<?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 Adini Rodini Rodney on Medium]]></title>
        <description><![CDATA[Stories by Adini Rodini Rodney on Medium]]></description>
        <link>https://medium.com/@papjm?source=rss-b9c7797febe2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*MuAN0_G69pS-7sDbDb2bdw.png</url>
            <title>Stories by Adini Rodini Rodney on Medium</title>
            <link>https://medium.com/@papjm?source=rss-b9c7797febe2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 10:49:31 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@papjm/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[Web Cache Deception: How A GraphQL Query in the Archives Led Me to Cached Profile Data]]></title>
            <link>https://medium.com/@papjm/web-cache-deception-how-a-graphql-query-in-the-archives-led-me-to-cached-profile-data-72fae6e03570?source=rss-b9c7797febe2------2</link>
            <guid isPermaLink="false">https://medium.com/p/72fae6e03570</guid>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[hacking]]></category>
            <dc:creator><![CDATA[Adini Rodini Rodney]]></dc:creator>
            <pubDate>Sat, 17 Oct 2026 10:36:21 GMT</pubDate>
            <atom:updated>2026-10-17T10:41:14.929Z</atom:updated>
            <content:encoded><![CDATA[<p>Recon sometimes reveals behavior that never appears in the application’s normal traffic.</p><p>While testing a target, every GraphQL request I observed used POST. Then I found something different in the web archives: a URL carrying a GraphQL query in its query string.</p><p>That archived request became the starting point for an investigation into cached user data.</p><p><em>The target, routes, operation names, and identifying details below have been replaced with illustrative examples.</em></p><h3>The archived request</h3><p>The URL looked roughly like this:</p><pre>GET /api/query/publicContent?query=...&amp;variables=...<br>Host: example.com</pre><p>I replayed it, and the server accepted it.</p><p>The response contained public information, so there was no sensitive exposure at this stage. But it revealed something useful: although the frontend sent GraphQL requests through POST, the server also accepted queries through GET.</p><p>I wanted to see whether that behavior extended to authenticated operations.</p><h3>Moving from public content to a user profile</h3><p>After registering and signing into my test account, I found a GraphQL operation that returned my profile information.</p><p>A simplified version looked like this:</p><pre>POST /api/query/accountProfile HTTP/1.1<br>Host: example.com<br>Content-Type: application/json<br>Cookie: session=REDACTED</pre><pre>{<br>  &quot;query&quot;: &quot;query AccountProfile($userId: String!) { accountProfile(userId: $userId) { fullName email phoneNumber } }&quot;,<br>  &quot;variables&quot;: {<br>    &quot;userId&quot;: &quot;TEST_ACCOUNT_ID&quot;<br>  }<br>}</pre><p>The actual response included additional personal information, including address and date-of-birth fields.</p><p>Access control appeared to work during my testing. Changing the identifier did not simply let me retrieve another account’s profile.</p><p>Next, I moved the GraphQL query and variables into the URL and changed POST to GET.</p><p>The server still returned my authenticated profile.</p><p>That alone did not demonstrate a vulnerability. The next question was whether the response could enter a shared cache.</p><h3>Adding a file extension</h3><p>I appended .pdf to the operation path:</p><pre>https://example.com/api/query/accountProfile.pdf?query=...&amp;variables=...</pre><p>The endpoint continued returning profile data.</p><p>On the first request, I observed a cache miss. Repeating the request produced a cache hit.</p><p>The important check came afterward: requesting the same URL without my session cookie returned the cached profile response.</p><p>This suggested a mismatch between how the application handled the modified route and how the caching layer treated it. The application still processed a dynamic GraphQL request, while the response appeared eligible for shared caching.</p><p>The .pdf suffix had not turned the response into a document. It was still user-specific JSON.</p><h3>The reproduction limitation</h3><p>My initial report described retrieving the cached response through a private browser window. During triage, however, that browser reproduction stopped working consistently.</p><p>Further testing required a crafted command-line request containing an Origin header. Populating the cache with authenticated profile data also required a valid session cookie.</p><p>The sequence I documented was:</p><ol><li>Send the modified GET request with my test account’s authentication cookie and the required header.</li><li>Observe the profile response being cached.</li><li>Request the same URL without the authentication cookie.</li><li>Retrieve the cached profile data.</li></ol><p>That distinction mattered.</p><p>I had demonstrated the behavior using my own authenticated session. I had not established a reliable way for an attacker to trigger it through an ordinary link click by another user.</p><p>The documented victim scenario required someone to execute a crafted command containing their session cookie. That was a substantial practical limitation.</p><h3>The report outcome</h3><p>The report was closed as <strong>Informative</strong> under a policy excluding issues that required unlikely user interaction.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/857/1*XK7wjEIilxu3GtHxSP7YGw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/999/1*wzCwovdH_qaf6Pn6r0nEkg.png" /></figure><p>The result was not an accepted bounty finding, but the investigation still taught me something useful: demonstrating cached personal data and demonstrating a realistic attack against another user are separate parts of a report.</p><p>Both need clear evidence.</p><h3>What I took away</h3><p>The strongest lead came from comparing archived behavior with the application’s current traffic.</p><p>The frontend used POST. An archived URL showed that GET was also accepted. Testing that behavior against an authenticated operation then exposed an unexpected interaction with the cache.</p><p>For future investigations, I would establish the victim-triggering conditions earlier and document exactly which headers, cookies, and browser behaviors are necessary.</p><p>I would also keep the demonstrated impact focused on the profile data actually reproduced, rather than extending the claim to other sensitive operations without equivalent evidence.</p><p>A cache hit is a clue. Retrieving another session’s response is stronger evidence. A realistic way to cause that response to be cached is what completes the attack story.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=72fae6e03570" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[An Email Address Was Enough: Two Account Deletion Findings Worth $850]]></title>
            <link>https://medium.com/@papjm/an-email-address-was-enough-two-account-deletion-findings-worth-850-89037f87dbc0?source=rss-b9c7797febe2------2</link>
            <guid isPermaLink="false">https://medium.com/p/89037f87dbc0</guid>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[bug-bounty-writeup]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <dc:creator><![CDATA[Adini Rodini Rodney]]></dc:creator>
            <pubDate>Thu, 15 Oct 2026 10:55:53 GMT</pubDate>
            <atom:updated>2026-10-15T10:55:53.419Z</atom:updated>
            <content:encoded><![CDATA[<h3>Testing what happens when customer support becomes part of the authorization process</h3><p>Account deletion testing usually starts with an endpoint.</p><p>Can I change the account ID? Does the application check ownership? Can a request from another session delete my account?</p><p>But there is another route worth examining: customer support.</p><p>Across two bug bounty programs, I found that support workflows allowed someone to request deletion of an account without proving ownership. Both findings involved accounts I controlled, and they earned me a combined <strong>$850</strong>.</p><p>The companies, domains, email addresses, and staff identities in this story have been redacted.</p><h3>Finding One: A Deletion Request From a Different Email Address</h3><p>For the first test, I created two accounts and completed their registration procedures.</p><p>I then sent a message to customer support from the email address associated with Account A, asking them to delete Account B.</p><p>The request was straightforward:</p><blockquote><em>Hello, I would like to close my account registered under </em><a href="/service/mailto:account-b@example.com"><em>account-b@example.com</em></a><em>.</em></blockquote><p>The important detail was the mismatch: <strong>the sender’s address was different from the address of the account being deleted.</strong></p><p>Support replied to the requesting address. After further processing, the deletion procedure began, and Account B was eventually deleted.</p><p>The missing check was ownership verification. Before acting, support had not required confirmation through Account B’s registered email address or an authenticated session belonging to that account.</p><p>I repeated the test using another account I controlled and supplied screenshots and a recording with the report.</p><p>The program acknowledged the issue and said it was reviewing its account deletion process. It assessed the finding as medium severity, with a CVSS score of 4.5.</p><h3>Finding Two: Separate Contact and Account Email Fields</h3><p>The second finding involved a support form with separate fields for:</p><ul><li>The email address where support should contact the requester.</li><li>Confirmation of that contact address.</li><li>The email address associated with the account, if different.</li></ul><p>That separation can serve legitimate purposes. The security question is what happens before support carries out a destructive action.</p><p>I entered an email address I controlled as the contact address and the address of a second test account as the account to be deleted.</p><p>Support sent its correspondence to the contact address I supplied. I continued the conversation from that address, and support ultimately deleted the second account.</p><p>A notification reached the deleted account’s registered address afterward.</p><p><strong>A notification after deletion does not provide authorization before deletion.</strong></p><p>The workflow established where to contact the requester, but it did not adequately establish that the requester controlled the account being deleted.</p><h3>The Shared Root Cause</h3><p>Both findings crossed the same trust boundary:</p><p><strong>Knowing an account’s email address was treated as sufficient authority to request its deletion.</strong></p><p>An email address identifies an account. It does not prove ownership.</p><p>Authorization checks need to cover every route to a sensitive action, including actions performed manually by support staff. A protected account deletion endpoint cannot compensate for a support process that performs the same action without equivalent verification.</p><h3>How I Would Test This Again</h3><p>Only test within a program’s rules and use accounts you own. Account deletion is destructive, so keep disposable test accounts separate from anything you need.</p><p>A useful starting point is to create Account A and Account B with the same display name and similar email addresses. This can help reveal whether support relies on superficial similarities when deciding who owns an account.</p><p>However, similar addresses can also introduce confusion. Repeat the test with clearly different addresses to establish whether the underlying problem is missing ownership verification.</p><p>Then:</p><ol><li>Complete registration for both accounts.</li><li>Contact support from Account A and request deletion of Account B.</li><li>Record where support sends its replies and any confirmation requests.</li><li>Check whether Account B must authorize the deletion before it happens.</li><li>Verify the final account state and preserve a minimal evidence trail.</li></ol><p>Be precise about the result. A support acknowledgment, a scheduled deletion, and a completed deletion are different outcomes.</p><p>Keep the initial request simple and truthful. You do not need an elaborate impersonation story to demonstrate a missing authorization check.</p><p>Also check the program’s policy before contacting support: some programs restrict social engineering or support interactions.</p><h3>What a Strong Report Should Show</h3><p>The clearest evidence is the relationship between three things:</p><ul><li>The address that submitted the request.</li><li>The address registered to the account being deleted.</li><li>The verification performed before deletion.</li></ul><p>Include the relevant correspondence and evidence of the final account state. Redact unrelated personal information.</p><p>Keep impact claims tied to what you demonstrated. In my tests, I showed that I could cause deletion of another account I controlled without confirming the request through that account.</p><p>That established an account availability and data integrity risk. It did not demonstrate access to the account’s confidential information.</p><h3>Fixing the Workflow</h3><p>Before deleting an account, support should require confirmation through a trusted channel already associated with that account, such as its registered email address or an authenticated session.</p><p>If the requester cannot access that channel, the request should go through a defined account recovery process. Requests from authorized representatives also need an appropriate verification process.</p><p>A requester-supplied contact address should not become proof of account ownership.</p><h3>What These Findings Changed About My Testing</h3><p>These two reports earned me a combined <strong>$850</strong>, but the broader lesson changed where I look for authorization flaws.</p><p>I now ask: who else can perform the same sensitive action, and how do they verify that the requester is allowed to ask for it?</p><p>Sometimes the missing authorization check is in a support workflow.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X5czrpJwHBQBj31ZQClnvA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*m1-BAHmcGarxzlwTGly60A.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=89037f87dbc0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A 403 Bypass Worth $900: How a Semicolon Exposed API Documentation]]></title>
            <link>https://medium.com/@papjm/a-403-bypass-worth-900-how-a-semicolon-exposed-api-documentation-698eecaf9308?source=rss-b9c7797febe2------2</link>
            <guid isPermaLink="false">https://medium.com/p/698eecaf9308</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[bug-bounty-writeup]]></category>
            <category><![CDATA[hacking]]></category>
            <dc:creator><![CDATA[Adini Rodini Rodney]]></dc:creator>
            <pubDate>Sun, 04 Oct 2026 10:16:06 GMT</pubDate>
            <atom:updated>2026-10-04T10:16:06.046Z</atom:updated>
            <content:encoded><![CDATA[<h3>A 403 Bypass Worth $900: How a Semicolon Exposed API Documentation</h3><p>A single semicolon turned a Cloudflare 403 Forbidden response into access to an API documentation page. When the original bypass stopped working, a different semicolon placement opened the same page again.</p><p>Two accepted reports earned me <strong>$450 each — a total of $900</strong>.</p><p>This started during reconnaissance in 2023 and continued through follow-up testing in 2025. Here is how I found the issue, bypassed the initial fix, and learned to think beyond one blocked URL.</p><p><em>The target has been anonymized. All target URLs use </em><em>example.com, and the paths and query parameters below are invented replacements that preserve the structure of the technique.</em></p><h3>Finding a route behind the 403</h3><p>During recon, I discovered an API subdomain. Visiting its root gave me a Cloudflare 403 Forbidden response.</p><p>Google searches did not give me useful leads, so I tried Yahoo instead. That search returned an indexed API URL with a structure like this:</p><pre>https://example.com/service/v4/catalog/en/items/482901/feedback?limit=10&amp;page=1&amp;sort=recent&amp;direction=desc</pre><p>That result revealed a useful starting point: the application had routes under /service/.</p><p>I visited:</p><pre>https://example.com/service</pre><p>The page offered an <strong>Interactive REST API documentation</strong> link. Following it led to a Swagger UI page, but Cloudflare blocked the request:</p><pre>https://example.com/service/v4/reference/start.html</pre><p>I now had a specific endpoint to investigate.</p><h3>The first bypass: one extra character</h3><p>I added a semicolon immediately after the version segment:</p><pre>https://example.com/service/v4;/reference/start.html</pre><p>The response changed to <strong>200 OK</strong>, and the API documentation loaded.</p><p>The difference was small:</p><p>Request pathObserved result/service/v4/reference/start.htmlCloudflare 403/service/v4;/reference/start.html200 OK; documentation accessible</p><p>The useful evidence was the content of the response. I could access the documentation that the original request had been prevented from reaching.</p><p>I reported the behavior. After triage and program review, the finding was rated <strong>Medium, with a CVSS score of 5.3</strong>, and I received a <strong>$450 bounty</strong>.</p><h3>Returning after the fix</h3><p>In 2025, the original semicolon payload no longer worked. Both the normal URL and the first bypass returned a 403.</p><p>I tested another variation, placing a semicolon with a name-value pair in a later part of the path:</p><pre>https://example.com/service/v4/reference/;foo=bar/start.html</pre><p>That variation reached the documentation again.</p><p>Request pathResult during follow-up testing/service/v4/reference/start.html403 Forbidden/service/v4;/reference/start.html403 Forbidden/service/v4/reference/;foo=bar/start.htmlDocumentation accessible</p><p>I submitted a second report demonstrating the new bypass. The team accepted it and awarded another <strong>$450</strong>, bringing the total to <strong>$900</strong>.</p><p>During later retesting, I also demonstrated these placements:</p><pre>https://example.com/service/v4/reference;/start.html<br>https://example.com/service/v4/reference/;/start.html</pre><p>The original report was reopened after the team reproduced the behavior. They explained that additional reports were unnecessary for those variants because the underlying fix was the same.</p><h3>Why a semicolon could matter</h3><p>The behavior suggested that the security layer and the application interpreted the path differently.</p><p>Semicolons can have meaning inside URL path segments. Some frameworks support semicolon-delimited path parameters, often called <strong>matrix variables</strong>. Spring’s documentation provides examples of this syntax, including parameters attached to individual path segments.</p><p>A plausible explanation is that the blocking rule matched one representation of the URL while another component still routed the modified path to the documentation.</p><p>I did not establish the target’s exact parsing behavior or confirm its framework. The observed result was that semicolon variations reached a resource whose ordinary path was blocked.</p><p>The lesson is to consider how each layer interprets the request: the WAF, any proxy, and the application router.</p><h3>Keeping the impact precise</h3><p>My demonstrated impact was <strong>bypassing the Cloudflare restriction protecting the API documentation page</strong>.</p><p>A reachable Swagger UI can help someone understand an API’s routes, parameters, and request formats. It does not automatically prove that protected API operations can be called without authorization.</p><p>I also noticed login pages and raised a separate concern about rate limiting. The triage team asked me to report that issue separately, so it is outside the finding described here.</p><p>Looking back, I would keep the evidence focused on what I could demonstrate. Access to the documentation did not prove access to configuration files, source repositories, user accounts, or sensitive records.</p><h3>What I learned</h3><p>Search engines can reveal useful routes even when a host’s root returns a 403. Trying Yahoo after Google produced no leads was what gave me a route to follow.</p><p>Retesting also needs to go beyond the original payload. Blocking one semicolon placement did not prevent other placements from reaching the same resource.</p><p>For developers, this points to consistent path handling across security and routing layers, alongside authentication and authorization on resources that require restricted access. If semicolon path parameters are unnecessary, rejecting them consistently can reduce ambiguity.</p><p>For me, the most useful part of this finding was learning to compare <strong>what a security rule blocks with what the application still accepts</strong>. That small difference led to two accepted reports and $900 in bounties.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2Jrm9ADYM66--lP_117UfA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QlaYcuL8sESxvfg3JTECBQ.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=698eecaf9308" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Changing One Customer ID Exposed Other Users’ Pet Recommendations and Earned Me a $1,000 Bounty]]></title>
            <link>https://medium.com/@papjm/how-changing-one-customer-id-exposed-other-users-pet-recommendations-and-earned-me-a-1-000-bounty-82c0179e62aa?source=rss-b9c7797febe2------2</link>
            <guid isPermaLink="false">https://medium.com/p/82c0179e62aa</guid>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <dc:creator><![CDATA[Adini Rodini Rodney]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 16:12:01 GMT</pubDate>
            <atom:updated>2026-10-01T16:12:01.351Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Researcher:</strong> papjm<br><strong>Vulnerabilities:</strong> Broken Access Control and Information Disclosure<br><strong>Reward:</strong> $1,000</p><blockquote><em>The company, hostname, API path, customer identifiers, database objects and triager identity have been redacted. No SQL injection was confirmed.</em></blockquote><h3>Summary</h3><p>I discovered two related vulnerabilities in an online retailer’s product-recommendation API.</p><p>First, changing a user-controlled customer_id parameter caused the API to return pet-product recommendations belonging to another customer.</p><p>Second, a Base64-encoded tracking_id in the response exposed internal SQL queries, database table names and column names.</p><h3>Discovery</h3><p>I signed in to my account and opened the account section while Burp Suite captured the application’s traffic.</p><p>In Burp’s HTTP history, I found a recommendation request containing several parameters:</p><pre>GET /[REDACTED]?<br>    category_id=[CATEGORY_ID]&amp;<br>    user_id=[USER_IDENTIFIER]&amp;<br>    customer_id=[MY_CUSTOMER_ID]</pre><p>I sent the request to Burp Repeater and changed only the customer_id to a nearby value:</p><pre>customer_id=[ANOTHER_CUSTOMER_ID]</pre><p>My authenticated session and the other parameters remained unchanged.</p><p>The server returned 200 OK with a different set of pet-product recommendations associated with the supplied customer ID.</p><p>This confirmed that the server trusted the customer ID from the URL without verifying that it belonged to the authenticated user.</p><p>If customer IDs were predictable or obtainable elsewhere, an attacker could potentially automate the process and enumerate recommendations belonging to multiple customers.</p><h3>The Base64-Encoded Tracking ID</h3><p>While examining the response, I noticed a long value inside a field named tracking_id.</p><p>The format and padding suggested that it was Base64-encoded. Base64 is encoding — not encryption — so I decoded it.</p><p>The decoded value contained internal SQL filters similar to this sanitized example:</p><pre>{<br>  &quot;filters&quot;: [<br>    {<br>      &quot;query&quot;: &quot;SELECT 1 FROM [INTERNAL_TABLE] WHERE [INTERNAL_CONDITIONS]&quot;,<br>      &quot;operation&quot;: &quot;in&quot;<br>    }<br>  ]<br>}</pre><p>The data exposed internal SQL query templates, database table names and column names used by the recommendation engine.</p><p>This did not prove that the application was vulnerable to SQL injection. However, the information could help an attacker understand the database structure and prepare more targeted attacks if another injection vulnerability existed.</p><h3>Impact</h3><p>An attacker with a normal account could potentially:</p><ul><li>Access pet-product recommendations associated with other customers.</li><li>Enumerate recommendation data by changing customer IDs.</li><li>Learn information about another customer’s pet-related interests.</li><li>Recover internal SQL query templates from the tracking value.</li><li>Identify database tables and columns that could support further reconnaissance.</li></ul><p>Product recommendations may not appear highly sensitive, but this was still a cross-account authorization failure. A user should never be able to retrieve another customer’s personalized data by changing a URL parameter.</p><h3>Root Cause</h3><p>The server accepted a client-controlled customer_id without confirming that it belonged to the authenticated session.</p><p>The client also received internal SQL information that was unnecessary for displaying recommendations.</p><h3>Recommended Remediation</h3><p>The application should:</p><ol><li>Derive the customer identity from the authenticated session.</li><li>Enforce object-level authorization before returning personalized data.</li><li>Reject requests for customer IDs that do not belong to the authenticated user.</li><li>Remove SQL queries and database details from client-visible responses.</li><li>Replace the encoded SQL data with an opaque server-generated tracking identifier.</li><li>Monitor repeated requests involving different customer IDs.</li></ol><h3>Disclosure and Outcome</h3><p>I reported the vulnerabilities through the company’s bug-bounty program.</p><p>The security team reviewed the evidence, triaged the report and awarded me a <strong>$1,000 bounty</strong> on February 9, 2024.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YLCKm9HFANONahKgeeKJig.png" /></figure><h3>Takeaway</h3><p>Whenever a request contains multiple identity-related parameters, change them one at a time and compare the responses carefully.</p><p>Also inspect values named tracking, metadata, debug, state or context. Data that appears opaque may simply be encoded.</p><p>The most important lesson is that personalized data must be selected using a server-verified identity — not an identifier controlled by the client.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=82c0179e62aa" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How a Predictable Activity ID Bypassed an Unguessable Invite Link and Earned Me a $1,000 Bounty]]></title>
            <link>https://medium.com/@papjm/how-a-predictable-activity-id-bypassed-an-unguessable-invite-link-and-earned-me-a-1-000-bounty-59b0e8b30016?source=rss-b9c7797febe2------2</link>
            <guid isPermaLink="false">https://medium.com/p/59b0e8b30016</guid>
            <category><![CDATA[hackerone]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[idor-vulnerability]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Adini Rodini Rodney]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 15:45:13 GMT</pubDate>
            <atom:updated>2026-10-01T15:49:00.487Z</atom:updated>
            <content:encoded><![CDATA[<h3>How a Predictable Activity ID Bypassed an Unguessable Invite Link and Earned Me a $1,000 Bounty</h3><p><strong>Researcher:</strong> papjm<br><strong>Vulnerability:</strong> Insecure Direct Object Reference (IDOR)<br><strong>Initial bounty:</strong> $1,000<br><strong>Total rewards including retests:</strong> $1,200</p><p><em>The company, application, hostname, API paths, usernames, and identifying values have been redacted.</em></p><h3>Summary</h3><p>I discovered an IDOR vulnerability in an activity-planning application that allowed an uninvited user to access private activity details and retrieve the activity’s invitation link.</p><p>The application protected private activities using long, unguessable invitation links. An activity creator was expected to share this link only with people permitted to participate.</p><p>However, the backend also identified each activity using a short, sequential numeric ID. This identifier appeared inside a delete request and could be used with a GET request to retrieve the private activity.</p><p>The server did not verify whether the requesting user had created or been invited to the activity.</p><h3>How Private Activities Were Supposed to Work</h3><p>When creating an activity, the organizer could set its visibility to private.</p><p>A private activity was not supposed to be discoverable by other users. Instead, the organizer received a long, unguessable invitation link that had to be shared directly with the intended participants.</p><p>The expected access model was therefore:</p><ol><li>The organizer creates a private activity.</li><li>The application generates an unguessable invitation link.</li><li>The organizer sends the link to selected participants.</li><li>Only users who receive the link can access the private activity.</li></ol><p>At first, this appeared secure because guessing the complete invitation link would have been extremely difficult.</p><p>The problem was that the same activity was represented internally by a much shorter and predictable numeric ID.</p><h3>Step-by-Step Discovery</h3><h3>Step 1: Create a Private Activity</h3><p>I signed in using my first account and started creating a new activity.</p><p>I completed the creation process by:</p><ol><li>Selecting myself as the activity organizer.</li><li>Choosing a destination.</li><li>Continuing without adding a route.</li><li>Entering the required activity information.</li><li>Changing the visibility from public to private.</li><li>Finishing the creation process.</li></ol><p>The application created the private activity and generated an unguessable invitation link for it.</p><p>At this point, another user should not have been able to access the activity unless I deliberately shared that link with them.</p><h3>Step 2: Discover the Internal Activity ID</h3><p>I opened the private activity and selected the option to delete it while monitoring the application’s network traffic through Burp Suite.</p><p>The application generated a request similar to:</p><pre>DELETE /[REDACTED]/[ACTIVITY_ID]</pre><p>The exact hostname and API path have been removed from this write-up.</p><p>The important part of the request was the ACTIVITY_ID. Instead of using the long invitation token, the backend referenced the activity using a short numeric identifier.</p><p>I recorded the identifier and dropped the delete request so that the activity would remain active.</p><h3>Step 3: Test the Identifier Using Another Account</h3><p>I signed out of the organizer’s account and signed in using a second account.</p><p>This second account:</p><ul><li>Did not create the private activity.</li><li>Had not been invited to it.</li><li>Did not possess the unguessable invitation link.</li><li>Should not have been permitted to view it.</li></ul><p>Using the second account, I took the endpoint discovered through the delete request and changed the HTTP method from DELETE to GET:</p><pre>GET /[REDACTED]/[ACTIVITY_ID]</pre><p>The server returned:</p><pre>HTTP/2 200 OK</pre><p>The response contained the private activity’s information, including its invitation link.</p><p>This meant that an unauthorized user could bypass the long, unguessable invitation link entirely. The user only needed the much shorter internal numeric identifier.</p><h3>Step 4: Test for Enumeration</h3><p>I then changed the numeric activity ID to nearby values.</p><p>Because the identifiers were sequential, requesting different numbers returned information about other private activities.</p><p>For example, if one valid activity used an identifier such as:</p><pre>52796</pre><p>an attacker could test nearby values by incrementing or decrementing the number.</p><p>This process could be automated using a tool such as Burp Intruder. Requests that returned 200 OK indicated valid activities.</p><p>The attacker did not need to guess the unguessable invitation links. The vulnerable API returned those links after the attacker supplied a valid numeric activity ID.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JV5h5_oE5J9nYnHiG77C6w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/233/1*170nANjCPChpSYAZwTiVuw.png" /></figure><h3>Why This Was Vulnerable</h3><p>The application relied on the secrecy of the invitation link to protect private activities.</p><p>However, the backend API did not enforce the same access restrictions.</p><p>When it received an activity ID, the server should have checked whether the requesting user was:</p><ul><li>The activity creator.</li><li>An invited participant.</li><li>An authorized member of the activity.</li><li>Otherwise permitted to access the requested object.</li></ul><p>Instead, the server returned the object solely because the supplied numeric ID existed.</p><p>This was a classic IDOR caused by missing object-level authorization.</p><h3>Impact</h3><p>An attacker with an ordinary account could potentially:</p><ul><li>Access private activities without receiving an invitation.</li><li>View information belonging to private activities.</li><li>Retrieve supposedly unguessable invitation links.</li><li>Share those invitation links with additional unauthorized users.</li><li>Enumerate other private activities using sequential IDs.</li><li>Automate the collection of private activity information.</li></ul><p>The unguessable invitation link provided little protection because a separate API endpoint exposed the same resource through a predictable identifier.</p><h3>Disclosure and Resolution</h3><p>I reported the vulnerability in October 2023.</p><p>The affected API was primarily associated with the company’s mobile application, even though it was also accessible through a web interface. The company accepted the report and awarded me a <strong>$1,000 bounty</strong>.</p><p>In June 2025, I was invited to retest the vulnerability. I repeated the original process and confirmed that the issue was still reproducible. I received an additional <strong>$50 retest reward</strong>.</p><p>In June 2026, the company contacted me for another retest after attempting to block access to the vulnerable endpoint.</p><p>My initial test showed that the vulnerability still worked. The company then discovered that the affected infrastructure existed in multiple regions and that each region needed to be reconfigured.</p><p>After the remaining infrastructure was updated, I tested the endpoint again. Unauthorized requests now returned:</p><pre>HTTP/2 403 Forbidden</pre><p>The private activity details and invitation link were no longer accessible. The company resolved the report and awarded me another <strong>$150 retest reward</strong>.</p><h3>Lessons Learned</h3><p>A long, unguessable link does not protect a resource if the same resource is available through another endpoint using a predictable identifier.</p><p>Whenever an application places an object ID inside a request:</p><ol><li>Test the identifier from a different account.</li><li>Test whether the object is private or user-specific.</li><li>Try other supported HTTP methods.</li><li>Check whether the identifiers are sequential.</li><li>Confirm that the server validates ownership or membership on every request.</li></ol><p>Identifiers are references — not authorization controls. Every request for a private object must be protected by a server-side permission check.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=59b0e8b30016" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How Changing a POST Request to GET Exposed Users’ PII and Earned Me a $5,000 Bounty]]></title>
            <link>https://medium.com/@papjm/how-changing-a-post-request-to-get-exposed-users-pii-and-earned-me-a-5-000-bounty-a9980c5b7df0?source=rss-b9c7797febe2------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9980c5b7df0</guid>
            <category><![CDATA[hackerone]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Adini Rodini Rodney]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 15:17:28 GMT</pubDate>
            <atom:updated>2026-10-01T15:17:28.139Z</atom:updated>
            <content:encoded><![CDATA[<h3>How Changing a POST Request to GET Exposed Users’ Privacy Requests</h3><p><strong>Researcher:</strong> papjm<br><strong>Severity:</strong> Critical<br><strong>Reward:</strong> $5,000 bounty + $150 retest</p><p><em>The company, third-party provider, hostname, and API paths have been intentionally redacted.</em></p><h3>Summary</h3><p>I discovered a broken access-control vulnerability in a company’s privacy-request portal. By changing a legitimate POST request to GET and removing the request body, I could retrieve privacy requests belonging to other users.</p><p>The exposed records contained sensitive personally identifiable information, including contact details submitted by people making deletion and other privacy-related requests.</p><h3>Discovery</h3><p>I started by submitting a deletion request using my own information. While reviewing the network traffic, I noticed a request being sent to the platform’s privacy-request API.</p><p>I forwarded the request to Burp Repeater, removed the POST body, and changed the HTTP method to GET.</p><p>Instead of rejecting the request, the server returned records belonging to other users without verifying that I was authorized to access them.</p><p>I then reviewed the platform’s public API documentation and found supported filtering and pagination parameters. These parameters made it possible to request specific categories of records and move through multiple pages of results.</p><p>An attacker could therefore automate the process and retrieve a large amount of sensitive user information.</p><h3>Impact</h3><p>The vulnerability could allow an unauthorized attacker to:</p><ul><li>Access privacy requests submitted by other users.</li><li>Retrieve sensitive personal and contact information.</li><li>Filter records by their source or category.</li><li>Enumerate and download records in bulk using pagination.</li></ul><p>Because the affected system processed privacy-related requests, the exposed information was particularly sensitive.</p><h3>Disclosure and Resolution</h3><p>I reported the vulnerability on June 17, 2025. The affected company independently reproduced the issue and determined that the vulnerability existed within a third-party service used to manage privacy requests.</p><p>The company instructed me to stop further testing while it coordinated remediation with the service provider.</p><p>The authorization controls were patched within two days. I was awarded a <strong>$5,000 Critical-severity bounty</strong> and invited to perform a paid retest.</p><p>During the retest, I followed the original reproduction process and confirmed that unauthorized access was no longer possible. The report was subsequently resolved, and I received an additional <strong>$150 retest reward</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jZNpN3URq10XEkJDAysyZg.png" /></figure><h3>Lesson</h3><p>Never assume an API endpoint is secure simply because the application normally accesses it using a specific HTTP method. Testing alternative methods, removing request bodies, and reviewing documented filtering or pagination features can expose serious authorization weaknesses hidden behind an otherwise normal workflow.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9980c5b7df0" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>