<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Vulnerability-Research on eikendev</title><link>https://www.eiken.dev/tags/vulnerability-research/</link><description>Recent content in Vulnerability-Research on eikendev</description><generator>Hugo</generator><language>en-US</language><managingEditor>hello@eiken.dev (eikendev)</managingEditor><webMaster>hello@eiken.dev (eikendev)</webMaster><lastBuildDate>Sun, 15 Nov 2026 17:08:50 +0100</lastBuildDate><atom:link href="https://www.eiken.dev/tags/vulnerability-research/index.xml" rel="self" type="application/rss+xml"/><item><title>Can open-weight models find a 0-day RCE?</title><link>https://www.eiken.dev/blog/2026/11/can-open-weight-models-find-a-0-day-rce/</link><pubDate>Sun, 15 Nov 2026 16:07:00 +0000</pubDate><author>hello@eiken.dev (eikendev)</author><guid>https://www.eiken.dev/blog/2026/11/can-open-weight-models-find-a-0-day-rce/</guid><description>&lt;p&gt;&lt;a href="https://www.eiken.dev/blog/2026/11/can-open-weight-models-reverse-engineer-a-stripped-rust-binary/"&gt;Last time&lt;/a&gt; I handed four open-weight models a stripped Rust binary and asked them to explain a particular detail of it.&#10;This time I wanted to know whether they can find a &lt;em&gt;real&lt;/em&gt; 0-day in somewhat popular software.&#10;So I gave the same four models the source code of &lt;a href="https://github.com/alexta69/metube"&gt;MeTube&lt;/a&gt;, a Python web app with 10k+ stars on GitHub, and asked them to find the most severe vulnerability they could.&lt;/p&gt;&#10;&lt;p&gt;I already knew the answer.&#10;MeTube had a &lt;a href="https://en.wikipedia.org/wiki/Arbitrary_code_execution"&gt;remote code execution&lt;/a&gt; (RCE) vulnerability reachable from its web API (triggered after the app restarts), and I had found and confirmed it before I let the models run.&#10;At the time of my experiment, the vulnerability was privately reported but still in triage.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;&#10;&lt;p&gt;So how did the models do?&#10;One of them found the vulnerability in every single run, and the best model from the reverse-engineering challenge came last.&lt;/p&gt;</description></item></channel></rss>