cd /news/structured-data/building-hbom-for-a-router · home › topics › structured-data › article
[ARTICLE · art-170235] src=worklifenotes.com ↗ pub= topic=structured-data verified=true sentiment=· neutral

Building HBOM for a Router

A hardware bill of materials for a D-Link R18 Wi-Fi router was generated by teardown and published in the ReARM repository, compiled against the unreleased CycloneDX 2.0-dev specification at commit ea8c360. The author used Claude to identify components from label and board photos and a 115200-baud UART bootlog, documenting an Econet EN7528HU SoC, Nanya NT5CC128M16JR-EK 256 MiB DRAM, Macronix MX35LF1GE4AB-Z4I SPI-NAND flash, and a Wi-Fi chip listed as a CycloneDX 2.0 component choice between MT7905DAN and MT7915DAN with MT7975 A-die. The exercise found gaps in the specification, including no way to represent a "known but not disclosed" state such as a serial number, and noted that teardown alone cannot establish chip origin.

by read4 min views1 publishedNov 16, 2026
Building HBOM for a Router
Image: Worklifenotes (auto-discovered)

With upcoming CycloneDX 2.0 release, a lot of improvements were made regarding its Hardware Bill of Materials (HBOM) capabilities. In this post, I’ll describe process how you can generate an HBOM for a Wi-Fi router.

In my case, I’m using D-Link R18 worth about ~30$ and sitting unused in my case. In parallel with HBOM I worked with Claude to add it OpenWRT support (which worked and which I would describe separately). But since I participated in HBOM group work, I thought this would be a good exercise.

First, important note that it’s always easier for manufacturer to create SBOMs, HBOMs and other BOMs. But since those rarely exist and if they are usually remain undistributed, I’ll show how we can make one ourselves.

Table of Contents

I Label

Router usually has label, which has some useful information, which in my case included product number, serial number (which I left out of the published HBOM), model number, country of origin, MAC address and several regulatory classifications.

If you don’t worry about exposing network default passwords, take a picture and feed it to Claude.

II Tear Down

This was easy on this router. Unscrew 2 screws on the bottom and pry open. Then take detailed pictures of everything including close ups of chips and feed everything to Claude.

In my case this was first overall board view, where we can see Econet SoC EN7528HU:

Then take close up picture of DRAM Flash, Nanya NT5CC128M16JR-EK (U1001), 256 MiB:

Next take a picture of the rear buttons for identification:

and another close up, this time of SPI-NAND Macronix MX35LF1GE4AB-Z4I (U15 Flash):

III Collect Bootlog

Next, identify boot UART serial (there were 3 candidate ports), connect serial and collect bootlog using picocom, in my case:

picocom -b 115200 /dev/ttyUSB0

and let Claude take the log and analyze it.

IV Compile against the spec

With all data gathered, just ask Claude to compile against the specification. While CycloneDX 2.0 is not released yet, I was using latest 2.0-dev branch at commit ea8c360, which should be close to the final release as HBOM-related work was already merged.

V Gaps

Main gap that I had was Wi-Fi chip protected by metal shield, which I didn’t want to remove (as I didn’t want to break the thing, at list yet). But the chip itself was down to 2 candidates. Therefore, we used CycloneDX 2.0 component choice addition which essentially means, this can be one of several components. In my case: MT7905DAN or MT7915DAN, with MT7975 A-die.

Further, evaluating the document against recent IST paper on HBOMs by Allan Friedman and Nicholas Leiserson, we could improve coverage by stating what levels are explicitly excluded or included. This would likely require follow up work in CycloneDX specification.

In this HBOM preparation, I was mostly focusing here on the major chips (SoC, DRAM, flash and Wi-FI), plus the connectors and other user-facing parts. Power-management chips, discrete semiconductors and passive components are out of scope for this HBOM.

Additionally, with preparing HBOM via teardown, it is nearly impossible to determine origin of chips. For this, deeper understanding and monitoring of actual supply chains is required.

Also, it looks follow up work might be needed in CycloneDX specification to represent “known but not disclosed” state. In my case, this would be, for example, serial number.

VI Result and Summary

Resulting HBOM is published in ReARM repository here with its provenance document, describing step-by-step generation with several versions that we had, which data came from label, from images, from boot log and from my observations (i.e. instructions to connect serial).

Overall, I spent couple of days on this while getting interrupted on other things and doing OpenWRT port work in parallel. If the goal would be to take HBOM only, the process should take about 3-4 hours per steps above and is fairly straightforward.

While it is unlikely that an end-user would be interested in producing HBOM documents, it is an invaluable tool in the supply chain security, especially when dealing with many devices of the same type.

Once HBOM becomes an inventory item, it can then be quickly analyzed for findings and violations, similar to SBOM. When propagating large number of devices through supply chains, few hours to collect data seems like a small investment relative to the benefits that such inventory document provides.

── more in #structured-data 4 stories · sorted by recency
── more on @d-link r18 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/building-hbom-for-a-…] indexed:0 read:4min 2026-11-16 · —