Overview
Command and control traffic is often encrypted, short and unremarkable one connection at a time. What gives it away is the rhythm: the same host reaching the same destination every minute, for hours.
Beacon telemetry looks for that rhythm in the IDS flow table and reports each regular pair, then reports it again after a quiet period. It decides nothing; that is left to you and your rules.
When you see this
- A host reaches the same external destination at a fixed interval.
- The traffic classifier rates a flow as plausibly C2 without it being the top class.
- Connections continue overnight, when nobody is working.
Steps
- 01
Turn on beacon telemetry
It needs ids.enabled and the XDP flow table: not AF_PACKET, and not Windows. Changing it takes a restart. If logging.eve_log.event_types is set, it must list beacon.
/etc/synapse/config.yaml ids: enabled: true capture_mode: xdp beacon: enabled: true min_conns: 8 min_span_secs: 600 realert_secs: 3600 ignore_dst_ports: [53, 123] - 02
Leave out what beacons by design
DNS, NTP and update checks are regular on purpose. ignore_dst_ports keeps those pairs tracked but never reported. Internal destinations are not reported unless report_internal_dst is on.
- 03
Read the records
Each regular pair produces one beacon record in the EVE output: after min_conns connections spanning at least min_span_secs, and again after realert_secs.
- 04
Confirm with a second signal
A steady interval alone is not proof: plenty of legitimate software polls. Look for a second signal on the same host: the traffic classifier's C2 probability, an IDS alert, a destination the feeds already know.
Smart firewall rule, action log ml.traffic.c2 ge 90 and ids.alert_count > 0DecisionDoes a second signal agree?Yes Contain the host.
No Keep the pair under review. If it is legitimate polling, leave it; if the port is noise, ignore it.
- 05
Contain the host
Cut its egress without pulling the cable: allow what the investigation needs, deny everything else, and keep watching. The containment playbook has the steps.
Common mistakes
- Treating a beacon record as an alert. It carries no score.
- Turning it on with AF_PACKET capture or on Windows, where it needs the XDP flow table it does not have.
- Leaving DNS and NTP to be reported instead of ignoring those ports.
Escalate when
- A beaconing host that also has an IDS alert.
- The same destination beaconed to from several hosts.
Where this stops working
- A pair is reported only after min_conns connections spanning min_span_secs: something that calls home rarely takes longer to show.
- It is telemetry, not prevention: nothing is blocked until a rule acts.