· via Hacker News – Front Page (native)
PcapAnalyzer: open-source tool diagnoses network faults from pcaps with 27 detectors
A new open-source, self-hosted tool called PcapAnalyzer reached Hacker News's front page, promising to pinpoint network faults from packet captures using 27 automated detectors while keeping all analysis on the operator's own machine.
The tool
PcapAnalyzer, an open-source project that reached the front page of Hacker News, is built to turn raw packet captures into specific fault diagnoses. According to the project's README on GitHub, a user uploads a .pcap or .pcapng file — gzip-compressed captures are also accepted — and the application reports what is failing, why, and what to check next, with the stated goal of shrinking diagnosis from hours to seconds. The tool is self-hosted, so capture files never leave the operator's network. It is licensed under Apache 2.0 and explicitly labelled beta, with a hosted demo available for anyone who wants to try it without installing anything.
Twenty-seven detectors
The core of the tool is a set of 27 detectors that run against every capture, each producing a severity rating and a suggested next step. The README groups them into six areas:
- TCP: firewall blocks signalled by SYN followed by RST, retransmissions, duplicate ACKs, loss and reordering, zero and full windows, resets after the handshake, slow handshakes.
- Path: PMTUD black holes, low MSS, IP fragmentation, ICMP errors, routing loops.
- Name and address services: unanswered or slow DNS, SERVFAIL responses, NXDOMAIN bursts, DHCP NAK, DECLINE or no answer, duplicate IPs, ARP storms.
- Applications: HTTP 4xx and 5xx errors, slow responses, TLS alerts, and legacy SSL 3.0, TLS 1.0 or 1.1.
- Tunnels: IPsec (IKE and ESP), WireGuard, OpenVPN, GRE, L2TP and Teredo sessions that are one-way, never complete a handshake, or renegotiate endlessly.
- Security: port scans and sweeps, cleartext logins over FTP, POP3, IMAP, SMTP AUTH and HTTP Basic (credentials are never stored or displayed), legacy Telnet, SMBv1 and SNMP v1-v2c, ARP spoofing, and overlapping IP fragments.
Capture quality and Wireshark-style views
The tool also checks whether the capture itself can support a reliable conclusion. It warns about frames duplicated by a mirror port, traffic observed in only one direction, truncated packets, and connections that were already open when the capture started — conditions that frequently mislead manual analysis.
For hands-on inspection, PcapAnalyzer offers Wireshark-style views: a packet list that supports real Wireshark display filters across the whole capture, colouring, CSV export, expert info, HTTP, DNS, DHCP and TLS tables, response-time measurement, and a flow diagram per conversation.
Optional AI investigation
An optional AI layer investigates a capture using tools such as real filters, stream detail and statistics, and cites the packets behind every claim; follow-up questions are supported. It runs on the user's own Anthropic API key, and without one a demo mode executes the same flow using text generated from the local analysis at no cost.
The README is specific about privacy boundaries here: the AI never receives the pcap file, only what its tools return — packet metadata by default, with packet content shared only if explicitly enabled for that session. While an AI investigation runs, the capture is held encrypted and is deleted after 60 minutes of inactivity or on a manual delete command. The self-hosted version collects no telemetry and requires no accounts.
Running it
The quick-start path needs Docker with Docker Compose. A start script generates a random access token — shown once and stored in .env — and brings the app up on port 8080 of localhost. On Windows, the script is meant to be run from Git Bash or WSL. A non-Docker route requires Python 3.9+, Node 20+ and tshark, the Wireshark command-line engine, which is GPL-licensed and runs as a separate program rather than being distributed with the tool.
The app binds to 127.0.0.1 only, and the README warns against exposing it to a network without an authenticated reverse proxy, because the API can use the operator's AI account. The interface ships in English and Spanish, with PDF reports and light and dark themes.
Caveats
The project has been tested against synthetic fixtures and public sample captures, and the maintainer is asking for feedback in the form of captures where the tool misses a genuine problem or reports a false one. Issue reporters are told not to attach captures containing real credentials or customer data. The detailed technical reference — JSON schema, detector thresholds, limits and API errors — currently lives in the Spanish README.
Why it matters
Reading packet traces has traditionally been a specialist skill, and pinpointing a network fault in Wireshark can take hours. PcapAnalyzer packages a meaningful chunk of that expertise into automated detectors with severity ratings and next-step guidance, which could help generalist engineers and small teams that lack a dedicated network specialist. The self-hosted design also addresses the obvious privacy risk: captures can contain credentials and sensitive traffic, so keeping analysis — including the optional AI layer — local is a sensible default. The beta status and reliance on synthetic test fixtures mean findings should still be verified before acting on them, but the combination of automated detection, Wireshark-compatible filtering and optional AI summarisation makes pcap diagnosis notably faster and more accessible.
- #networking
- #open-source
- #packet-analysis
- #diagnostics
- #docker