<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Firmware on Martin&#39;s Blog</title>
    <link>https://mac.sploit.dk/tags/firmware/</link>
    <description>Recent content in Firmware on Martin&#39;s Blog</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Wed, 15 Apr 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://mac.sploit.dk/tags/firmware/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Synology findhostd: Auditing an Unauthenticated Endpoint Running as Root</title>
      <link>https://mac.sploit.dk/blog/synology-agent-audit/</link>
      <pubDate>Wed, 15 Apr 2026 00:00:00 +0000</pubDate>
      <guid>https://mac.sploit.dk/blog/synology-agent-audit/</guid>
      <description>&lt;p&gt;This started with a simple question to a Claude model: &lt;em&gt;&amp;ldquo;My NAS is at&#xA;192.168.0.103. What version is it running?&amp;rdquo;&lt;/em&gt;&lt;/p&gt;&#xA;&lt;p&gt;I wanted to see how far I could push a conversational AI model to perform real&#xA;security research on my personal Synology NAS. The setup was a GitHub Copilot&#xA;CLI with Claude Opus 4.6 running in a terminal agent (Ubuntu) with access to&#xA;standard tools, plus a handful of custom skills I had written previously:&#xA;&lt;code&gt;decompile-idapro&lt;/code&gt; (drives IDA Pro 9.3 headless decompilation),&#xA;&lt;code&gt;binary-analysis&lt;/code&gt; (triage and inspection of ELF/PE binaries), and &lt;code&gt;find-vulns&lt;/code&gt;&#xA;(structured vulnerability scanning with CWE tagging). The skills give the model&#xA;capabilities it would not have out of the box, but the research direction was&#xA;entirely unscripted &amp;ndash; just natural language prompts, one leading into the next,&#xA;following whatever thread looked interesting.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
