<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Soren Vestergaard</title>
    <description>The latest articles on DEV Community by Soren Vestergaard (@soren_vest).</description>
    <link>https://dev.to/soren_vest</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4215013%2F5e5af650-4749-477d-b5b7-52ca444c2379.jpg</url>
      <title>DEV Community: Soren Vestergaard</title>
      <link>https://dev.to/soren_vest</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/soren_vest"/>
    <language>en</language>
    <item>
      <title>Threat modeling a feature in 20 minutes</title>
      <dc:creator>Soren Vestergaard</dc:creator>
      <pubDate>Sun, 01 Nov 2026 15:07:07 +0000</pubDate>
      <link>https://dev.to/soren_vest/threat-modeling-a-feature-in-20-minutes-i92</link>
      <guid>https://dev.to/soren_vest/threat-modeling-a-feature-in-20-minutes-i92</guid>
      <description>&lt;p&gt;I used to treat threat modeling as a whiteboard meeting. Now I do it in 20 minutes with a text file, and it catches more.&lt;/p&gt;

&lt;p&gt;The four questions:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;What are we building?&lt;/strong&gt; One paragraph, assets and actors. No diagrams needed yet.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;What can go wrong?&lt;/strong&gt; Walk the trust boundaries: browser → API → queue → worker → database. At each boundary, ask who can lie.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;What do we do about it?&lt;/strong&gt; Every abuse case gets one of: validate, authorize, rate limit, log, or accept.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Did we miss anything?&lt;/strong&gt; Check the list from last time.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;The output is a checklist that goes in the PR, not a document nobody opens.&lt;/p&gt;

&lt;p&gt;The trick that keeps it short: only model what changed. You don't need a new threat model for a copy change.&lt;/p&gt;

&lt;p&gt;How do you keep the model from becoming a document graveyard?&lt;/p&gt;

</description>
      <category>security</category>
      <category>appsec</category>
      <category>architecture</category>
    </item>
    <item>
      <title>Notes from triaging forty SSRF reports</title>
      <dc:creator>Soren Vestergaard</dc:creator>
      <pubDate>Thu, 29 Oct 2026 10:31:38 +0000</pubDate>
      <link>https://dev.to/soren_vest/notes-from-triaging-forty-ssrf-reports-550h</link>
      <guid>https://dev.to/soren_vest/notes-from-triaging-forty-ssrf-reports-550h</guid>
      <description>&lt;p&gt;Forty SSRF reports in a year sounds impressive until you see the distribution.&lt;/p&gt;

&lt;p&gt;About half were the same bug: a URL parameter that the backend fetched for an image preview. The fix was one allowlist on the fetch client, applied in three services. The reports were real, but they were one finding counted forty times.&lt;/p&gt;

&lt;p&gt;A quarter were redirects. Someone validated the host, then followed a 302 to 169.254.169.254. The lesson isn't 'validate twice', it's that the fetch client has to be the thing that enforces the rule, not the controller above it.&lt;/p&gt;

&lt;p&gt;The rest were genuinely interesting, including two that reached internal admin panels through a proxy we forgot we had.&lt;/p&gt;

&lt;p&gt;What changed after: every fetch goes through one internal client with an allowlist and a five-second timeout, and it logs the resolved IP, not the hostname. We also wrote a tiny canary endpoint that reports if it ever gets hit.&lt;/p&gt;

&lt;p&gt;Volume is not signal. One allowlist did more than forty tickets.&lt;/p&gt;

</description>
      <category>security</category>
      <category>appsec</category>
      <category>webdev</category>
    </item>
  </channel>
</rss>
