Hunting For Beaconing With Frequency Analysis

Hunting for Beaconing with Frequency Analysis

When to Use

• When proactively searching for compromised endpoints calling back to C2 infrastructure
• After threat intelligence reports indicate active C2 frameworks targeting your sector
• When network logs show periodic outbound connections to unfamiliar destinations
• During purple team exercises validating C2 detection capabilities
• When investigating a potential breach and need to identify active C2 channels

Detection Gaps & Validation

• High jitter defeats CV thresholds: a CV < 0.20 rule misses beacons configured with large jitter (Cobalt Strike jitter 50, or Sunburst's minutes-scale randomization) — widen the window, score on data-size consistency, and use FFT/autocorrelation rather than CV alone for jittered beacons.
• Long sleep / dormancy: beacons sleeping hours-to-days produce too few connections in a 24h window to be significant — extend the lookback to 7–14 days and lower the min connections floor for rare destinations.
• Destination filtering breaks on domain fronting / cloud abuse: C2 over *.cloudfront.net, Slack, Discord, Telegram, or Graph API rides allowlisted FQDNs — do not exclude major CDNs/SaaS wholesale; pivot on JA3/JA4 and bytes-out regularity instead.
• Log gaps: proxy logs that aggregate or sample connections destroy the inter-arrival timing the math depends on — confirm per-connection Zeek conn.log/firewall records, not summarized flows.
• Validate: run Atomic Red Team T1071.001 or a Cobalt Strike/Sliver test profile with a known interval+jitter and confirm the SPL/KQL streamstats query surfaces it.
• Tune FPs: Windows Update, AV definition pulls, NTP, telemetry, and SaaS heartbeats are textbook periodic — allowlist by destination + process (Sysmon EID 1/3 correlation), not by interval alone.

Prerequisites

• Network proxy/firewall logs with timestamps and destination data (minimum 24 hours)
• Zeek conn.log, dns.log, and ssl.log or equivalent NetFlow/IPFIX data
• SIEM platform with statistical analysis capability (Splunk, Elastic, Microsoft Sentinel)
• RITA (Real Intelligence Threat Analytics) or AC-Hunter for automated beacon analysis
• Threat intelligence feeds for domain/IP reputation enrichment

Workflow

1. Define Beacon Parameters: Establish detection thresholds -- coefficient of variation (CV) below 0.20 indicates strong periodicity, minimum 50 connections over 24 hours, average interval between 30 seconds and 24 hours.
2. Collect Network Telemetry: Aggregate proxy logs, DNS queries, firewall connection logs, and Zeek metadata into the analysis platform.
3. Calculate Connection Intervals: For each source-destination pair, compute the time delta between consecutive connections and derive mean interval, standard deviation, and CV.
4. Apply Jitter Analysis: Sophisticated C2 frameworks like Cobalt Strike add jitter (randomness) to beacon intervals. The Sunburst backdoor beaconed every 15 minutes plus/minus 90 seconds. Analyze jitter patterns to detect even randomized beaconing.
5. Filter Legitimate Periodic Traffic: Exclude known-good beaconing sources including Windows Update, antivirus definition updates, NTP synchronization, SaaS heartbeat services, and CDN health checks.
6. Analyze Data Size Consistency: C2 heartbeat packets typically have consistent payload sizes. Calculate the CV of bytes transferred per connection -- low variance suggests automated communication.
7. Enrich with Threat Intelligence: Check identified beaconing destinations against VirusTotal, WHOIS registration data (flag domains under 30 days old), certificate transparency logs, and passive DNS history.
8. Correlate with Endpoint Telemetry: Map beaconing source IPs to endpoint hostnames via DHCP logs, then correlate with process creation events (Sysmon Event ID 1, 3) to identify the responsible process.
9. Score and Prioritize: Assign risk scores based on CV value, domain age, TI matches, data size consistency, and suspicious port usage. Escalate high-confidence findings.

Key Concepts

Tools & Systems

Detection Queries

Splunk -- HTTP/S Beacon Frequency Analysis

index=proxy OR index=firewall
| where NOT match(dest, "(?i)(microsoft|google|amazonaws|cloudflare|akamai)")
| bin _time span=1s
| stats count by src_ip dest _time
| streamstats current=f last(_time) as prev_time by src_ip dest
| eval interval=_time-prev_time
| stats count avg(interval) as avg_interval stdev(interval) as stdev_interval
  min(interval) as min_interval max(interval) as max_interval by src_ip dest
| where count > 50
| eval cv=stdev_interval/avg_interval
| where cv < 0.20 AND avg_interval > 30 AND avg_interval < 86400
| sort cv
| table src_ip dest count avg_interval stdev_interval cv

KQL -- Microsoft Sentinel Beacon Detection

DeviceNetworkEvents
| where Timestamp > ago(24h)
| where RemoteIPType == "Public"
| summarize ConnectionTimes=make_list(Timestamp), Count=count() by DeviceName, RemoteIP, RemoteUrl
| where Count > 50
| extend Intervals = array_sort_asc(ConnectionTimes)
| mv-apply Intervals on (
    extend NextTime = next(Intervals)
    | where isnotempty(NextTime)
    | extend IntervalSec = datetime_diff('second', NextTime, Intervals)
    | summarize AvgInterval=avg(IntervalSec), StdDev=stdev(IntervalSec)
)
| extend CV = StdDev / AvgInterval
| where CV < 0.2 and AvgInterval > 30
| sort by CV asc

Sigma Rule -- Beaconing Pattern Detection

title: Potential C2 Beaconing Pattern Detected
status: experimental
logsource:
    category: proxy
detection:
    selection:
        dst_ip|cidr: '!10.0.0.0/8'
    timeframe: 24h
    condition: selection | count(dst) by src_ip > 50
level: medium
tags:
    - attack.command_and_control
    - attack.t1071.001

Common Scenarios

1. Cobalt Strike Beacon: Default 60-second interval with configurable 0-50% jitter over HTTPS. Malleable C2 profiles can mimic legitimate traffic patterns.
2. Sunburst/SUNSPOT: 12-14 day dormancy period, then beaconing every 12-14 minutes with randomized jitter, designed to evade frequency analysis.
3. DNS Tunneling C2: Encoded data exfiltration via DNS TXT/CNAME queries to attacker-controlled domains, detectable via high subdomain entropy and query volume.
4. Sliver C2: Modern C2 framework with HTTPS, mTLS, and WireGuard protocols, configurable beacon intervals with built-in jitter support.
5. Legitimate Service Abuse: C2 communication over Slack, Discord, Telegram, or cloud storage APIs, making destination-based filtering ineffective.

Output Format

Hunt ID: TH-BEACON-[DATE]-[SEQ]
Source IP: [Internal IP]
Source Host: [Hostname from DHCP/DNS]
Destination: [Domain/IP]
Protocol: [HTTP/HTTPS/DNS]
Beacon Interval: [Average seconds]
Jitter Estimate: [Percentage]
Coefficient of Variation: [CV value]
Connection Count: [Total connections in window]
Data Size CV: [Payload consistency metric]
Domain Age: [Days since registration]
TI Match: [Yes/No -- source]
Risk Score: [0-100]
Risk Level: [Critical/High/Medium/Low]
Indicators: [List of triggered risk factors]