| name | detecting-modbus-protocol-anomalies |
| description | This skill covers detecting anomalies in Modbus/TCP and Modbus RTU communications in industrial control systems. It addresses function code monitoring, register range validation, timing analysis, unauthorized client detection, and deep packet inspection for malformed Modbus frames. The skill leverages Zeek with Modbus protocol analyzers, Suricata IDS with OT rules, and custom Python-based detection using Markov chain models for normal Modbus transaction sequences.
|
| domain | cybersecurity |
| subdomain | ot-ics-security |
| tags | ["ot-security","ics","scada","industrial-control","iec62443","modbus","protocol-anomaly"] |
| version | 1.0.0 |
| author | mahipal |
| license | Apache-2.0 |
| nist_ai_rmf | ["MEASURE-2.7","MAP-5.1","MANAGE-2.4"] |
| atlas_techniques | ["AML.T0070","AML.T0066","AML.T0082"] |
| nist_csf | ["PR.IR-01","DE.CM-01","ID.AM-05","GV.OC-02"] |
| mitre_attack | ["T1078","T1190","T1059","T0816","T0836"] |
| source | https://github.com/mukul975/Anthropic-Cybersecurity-Skills |
| source_commit | 04450304b12645cb2b974ab96d28c0664758a88d |
| note | Vendored verbatim from an external Apache-2.0 security-skill library, pinned by commit. Exceeds the internal 300-line skill guideline (agent-code-constraints.md) -- kept as-is because this is vendored reference material (forensics/threat-intel procedure), not Yana AI-authored content, and trimming would damage technical accuracy. |
Detecting Modbus Protocol Anomalies
When to Use
- When deploying Modbus-specific intrusion detection in an OT environment
- When building baseline models for deterministic Modbus polling patterns
- When investigating suspicious Modbus traffic flagged by OT monitoring tools
- When implementing function code allowlisting on industrial firewalls
- When detecting unauthorized Modbus write commands that could manipulate process setpoints
Do not use for securing Modbus communications end-to-end (Modbus has no native security; see implementing-network-segmentation-for-ot for firewall-based controls), for non-Modbus protocol monitoring (see detecting-anomalies-in-industrial-control-systems for multi-protocol), or for active fuzzing of Modbus implementations (see performing-plc-firmware-security-analysis).
Prerequisites
- Network SPAN/TAP access to monitor Modbus/TCP traffic (port 502)
- Zeek (formerly Bro) with Modbus protocol analyzer or Suricata with OT rulesets
- Python 3.9+ with scapy and pymodbus for custom analysis
- Baseline capture of normal Modbus traffic (minimum 1-2 weeks)
- Documentation of authorized Modbus clients, function codes, and register maps
Workflow
Step 1: Capture and Parse Modbus Traffic
Deploy passive monitoring to capture all Modbus/TCP traffic and parse it into structured records for analysis.
"""Modbus Protocol Anomaly Detection System.
Monitors Modbus/TCP traffic for anomalies including unauthorized
function codes, unusual register access, timing deviations,
and rogue client devices.
"""
import json
import struct
import sys
import time
from collections import defaultdict, deque
from dataclasses import dataclass, field
from datetime import datetime
from statistics import mean, stdev
try:
from scapy.all import sniff, rdpcap, IP, TCP
except ImportError:
()
sys.exit()
MODBUS_FUNCTION_CODES = {
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
: (, ),
}
:
timestamp:
anomaly_type:
severity:
src_ip:
dst_ip:
unit_id:
func_code:
detail:
mitre_technique: =
:
src_ip:
dst_ip:
func_codes_seen: = field(default_factory=: defaultdict())
register_ranges: = field(default_factory=)
intervals: = field(default_factory=: deque(maxlen=))
last_timestamp: =
request_count: =
write_count: =
:
():
.sessions = {}
.baseline_sessions = {}
.anomalies = []
.authorized_clients = ()
.authorized_func_codes = {}
.packet_count =
():
.authorized_clients = (clients)
():
.authorized_func_codes[session_key] = (func_codes)
():
(baseline_file) f:
baseline = json.load(f)
key, data baseline.get(, {}).items():
.baseline_sessions[key] = data
.authorized_func_codes[key] = (data.get(, []))
()
():
pkt.haslayer(TCP) pkt.haslayer(IP):
pkt[TCP].dport != pkt[TCP].sport != :
payload = (pkt[TCP].payload)
(payload) < :
.packet_count +=
timestamp = (pkt.time)
ts_str = datetime.fromtimestamp(timestamp).isoformat()
:
trans_id = struct.unpack(, payload[:])[]
proto_id = struct.unpack(, payload[:])[]
length = struct.unpack(, payload[:])[]
unit_id = payload[]
func_code = payload[]
(IndexError, struct.error):
pkt[TCP].dport == :
src_ip = pkt[IP].src
dst_ip = pkt[IP].dst
is_request =
:
src_ip = pkt[IP].dst
dst_ip = pkt[IP].src
is_request =
is_request:
session_key =
session_key .sessions:
.sessions[session_key] = ModbusSession(src_ip=src_ip, dst_ip=dst_ip)
session = .sessions[session_key]
session.request_count +=
session.func_codes_seen[func_code] +=
.authorized_clients src_ip .authorized_clients:
.anomalies.append(ModbusAnomaly(
timestamp=ts_str,
anomaly_type=,
severity=,
src_ip=src_ip, dst_ip=dst_ip,
unit_id=unit_id, func_code=func_code,
detail=,
mitre_technique=,
))
allowed_fcs = .authorized_func_codes.get(session_key)
allowed_fcs func_code allowed_fcs:
fc_info = MODBUS_FUNCTION_CODES.get(func_code, (, ))
severity = fc_info[] ==
.anomalies.append(ModbusAnomaly(
timestamp=ts_str,
anomaly_type=,
severity=severity,
src_ip=src_ip, dst_ip=dst_ip,
unit_id=unit_id, func_code=func_code,
detail=,
mitre_technique=,
))
func_code (, , , , , ):
session.write_count +=
fc_name = MODBUS_FUNCTION_CODES.get(func_code, (, ))[]
(payload) >= :
register_addr = struct.unpack(, payload[:])[]
session.register_ranges.add((func_code, register_addr))
.anomalies.append(ModbusAnomaly(
timestamp=ts_str,
anomaly_type=,
severity=,
src_ip=src_ip, dst_ip=dst_ip,
unit_id=unit_id, func_code=func_code,
detail=,
mitre_technique=,
))
session.last_timestamp > :
interval = (timestamp - session.last_timestamp) *
session.intervals.append(interval)
baseline = .baseline_sessions.get(session_key)
baseline (session.intervals) > :
expected_interval = baseline.get(, ) *
expected_std = baseline.get(, ) *
expected_std > :
z_score = (interval - expected_interval) / expected_std
z_score > :
.anomalies.append(ModbusAnomaly(
timestamp=ts_str,
anomaly_type=,
severity=,
src_ip=src_ip, dst_ip=dst_ip,
unit_id=unit_id, func_code=func_code,
detail=(
),
mitre_technique=,
))
proto_id != :
.anomalies.append(ModbusAnomaly(
timestamp=ts_str,
anomaly_type=,
severity=,
src_ip=src_ip, dst_ip=dst_ip,
unit_id=unit_id, func_code=func_code,
detail=,
mitre_technique=,
))
unit_id == func_code (, , , ):
.anomalies.append(ModbusAnomaly(
timestamp=ts_str,
anomaly_type=,
severity=,
src_ip=src_ip, dst_ip=dst_ip,
unit_id=unit_id, func_code=func_code,
detail=,
mitre_technique=,
))
session.last_timestamp = timestamp
():
()
packets = rdpcap(pcap_file)
pkt packets:
.process_packet(pkt)
()
():
()
()
()
()
()
()
severity_counts = defaultdict()
type_counts = defaultdict()
a .anomalies:
severity_counts[a.severity] +=
type_counts[a.anomaly_type] +=
()
sev [, , , ]:
severity_counts[sev]:
()
()
atype, count (type_counts.items(), key= x: -x[]):
()
()
a .anomalies[:]:
()
__name__ == :
detector = ModbusAnomalyDetector()
(sys.argv) > :
(sys.argv) > :
detector.load_baseline(sys.argv[])
detector.analyze_pcap(sys.argv[])
detector.generate_report()
:
()
