Skip to content

What's done

Last updated 19 Nov 2026.

Research: how Nessus works

All of this lives in the nessus-dissection repo.

  • [x] Unpacked the Nessus 10.12.5 macOS installer with our own small tools (DMG, pkg and universal binary extractors)
  • [x] Found that most of Nessus (web server, API, CLI, all checks) is written in its own script language, NASL, and runs inside one big C/C++ program
  • [x] Pulled out the database layout, around 280 API routes, about 400 scan settings and the full plugin function list
  • [x] Decompiled the small watchdog binary with Ghidra (the main 13 MB engine is still to do)
  • [x] Ran Nessus Essentials live and measured two scans: one host took 14 to 17 minutes, and 64% to 72% of that time was spent stuck at "99%"
  • [x] Found that Nessus looks at all ~345,000 plugins for every host and skips about 98% of them
  • [x] Collected public complaints about Nessus (slow, hard to set up, false positives, paywalls)
  • [x] Wrote down 17 weak points and how Nuqs should do better

Decisions

  • [x] Language: Go (see Decisions)
  • [x] One core with the CLI now, and the API and web UI later as thin layers on top
  • [x] Nmap does the low-level scanning, we don't reinvent it
  • [x] Plan for keeping up with new CVEs (see Keeping up with new CVEs)
  • [x] 4-month roadmap, tightened so each week does a lot more
  • [x] Web UI served by the binary, no Qt desktop app (see Decisions)
  • [x] Repo moved to nuqs-org/nuqs, with all commits kept
  • [x] Roadmap split into FYP-I (presentation and open house on 5 Dec) and FYP-II

Nuqs core (milestone 0)

  • [x] Target validation (IPs, networks, ranges, hostnames, a safety limit, and blocking Nmap option injection)
  • [x] Nmap runner and XML parser
  • [x] The Check interface, the main way to extend Nuqs
  • [x] Scan engine: picks checks per service, runs them in parallel with timeouts, survives errors, hangs and panics
  • [x] 3 checks: telnet-open, http-security-headers, ftp-anonymous-login
  • [x] Text and JSON reports, every finding has evidence and a fix
  • [x] nuqs scan, nuqs checks, nuqs version
  • [x] Tests: unit tests for everything, fake servers for each check, fuzz tests for both parsers, an integration test with real Nmap, ~97% coverage
  • [x] CI: format, vet, lint (including "everything public is documented"), race detector, coverage gate, fuzzing, integration, govulncheck, build
  • [x] Docs website built and published from CI, with API pages generated from code comments
  • [x] Architecture page (diagram, package table and file map) generated from the code and updated by CI on every push, with a test that checks it against go list
  • [x] Every file starts with a one-line description, the file map on the architecture page is built from them, and a test fails if one is missing
  • [x] Nicer API pages: one heading per item, the description first, and a link to the code
  • [x] scripts/docs.sh installs what the docs need and builds or serves them in one command
  • [x] Zero dependencies, standard library only

Next

Week 1 of the roadmap: version comparison, confidence on findings, the test lab, 5 checks, UI wireframes, and the first docs (background, survey, scope). Everything builds towards the FYP-I presentation on 5 Dec.