alerts, add DNS, web and TLS records, and add flow last.

Does “allowed” in an alert mean the attack succeeded? No. It means Suricata did not block the packet. In detection mode it never blocks, so every alert says allowed. Check the target’s own logs to see what happened next.

Can Suricata send logs straight to a SIEM without syslog? Suricata can write EVE to a file, to syslog, to a Unix socket or to Redis. It does not open a network connection to a SIEM itself, so you need the local syslog service or a log shipper in between.

Which Suricata version should a sensor run? A supported one. As of November 2026 that is the 8.0 line. The 7.0 line ended on July 7, 2026 and the 6.0 line on August 1, 2024.

Do I still need Suricata if I have a firewall sending logs to the SIEM? They answer different questions. A firewall log says a connection was allowed or denied. Suricata looks inside allowed traffic and says what it contained.

Sources