<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by alaric on Medium]]></title>
        <description><![CDATA[Stories by alaric on Medium]]></description>
        <link>https://medium.com/@alaric0?source=rss-d4b14241809d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*OGvjNc6MLBxuX1AwcUIyWA.png</url>
            <title>Stories by alaric on Medium</title>
            <link>https://medium.com/@alaric0?source=rss-d4b14241809d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 23:28:27 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@alaric0/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[My First Bug Bounty: How I Found Leaked Credentials in NASA’s Git History]]></title>
            <link>https://infosecwriteups.com/my-first-bug-bounty-how-i-found-leaked-credentials-in-nasas-git-history-785316ba7eb8?source=rss-d4b14241809d------2</link>
            <guid isPermaLink="false">https://medium.com/p/785316ba7eb8</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[git]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[infosec]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <dc:creator><![CDATA[alaric]]></dc:creator>
            <pubDate>Sun, 11 Oct 2026 09:01:24 GMT</pubDate>
            <atom:updated>2026-10-11T09:01:24.679Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Hardcoded secrets, CI/CD supply chains, and a lesson in writing reports that actually get triaged.</blockquote><p><strong>TL;DR:</strong> While browsing NASA’s public GitLab instance, I cloned a repository and scanned its full commit history, not just the current files, with a secrets scanner called Gitleaks. It found GitLab CI/CD tokens committed back in 2023. I reported them through NASA’s Vulnerability Disclosure Program on Bugcrowd. The report was triaged as P1, the issue was fixed, and I received a Letter of Recognition.</p><h3>The part nobody tells you about your first bug</h3><p>My first bug bounty finding was not a movie-style hack. There was no green text rain, no “I’m in,” no dramatic music.</p><p>It was a git clone, a command that took eleven seconds to run, and a JSON file that made me sit up straight in my chair.</p><p>But it was real. It was reported, accepted, and fixed. If you’re waiting for your own first accepted finding, I want to show you that the path is a lot more ordinary, and more learnable, than the highlight reels make it look.</p><h3>Finding something worth looking at</h3><p>Every bug hunter eventually asks the same question: where do I even start?</p><p>Public bug bounty programs are picked clean by thousands of people. So I did something simple. I went looking for publicly accessible infrastructure that belongs to a large organization but is easy to overlook: public Git servers, self-hosted DevOps tooling, and internal-ish repos that were never really meant to be browsed.</p><p>That’s how I landed on NASA’s HelioCloud project and its GitLab instance at git.smce.nasa.gov.</p><p>HelioCloud is genuinely cool. It’s a cloud platform that gives researchers access to geospace data and notebooks. It’s the kind of project where a small team moves fast and ships a lot of infrastructure-as-code, and fast-moving infrastructure teams commit a lot of config files.</p><p>Config files are where secrets go to die.</p><h3>Why commit history is the real attack surface</h3><p>Here’s the mental shift that led to the bug:</p><blockquote><em>Developers think in terms of files. Attackers think in terms of history.</em></blockquote><p>When someone leaks a token, they usually do the responsible thing eventually: remove it from the file, rotate it, and move on. The current version of the repo looks clean, and everyone sleeps well.</p><p>But Git is a time machine. Every commit is preserved. That token you deleted last year is still sitting in an old commit, byte for byte, waiting for anyone who thinks to look.</p><p>So instead of reading files, I decided to read the timeline.</p><h3>The tool: Gitleaks</h3><p>I used Gitleaks, a widely used open-source secret scanner. It doesn’t just read the current state of a repo. It walks the entire commit history and matches it against hundreds of rules for known credential formats.</p><p>The commands were genuinely this simple:</p><pre># Clone the repository<br>git clone https://git.smce.nasa.gov/heliocloud/heliocloud-docker-images</pre><pre># Scan the full commit history<br>gitleaks detect --source heliocloud-docker-images/ \<br>  --report-format json \<br>  --report-path result.json</pre><pre># Read the findings<br>cat result.json</pre><p>That’s it. No exploitation, no fuzzing, no clever payloads. Just looking at what history had left behind.</p><p>Gitleaks came back with results.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/762/1*lq2VSFLRRMQLcOlD7W0S4Q.png" /></figure><h3>What it found</h3><p>The findings were hardcoded credentials committed in April and May of 2023 across a couple of Kubernetes/Helm values files. I’m masking the actual values here. The tokens are dead now, but a public writeup doesn’t need to republish secrets.</p><p><strong>1. GitLab Runner Registration Token</strong> in config-values.yaml</p><pre># config-values.yaml (historical commit)<br>runnerRegistrationToken: &quot;GR1348941••••••••••••••••••••&quot;</pre><p><strong>2. GitLab Agent Access Token</strong> in gitlab-agent-values.yaml</p><pre># gitlab-agent-values.yaml (historical commit)<br>token: &quot;PjRn5hrx••••••••••••••••••••••••••••••&quot;</pre><p>A trimmed version of the Gitleaks report looked like this:</p><pre>[<br>  {<br>    &quot;Description&quot;: &quot;GitLab Runner Registration Token&quot;,<br>    &quot;File&quot;: &quot;config-values.yaml&quot;,<br>    &quot;Commit&quot;: &quot;113e0ebb••••••••&quot;,<br>    &quot;Author&quot;: &quot;••••••••@nasa.gov&quot;,<br>    &quot;Date&quot;: &quot;2023-05-04T22:37:50Z&quot;,<br>    &quot;RuleID&quot;: &quot;gitlab-rrt&quot;<br>  },<br>  {<br>    &quot;Description&quot;: &quot;Generic API Key&quot;,<br>    &quot;File&quot;: &quot;gitlab-agent-values.yaml&quot;,<br>    &quot;Commit&quot;: &quot;e5b934f6••••••••&quot;,<br>    &quot;Author&quot;: &quot;••••••••@nasa.gov&quot;,<br>    &quot;Date&quot;: &quot;2023-04-06T16:03:00Z&quot;,<br>    &quot;RuleID&quot;: &quot;generic-api-key&quot;<br>  }<br>]</pre><p>To confirm each secret was real and not a placeholder, I pulled the exact file contents straight out of the historical commits:</p><pre>git show 113e0ebb:config-values.yaml<br>git show e5b934f6:gitlab-agent-values.yaml</pre><p>There they were, in black and white, dated 2023.</p><h3>Why this actually matters</h3><p>Precision is the difference between a report that gets paid and one that gets closed as informational, so let me be precise.</p><p>This is <strong>CWE-798: Use of Hardcoded Credentials</strong>. The interesting part isn’t the category, though. It’s what those two specific tokens do.</p><p><strong>A GitLab Runner Registration Token</strong> is used to register a runner with GitLab. If it’s live, anyone holding it can register a rogue runner to the project. That runner then picks up CI/CD jobs, which means it can end up executing pipeline code, touching build artifacts, and reading CI/CD environment variables, which is usually where more secrets live.</p><p>That’s a software supply chain risk. You’re not just compromising a box. You’re inserting yourself into the assembly line that produces the software.</p><p><strong>A GitLab Agent token</strong> (for the GitLab agent for Kubernetes) is used by the cluster’s agent to authenticate to GitLab. Depending on how it’s configured and scoped, it’s a foothold aimed at cluster-side access.</p><p>And the kicker: all of this lives in history. Even after the files were cleaned up, the tokens stayed in the old commits. Unless someone explicitly revokes them in GitLab, they may still be valid years later.</p><h3>Testing responsibly, and a limitation I was upfront about</h3><p>This is where a lot of first-timers get themselves, and their reports, into trouble, so I want to be clear.</p><p>I did not abuse anything. I only performed validation that stays within authorized, non-destructive bounds.</p><p>I made a single API call to see whether the token was still active:</p><pre>curl --header &quot;Authorization: Bearer PjRn5hrx••••••••••••&quot; \<br>  &quot;https://git.smce.nasa.gov/api/v4/projects&quot;</pre><p>The response was <strong>401 Unauthorized</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nYDgxmO9eMkMB-ZZabMOBA.png" /></figure><p>That told me the token had already been revoked or had expired since 2023. I said so plainly in my report. I didn’t pretend the token was live. The honest framing was:</p><blockquote><em>The credential is currently inactive, but its presence in version history is still a security risk. It violates secure-coding practice (CWE-798), it shows that secrets were being committed to the repo, and the history needs to be scrubbed so this can’t bite the project in the future.</em></blockquote><p>That’s the real value of a finding like this. It isn’t “I got in.” It’s “your pipeline has a habit, and here’s the evidence.”</p><h3>The report (and the part I got wrong first)</h3><p>I submitted through NASA’s Vulnerability Disclosure Program on Bugcrowd.</p><p>My first version had the finding but not the storytelling. The triage team came back almost immediately:</p><blockquote><em>“Your steps are unclear. Could you please consolidate all the steps… use full URLs where possible… provide the raw HTTP requests in Markdown format.”</em></blockquote><p>This was the single most useful thing that happened to me on this entire journey.</p><p>I had assumed that because I could reproduce it, everyone else could too. I was wrong. A report isn’t a diary. It’s an instruction manual for a stranger.</p><p>So I rewrote it into three clean, numbered, copy-pasteable steps:</p><ol><li><strong>Discovery:</strong> clone and Gitleaks scan, with exact commands.</li><li><strong>Manual verification:</strong> git show &lt;commit&gt;:&lt;file&gt; for each secret.</li><li><strong>Proof of Concept:</strong> the scoped validation check, with an explicit note to only test systems you own or are authorized to test.</li></ol><p>I also attached a walkthrough video and screenshots so the reproduction matched my written steps one to one. The submission moved to Triaged the same day, with severity set to <strong>P1</strong>.</p><h3>The outcome</h3><ul><li><strong>Submitted:</strong> 06 Apr 2026</li><li><strong>Triaged as P1:</strong> 17 Apr 2026</li><li><strong>Retested and resolved:</strong> 25 Jun 2026</li><li><strong>Recognition:</strong> Letter of Recognition from NASA</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*b62B89XbIzqG7COtaCAk9Q.jpeg" /></figure><p>NASA’s team also asked researchers to allow 90 days before publishing anything publicly. That’s a reasonable norm, and it’s exactly what I did. That waiting period is why you’re reading this now and not in April.</p><p>No fancy trophy. Just a PDF letter and the quiet satisfaction of knowing a real team changed something real because of work I did.</p><h3>What I’d do differently, and what you should do</h3><p>If you take nothing else from this, take these:</p><ol><li><strong>Scan history, not just files.</strong> gitleaks detect walked the whole timeline. That&#39;s where the value was.</li><li><strong>Write for a stranger.</strong> Numbered steps, full URLs, raw requests, copy-pasteable commands. My report got triaged the day I fixed the steps.</li><li><strong>Be honest about impact.</strong> Saying “the token is already revoked” didn’t weaken my report, it strengthened it. Reviewers can tell when you’re padding.</li><li><strong>Never test beyond your authorization.</strong> A one-line, non-destructive check confirmed the token’s state, and that was plenty. You don’t need to register a rogue runner to prove a point.</li><li><strong>Respect the disclosure window.</strong> Let the organization fix it first.</li></ol><h3>Closing thoughts</h3><p>When I started, I thought a “real” bug meant exploitation: some clever chain, some flashy payload. My first accepted finding was a git command and a bit of curiosity about what people leave behind.</p><p>The secret wasn’t a secret at all. It was just sitting in the past, waiting for someone to look.</p><p>If you’re chasing your first bug, go look at the places that are easy to forget: the history, the backup, the config, the thing someone deleted a year ago.</p><p>And write it down like the person reading it has never seen your screen.</p><p>That’s the whole trick.</p><p><em>Found with Gitleaks. Reported responsibly through NASA’s Vulnerability Disclosure Program via Bugcrowd. All secrets shown in this article are masked, and the underlying credentials were reported, not published. Thanks to the NASA security team and Bugcrowd triage for a clear and fair process.</em></p><p>#Cybersecurity, #BugBounty, #Infosec, #DevOps, #Git</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=785316ba7eb8" width="1" height="1" alt=""><hr><p><a href="https://infosecwriteups.com/my-first-bug-bounty-how-i-found-leaked-credentials-in-nasas-git-history-785316ba7eb8">My First Bug Bounty: How I Found Leaked Credentials in NASA’s Git History</a> was originally published in <a href="https://infosecwriteups.com/">InfoSec Write-ups</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>