<?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 Cyberrado on Medium]]></title>
        <description><![CDATA[Stories by Cyberrado on Medium]]></description>
        <link>https://medium.com/@cyberrado57?source=rss-b49b58fd445e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*KfQqmRnW24tB2WSJ</url>
            <title>Stories by Cyberrado on Medium</title>
            <link>https://medium.com/@cyberrado57?source=rss-b49b58fd445e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 18:22:32 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@cyberrado57/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[Exploiting Supabase Misconfigurations: Authentication Bypass and Privilege Escalation for $1,800]]></title>
            <link>https://medium.com/@cyberrado57/exploiting-supabase-misconfigurations-authentication-bypass-and-privilege-escalation-for-1-800-f20d02557e47?source=rss-b49b58fd445e------2</link>
            <guid isPermaLink="false">https://medium.com/p/f20d02557e47</guid>
            <dc:creator><![CDATA[Cyberrado]]></dc:creator>
            <pubDate>Sun, 26 Apr 2026 17:21:03 GMT</pubDate>
            <atom:updated>2026-04-26T17:26:29.136Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1PpyJ-baPHp7-agnQFIupg.png" /></figure><p>Hey everyone, <a href="https://x.com/RadoGold57"><strong>CyberR</strong></a> here.</p><p>Backend-as-a-Service (BaaS) platforms like Supabase and Firebase have drastically accelerated development cycles. However, the shared responsibility model means that while the provider secures the infrastructure, the developer is responsible for the configuration. When those configurations are slightly off, the results can be catastrophic.</p><p>Recently, I was hunting on the program and found a critical misconfiguration in their Supabase authentication implementation. By chaining an auto-confirmed email flaw with a user metadata injection, I was able to bypass authentication and escalate my privileges to admin.</p><p>Here is the breakdown of the $1,800 critical vulnerability, how to exploit it, and how to secure against it.</p><h3>The Target</h3><p><strong>Company:</strong> Redacted <strong>Vulnerability Type:</strong> Improper Authentication — Generic (CWE-287) <strong>CVSS Score:</strong> 9.1 (Critical)<strong>Bounty:</strong> $1,800</p><h3>The Discovery: Leveraging the Public anon Key</h3><p>While analyzing ******, I noticed the application relied heavily on Supabase for its backend operations. By inspecting the network traffic in Chrome Developer Tools, I grabbed the Supabase REST URL (https://******) and the public anon key passed in the apikey header.</p><p>A common misconception in bug bounty is that finding the anon key is a vulnerability in itself. It is not. The anon key is public by design, intended to be used by the frontend to communicate with Supabase. However, this key is the doorway. The real vulnerabilities lie in <em>how the backend handles the requests</em> coming through that door.</p><p>In a previous exploit on this same program, I had leaked a list of valid client organization domains (e.g., ******.com) from a poorly secured partners table. I decided to see if I could register an account using one of these legitimate corporate domains.</p><h3>The Flaw: Auto-Confirmed Signups</h3><p>Normally, when a user registers for an enterprise application using a corporate email, the system enforces an email verification step (a magic link or OTP) before granting an active session.</p><p>I sent a POST request to the Supabase /auth/v1/signup endpoint using the exposed anon key, registering a fake email on the target partner&#39;s domain (test_audit@******.com).</p><p>To my surprise, the API immediately responded with a 200 OK and a valid Bearer JWT.</p><p>JSON</p><pre>{<br>  &quot;access_token&quot;: &quot;eyJhbGciOiJIUzI1Ni...&quot;,<br>  &quot;token_type&quot;: &quot;bearer&quot;,<br>  &quot;user&quot;: {<br>    &quot;role&quot;: &quot;authenticated&quot;,<br>    &quot;email&quot;: &quot;test_audit@******.com&quot;,<br>    &quot;email_confirmed_at&quot;: &quot;2026-01-27T03:48:14.160Z&quot;,<br>    &quot;last_sign_in_at&quot;: &quot;2026-01-27T03:48:14.179Z&quot;<br>  }<br>}</pre><p>The system was misconfigured to automatically confirm email addresses (&quot;email_confirmed_at&quot; was populated immediately). I now had a valid, authenticated session as an employee of a partner organization, without ever needing access to their email inbox.</p><h3>The Escalation: Injecting User Metadata for Admin Access</h3><p>While an authentication bypass is a great find, I wanted to push the impact further.</p><p>Supabase’s GoTrue auth API allows users to pass a data object during signup, which populates the user_metadata field in the resulting JWT. If developers mistakenly rely on this client-provided metadata for authorization or Row Level Security (RLS) policies, it leads to immediate privilege escalation.</p><p>I modified my signup payload to inject a custom role:</p><pre>POST /auth/v1/signup HTTP/2<br>Host: ******<br>Content-Type: application/json;charset=UTF-8<br>Apikey: eyJhbGciOiJIUzI1... &lt;ANON_KEY&gt;<br>Authorization: Bearer eyJhbGciOiJIUzI1... &lt;ANON_KEY&gt;</pre><pre>{<br>    &quot;email&quot;: &quot;admin_audit@******.com&quot;,<br>    &quot;password&quot;: &quot;Password123!&quot;,<br>    &quot;data&quot;: {<br>        &quot;full_name&quot;: &quot;Audit Test&quot;,<br>        &quot;role&quot;: &quot;admin&quot;<br>    }<br>}</pre><p>When I decoded the resulting JWT access token using jwt.io, my custom &quot;role&quot;: &quot;admin&quot; key-value pair was baked right into the payload. Because the application’s frontend and RLS policies relied on this metadata to dictate access, this account was immediately granted administrative privileges over the partner&#39;s tenant.</p><h3>The Triage Process</h3><p>This report required some back-and-forth. Initially, the triage team struggled to reproduce the issue because they were unsure how to obtain the API key and didn’t immediately see the privilege escalation vector.</p><p>I had to clarify two crucial points:</p><ul><li>The anon key is dynamically loaded in the frontend and is inherently accessible to anyone intercepting traffic.</li><li>The core impact wasn’t <em>just</em> creating a phantom account, but the arbitrary injection of the &quot;role&quot;: &quot;admin&quot; parameter into the JWT.</li></ul><p>After a bit of a wait (three weeks of radio silence can be agonizing!), the team prioritized the ticket, successfully reproduced the exploit, and awarded an $1,800 bounty for a Critical (9.1) finding.</p><h3>Key Takeaways for DevSecOps</h3><p>BaaS misconfigurations are incredibly common. If you are building with Supabase or Firebase, keep these principles in mind:</p><ul><li><strong>Enforce Email Confirmations:</strong> Always ensure email_confirm is set to true in your auth settings. Never issue an active access_token until the user has proven ownership of the inbox, especially in B2B or multi-tenant SaaS environments.</li><li><strong>Never Trust Client-Side Metadata for AuthZ:</strong> The user_metadata field is generally writable by the user. <strong>Do not</strong> use it to store sensitive permission flags like roles or subscription tiers. Store roles in a secure, backend-only Postgres table (e.g., a public.users table) and use Row Level Security (RLS) to ensure users cannot modify their own roles.</li><li><strong>Audit Your RLS Policies:</strong> Ensure your database policies aren’t relying strictly on email domains (e.g., auth.email() LIKE &#39;%@******.com&#39;) unless you have strictly enforced email verification.</li></ul><p>A secure architecture relies on defense in depth. Don’t let a single toggle switch compromise your entire perimeter.</p><p>Until the next bug, happy hunting!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f20d02557e47" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Hunting Bugs in High Fashion: Chaining SSO Misconfigurations and Broken Access Control at a Major…]]></title>
            <link>https://medium.com/@cyberrado57/hunting-bugs-in-high-fashion-chaining-sso-misconfigurations-and-broken-access-control-at-a-major-62546a3e42f1?source=rss-b49b58fd445e------2</link>
            <guid isPermaLink="false">https://medium.com/p/62546a3e42f1</guid>
            <dc:creator><![CDATA[Cyberrado]]></dc:creator>
            <pubDate>Sun, 26 Apr 2026 12:30:55 GMT</pubDate>
            <atom:updated>2026-04-26T12:30:55.501Z</atom:updated>
            <content:encoded><![CDATA[<h3>Hunting Bugs in High Fashion: Chaining SSO Misconfigurations and Broken Access Control at a Major Luxury Brand</h3><p>Hey everyone, <strong>CyberR</strong> here.</p><p>When hunting on major programs, finding a standalone bug is great, but chaining multiple low-impact misconfigurations into a concrete exploit is where the real fun begins. Recently, I spent some time analyzing a global luxury brand’s infrastructure and managed to chain an exposed configuration file with a completely unrestricted Single Sign-On (SSO) integration.</p><p>This ultimately led to Broken Access Control on an internal microservice, allowing arbitrary data injection. The best part? The systemic nature of the misconfiguration meant it affected their Integration (INT), Pre-Production (PRP), and Production (PRD) environments — resulting in separate payouts for each.</p><p>Here is the technical breakdown of the vulnerability, the attack chain, and how a few simple oversights can compromise an internal admin portal.</p><h3>The Target</h3><p><strong>Company:</strong> Redacted Luxury Brand</p><p><strong>Program:</strong> YesWeHack</p><p><strong>Vulnerability Type:</strong> Improper Access Control — Generic (CWE-284)</p><p><strong>Scope:</strong> *.redacted.com</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*doxGP9Z-RiYq0uEfKZtyjg.png" /></figure><h3>The Discovery Phase: Leaking the Keys</h3><p>The journey started on an internal-facing admin portal: admin-int.redacted.com. While the portal itself was shielded by a login page, the front-end assets were not perfectly secured.</p><p>I discovered a publicly accessible configuration file located at /public/config.js. When I pulled this file, it became a goldmine of internal infrastructure details. It exposed not only the API credentials (client ID and secret) for their Survey API but also their <strong>LINE Front-end Framework (LIFF) ID</strong>.</p><p>JavaScript</p><pre>// LIFF ID — allows authentication via LINE<br>&quot;VITE_LIFF_ID&quot;: &quot;2006031973-******&quot;</pre><pre>// Survey API credentials <br>&quot;survey&quot;: {<br>    &quot;path&quot;: &quot;cst-as/smbindsjp-line/v1&quot;,<br>    &quot;id&quot;: &quot;4580fb32007a446ba34990cceda154f1&quot;,<br>    &quot;secret&quot;: &quot;fE9f4Cd137Ac4468aF2df864E4056d55&quot;<br>}</pre><p>Exposing a client_id and client_secret in a front-end build is a classic mistake, but without an authorization token, these credentials were just dormant keys. I needed a way to authenticate.</p><h3>The SSO Misconfiguration: Unrestricted LINE Login</h3><p>As someone who spends a lot of time analyzing OAuth misconfigurations and SSO flows, the exposed LIFF ID immediately caught my attention. I navigated to the exposed LIFF URL:</p><p>https://liff.line.me/2006031973-******</p><p>This redirected me to the brand’s test survey endpoint and triggered a LINE Login flow.</p><p>Here was the critical flaw: <strong>There was absolutely no authorization check on who could log in.</strong> The application lacked an allowlist, domain restriction, or role-based access control for the SSO integration. <em>Any</em> user with a standard LINE account — out of their 200+ million user base — could complete the login flow.</p><p>Upon successfully authenticating with my personal LINE account, the application happily issued me a valid Bearer JWT token.</p><h3>The Exploit: Bypassing Access Controls</h3><p>Now armed with a valid JWT Bearer token and the exposed API credentials from the config.js file, it was time to test the internal APIs.</p><p>I fired up Burp Suite and targeted the Survey API endpoint int.api.redacted.com. I crafted a POST request, injecting the leaked Client_id and Client_secret as headers, alongside my newly minted Bearer token:</p><pre>POST /cst-as/smbindsjp-line/v1/lvjint/survey HTTP/2<br>Host: int.api.redacted.com<br>Content-Type: application/json<br>Authorization: Bearer &lt;JWT_TOKEN&gt;<br>Client_id: 4580fb32007a446ba34990cceda154f1<br>Client_secret: fE9f4Cd137Ac4468aF2df864E4056d55<br>Origin: https://admin-int.redacted.com<br>Referer: https://admin-int.redacted.com/</pre><pre>{&quot;FirstName&quot;:&quot;Cyber&quot;,&quot;LastName&quot;:&quot;R&quot;,&quot;Department&quot;:&quot;DIGITAL&quot;,&quot;isPreview&quot;:true,&quot;Consent&quot;:true}</pre><p>The server response? A beautiful HTTP 200 OK.</p><pre>HTTP/2 200 OK<br>Content-Type: text/plain; charset=utf-8<br>X-Ratelimit-Remaining: 499<br>X-Ratelimit-Limit: 500<br>X-Ratelimit-Reset: 556</pre><pre>{&quot;success&quot;: true}</pre><h3>Escalating the Impact across Environments</h3><p>With this access, an attacker could:</p><ul><li><strong>Pollute Survey Data:</strong> The rate limit allowed for 500 requests every ~9 minutes. An attacker could automate this to inject roughly 80,000 fake entries per day.</li><li><strong>Forge GDPR Consent:</strong> Notice the &quot;Consent&quot;:true flag in the payload? The API accepted this without backend validation, meaning an attacker could forge legal consent records.</li><li><strong>Data Disclosure:</strong> After a bit more testing, I found I could change the HTTP method to GET and hit the /list-departments endpoint. This successfully leaked internal department identifiers ([&quot;CDDR&quot;,&quot;DIGITAL&quot;,&quot;NEO&quot;]).</li></ul><p>While testing other microservices listed in the config, I received 401 Unauthorized. This confirmed that while the infrastructure had an authorization middleware, it was completely missing on the survey microservice, a textbook Broken Access Control vulnerability.</p><p><strong>A Systemic Flaw:</strong> The most lucrative part of this discovery was that this wasn’t an isolated incident. By enumerating their subdomains, I discovered that this exact architectural flaw was replicated across multiple environments. The public/config.jsleak and the missing authorization middleware were present in:</p><ul><li><strong>Integration (INT):</strong> admin-int.redacted.com</li><li><strong>Pre-Production (PRP):</strong> admin-prp.redacted.com</li><li><strong>Production (PRD):</strong> admin-prd.redacted.com</li></ul><h3>The Triage and Multi-Environment Payout</h3><p>Initially, the security team had trouble reproducing the bug, returning a 403 Forbidden. I clarified that they specifically needed to authenticate through the LIFF link to generate the initial Bearer token before hitting the internal API.</p><p>Once they followed the exact SSO flow, they reproduced the vulnerability. The bug was originally evaluated at a High CVSS (8.2), but was adjusted to a Medium (6.5) because the impacted data was considered low-sensitivity.</p><p>However, because I was able to demonstrate that the INT, PRP, and PRD environments were all vulnerable via their respective endpoints, <strong>the program treated them as distinct assets and awarded separate bounties for each environment.</strong> By late March, the authorization checks were patched across all environments, fully closing the loop.</p><h3>Key Takeaways for DevSecOps</h3><ul><li><strong>Never hardcode secrets in front-end builds:</strong> If it’s in a .js file served to the client, it&#39;s public.</li><li><strong>Validate your SSO implementations:</strong> Implementing third-party login (OAuth/OIDC/LINE) is only step one. You must enforce authorization checks (like allowlists or domain validation) to ensure only <em>intended</em> users receive administrative tokens.</li><li><strong>Consistent Middleware:</strong> In microservice architectures, ensure your authorization middleware is applied universally. Leaving a single service (like the survey API) unprotected can compromise the integrity of your data.</li><li><strong>Hunt Across Environments:</strong> If you find a misconfiguration in staging or testing, always check if the developers pushed the same flaw to production. It often pays out to look at the bigger picture.</li></ul><p>Until the next bug, happy hunting!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=62546a3e42f1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Exploiting OData Wildcards: How I Scraped Tesla’s Internal Employee Directory for a $2,000 Bounty]]></title>
            <link>https://medium.com/@cyberrado57/exploiting-odata-wildcards-how-i-scraped-teslas-internal-employee-directory-for-a-2-000-bounty-3fe19887f18d?source=rss-b49b58fd445e------2</link>
            <guid isPermaLink="false">https://medium.com/p/3fe19887f18d</guid>
            <category><![CDATA[hackerone]]></category>
            <category><![CDATA[bugcrowd]]></category>
            <category><![CDATA[web-security]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <dc:creator><![CDATA[Cyberrado]]></dc:creator>
            <pubDate>Tue, 31 Mar 2026 08:17:14 GMT</pubDate>
            <atom:updated>2026-03-31T08:17:14.560Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/891/1*BQK2efFxkl9Ww1GKzeF_8g.png" /></figure><p><strong>Introduction</strong><br>Finding a vulnerability in a massive infrastructure like Tesla’s is always an adrenaline rush. While many researchers hunt for complex memory corruptions, sometimes the most impactful bugs hide in a “feature” — specifically, the way an API handles complex queries and internal metadata.</p><p>In this post, I’ll walk through a series of vulnerabilities I discovered on `[REDACTED].tesla.com` that allowed for mass PII enumeration, database schema disclosure, and global business metric leaks.</p><p><strong>The Target: [REDACTED].tesla.com<br></strong>The subdomain `[REDACTED].tesla.com` facilitates coordination between Tesla and its external partners (Bodyshops/Suppliers). Accessing it requires partner-specific provisioning.</p><p>Once inside, I realized the application used a highly flexible, query-based architecture that trustingly exposed internal data to external authenticated users.</p><p>— -</p><p><strong>Finding 1: The `profilengroup` BOLA (PII Leak)<br></strong>The first major issue lived in the user profile look-up feature.</p><p>**Endpoint:**<br>`GET /api/user/profilengroup?take=5&amp;query=like(displayName,’cath%’)`</p><p>The `query` parameter used a syntax similar to OData or a custom “Workforce Query Language” (WQL). By injecting wildcards like `like(displayName,’%’)`, I was able to iterate through the entire internal employee directory.</p><p><strong>The Impact:<br></strong>This wasn’t just a list of names; the API “over-shared” sensitive fields including:<br>- **Personal Mobile Numbers** of internal staff.<br>- **Organizational Hierarchy** (the `directReports` field allowed mapping the entire reporting structure).<br>- **Physical Office Locations** (exact building addresses).<br>- **Internal AD/LDAP Groups** (revealing project team memberships and permission levels).</p><p>— -</p><p><strong>Finding 2: The `manage/dataset` Schema &amp; Admin Leak<br></strong>Beyond user profiles, I found that the application exposed its own architectural blueprint.</p><p>**Endpoint:**<br>`GET /api/manage/dataset?query=space==’bodyshop-eu’&amp;take=50`</p><p>By querying the dataset configuration, I was able to retrieve:<br>1. **Internal Developer Emails**: Fields like `createdBy` and `lastModifiedBy` leaked the `<a href="http://twitter.com/tesla">@tesla</a>.com` emails of the internal admins and developers managing the platform — perfect targets for spear-phishing.<br>2. **Database Schema Documentation**: The response contained the full JSON schema for the backend tables, including hidden fields (e.g., `internalnotes`, `isdeleted`) and validation rules.</p><p>— -</p><p><strong>Finding 3: The `search` Global Count BOLA (Business Intel)<br></strong>Finally, I investigated the search functionality for tickets. While I could only see the details of my own submissions, the API leaked the **Global Total Count**.</p><p>**Endpoint:**<br>`GET /api/search?space=bodyshop-na&amp;category=supplier&amp;take=0&amp;query=*`</p><p>By setting `take=0` and using a wildcard query, the server responded with the total number of records in the system.</p><p><strong>Business Metrics Exposure:<br></strong>- **North America (NA)**: 10,000+ tickets.<br>- **Europe (EU)**: 1,963 tickets.</p><p>— -</p><p><strong>The Bounty &amp; Conclusion<br></strong>Tesla’s security team triaged the findings as a consolidated **P2 (High Impact)** issue due to the broad exposure of PII and internal operational data.</p><p>- **Severity**: P2<br>- **Bounty**: $2,000</p><p><strong>Key Takeaways<br></strong>- **Sanitize Metadata**: Never expose `createdBy` emails or full `jsonSchema` data to external users.<br>- **Enforce Strict Scoping on Counts**: Always filter “Total Count” responses to reflect only the records the user is authorized to see.<br>- **Avoid Wildcard Trust**: If your API supports `LIKE` or `*` operators, ensure they are strictly limited to non-sensitive datasets.</p><p>Happy hunting!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3fe19887f18d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>