<?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 Riya Limba on Medium]]></title>
        <description><![CDATA[Stories by Riya Limba on Medium]]></description>
        <link>https://medium.com/@riyalimba?source=rss-7e38590ffb2f------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*XxjbYaTB2kEjIM2xm0XSVA.jpeg</url>
            <title>Stories by Riya Limba on Medium</title>
            <link>https://medium.com/@riyalimba?source=rss-7e38590ffb2f------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 11:07:30 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@riyalimba/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[HackerOne’s Data Says AI Reports Are Getting Better. So Why Do Mine Keep Getting Closed?]]></title>
            <link>https://medium.com/@riyalimba/hackerones-data-says-ai-reports-are-getting-better-so-why-do-mine-keep-getting-closed-0dde2b22fe33?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/0dde2b22fe33</guid>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[osint]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Tue, 27 Oct 2026 16:58:08 GMT</pubDate>
            <atom:updated>2026-10-27T16:58:08.456Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pjo6JrowncInT_ixH3X7xA.png" /></figure><h3>60% of security leaders say AI-assisted submissions improved in quality. I had to figure out what they were doing that I wasn’t.</h3><p>I read HackerOne’s annual report on a Wednesday and found a number that contradicted everything I’d been writing for months.</p><p>60% of security leaders said the quality of AI-assisted submissions is improving .</p><p>I sat with that for a long time.</p><p>Because I’d been writing about AI slop. About Google pausing its OSS VRP because automated submissions were “not valid.” About curl ending its bounty program because confirmed vulnerabilities fell below 5%. About Intel suspending its program entirely.</p><p>Those stories are real. The slop is real.</p><p>But the HackerOne data says something else is also real: a majority of the AI-assisted reports landing in triage queues are getting better, not worse.</p><p>Which means the problem isn’t AI. It’s how I’ve been using it.</p><h3>What the Report Actually Says</h3><p>The HackerOne report includes two numbers that sit side by side.</p><p>53% of security leaders report a higher volume of AI-assisted submissions. That’s the flood everyone talks about.</p><p>60% report better quality .</p><p>Read that again. More reports, and better ones.</p><p>The report also found that 85% of researchers are actively developing AI skills, and nearly three-quarters said AI meaningfully increased the valid findings they produce .</p><p>A top-earning researcher put it plainly: <em>“AI hasn’t replaced how I think as a security researcher. It’s changed how much ground I can cover”</em> .</p><p>I’d been reading that quote as inspiration. Now I read it as a diagnosis.</p><p>The researchers getting value from AI aren’t using it to generate reports. They’re using it to cover more ground, then applying their own judgment to what it finds.</p><p>I’d been doing the opposite. I’d been using AI to generate the report, then hoping the triager would do the judgment for me.</p><h3>What I Was Doing Wrong</h3><p>I used AI to help me find bugs. That part wasn’t wrong.</p><p>But when it came time to write the report, I let the AI do the thinking.</p><p>It wrote the summary. It described the impact. It suggested the severity. I copied it into the form, added a screenshot, and hit submit.</p><p>The report looked polished. It sounded technical. It used words like “exploitability” and “attack vector.”</p><p>But when the triager read it, they couldn’t reproduce the bug. Because I hadn’t actually verified that the bug was real. The AI said it was. I believed it.</p><p>The 60% of security leaders seeing better quality aren’t seeing better AI-generated reports. They’re seeing reports from researchers who used AI to find things, then verified those things themselves before writing a single word.</p><h3>What the Numbers Actually Mean</h3><p>The HackerOne report also found that 68% of researchers have shifted toward more complex bugs with higher bounties .</p><p>That’s the signal. The researchers who are succeeding aren’t submitting more reports. They’re submitting better ones. They’re using AI to handle the tedious parts — reading documentation, scanning large codebases, identifying patterns — and then spending their own time on the parts AI can’t do: understanding why something is broken and proving it.</p><p>I’d been using AI to skip the understanding. That’s why my reports kept getting closed.</p><h3>What I’m Doing Differently</h3><p>I’m not going to stop using AI. That would be stupid. The researchers who succeed in the next few years will be the ones who use it well.</p><p>But I’m changing when I use it.</p><p>First, AI finds. I verify. If the AI says an endpoint is vulnerable, I test it myself. I run the request. I check the response. I confirm the behavior. If I can’t reproduce it with my own hands, it’s not a finding. It’s a guess.</p><p>Second, AI explains. I write. I’ll ask the AI to help me understand a code path or a configuration. But I write the report myself, in my own words, with my own proof. Because the triager needs to see that a human understands what they’re reading.</p><p>Third, I’m measuring myself by quality, not volume. The old me submitted 47 reports in three months. The new me submits fewer, but each one is verified before it leaves my hands. A report that gets closed as “informative” is worse than a report I never submitted, because it wasted someone’s time.</p><h3>The Uncomfortable Truth</h3><p>The HackerOne data says AI reports are getting better. That means the researchers using AI well are pulling ahead. The ones using it badly are getting closed out.</p><p>I’ve been in the second group.</p><p>I don’t have a CVE. I’ve found one confirmed bug. I’m still learning.</p><p>But I know what the 60% figure means now. It means the problem was never AI. The problem was me using it as a shortcut.</p><p>Shortcuts don’t work in bug bounty. Verification does.</p><p><em>If you’re also learning to use AI without letting it do the thinking for you, I write about what I’m actually figuring out — mistakes included. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0dde2b22fe33" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Tracked Down What Happened to the Bug I Reported. It Wasn’t Fixed.]]></title>
            <link>https://medium.com/@riyalimba/i-tracked-down-what-happened-to-the-bug-i-reported-it-wasnt-fixed-c20b95066af9?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/c20b95066af9</guid>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[osint]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Sat, 24 Oct 2026 14:03:13 GMT</pubDate>
            <atom:updated>2026-10-24T14:03:13.633Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zLx67_vx0Y2NNwVxjK_GRQ.png" /></figure><h3>HackerOne paid a record $89 million this year. The backlog of unfixed bugs grew 131%. I think I know why.</h3><p>I found my first confirmed bug four months ago.</p><p>It was a broken access control issue. Nothing dramatic — an API endpoint that returned more data than it should have. I wrote the report, included the raw requests, explained the impact in one sentence. The triager confirmed it. Paid me.</p><p>I felt good about that. I thought I had contributed something. Made a system a little safer.</p><p>Then I read HackerOne’s annual report last week and found a number I couldn’t stop thinking about.</p><p>Exposure debt rose 131% .</p><p>Not technical debt. Not financial debt. Exposure debt — the accumulated pile of vulnerabilities that organizations know about but haven’t fixed yet.</p><p>I went back and checked the status of my report. It was closed. Paid. Resolved.</p><p>But the bug? I don’t know if it was ever actually patched. The report was “resolved” from the platform’s perspective. That doesn’t mean the vulnerability was fixed.</p><h3>What Exposure Debt Actually Means</h3><p>I had to read the report twice to understand the mechanics.</p><p>HackerOne says average resolution time improved by 54% — from 135 days down to 62 days . Organizations are fixing individual bugs faster than ever.</p><p>But the backlog still grew. Because new findings are arriving faster than teams can clear them.</p><p>Nearly 70% of security leaders confirmed this: validated findings are entering their systems faster than their teams can remediate them .</p><p>A team can get better at fixing bugs and still watch the pile grow. That’s the trap. The drain is wider than it’s ever been. The faucet is running at a rate the drain was never designed for.</p><h3>Why This Matters to Me Specifically</h3><p>I’ve found one confirmed bug. I’m waiting to be paid on a second.</p><p>Reading about exposure debt, I felt something shift.</p><p>The question I’d been asking was: <em>How do I find more bugs?</em></p><p>The better question — the one the data points to — is: What happens to the bugs I find after I submit them?</p><p>If my report enters a backlog and never gets fixed, I got paid. The triager did their job. But the vulnerability is still there. The data is still exposed. The system is still vulnerable.</p><p>I contributed to someone’s backlog. Not to their security.</p><h3>What the Report Says About Researchers Like Me</h3><p>The HackerOne report includes data that feels personal.</p><p>85% of surveyed researchers are actively developing AI skills. Nearly three-quarters said AI meaningfully increased the valid findings they produce .</p><p>But 68% said they’ve shifted toward more complex bugs with higher bounties .</p><p>That’s the signal. The researchers who are succeeding are going deeper, not wider. They’re using AI to cover more ground, then focusing human judgment on the bugs that require creativity and context.</p><p>A top-earning researcher put it plainly: <em>“AI hasn’t replaced how I think as a security researcher. It’s changed how much ground I can cover”</em> .</p><h3>What I’m Doing Differently</h3><p>I can’t fix the remediation crisis. I can’t force organizations to patch faster. But I can change three things.</p><p>First, I’m treating “reproducible” as a higher bar than “valid.” A valid bug that takes an hour to reproduce is less useful than a valid bug with a one-line PoC. The easier I make it for a developer to fix, the more likely it actually gets fixed.</p><p>Second, I’m looking at remediation history, not just bounty size. A program that pays well but never patches is accumulating debt. A program with lower payouts but active patching is contributing to actual security. HackerOne’s CEO said it directly: pausing programs may ease volume, but it loses the work of researchers finding real issues .</p><p>Third, I’m remembering that the bottleneck isn’t me. The report says nearly 70% of security leaders confirmed that validated findings are arriving faster than teams can fix them . My job isn’t just to find bugs. It’s to make the fix as easy as possible.</p><h3>The Uncomfortable Truth</h3><p>HackerOne paid a record $89 million to researchers this year. That’s genuinely good news. The work is valued. Companies are still investing in finding bugs before attackers do .</p><p>But the 131% exposure debt increase tells a different story. It says the system is finding faster than it’s fixing. It says the gap is growing.</p><p>The bugs I find aren’t just entries in my report history. They’re entries in someone’s backlog. And if that backlog keeps growing, the whole system becomes unsustainable.</p><p>I don’t have a CVE. I’ve found one confirmed bug. I’m still learning.</p><p>But I know what exposure debt means now. And I don’t want to add to it.</p><p><em>If you’re also learning bug bounty in a system where discovery outruns remediation, I write about what I’m actually figuring out — numbers included. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c20b95066af9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Just Found Out What Happens to the Bugs I Report. It Changed How I Hunt.]]></title>
            <link>https://medium.com/@riyalimba/i-just-found-out-what-happens-to-the-bugs-i-report-it-changed-how-i-hunt-09a74042ba43?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/09a74042ba43</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[osint]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Fri, 23 Oct 2026 13:12:52 GMT</pubDate>
            <atom:updated>2026-10-23T13:12:52.226Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8eRaezIkQnXIff_-m4z9CQ.png" /></figure><h3>HackerOne says fix times improved 54%. The backlog still grew 131%. Here’s the math nobody explains to beginners.</h3><p>I read HackerOne’s 2026 Security Research Report on a Wednesday and found a number I couldn’t stop thinking about.</p><p>Exposure debt rose 131%.</p><p>Not technical debt. Not financial debt. Exposure debt — the accumulated pile of validated vulnerabilities that organizations know about but haven’t fixed yet .</p><p>But here’s the part that made me put my coffee down. Average resolution time improved by 54% — from 135 days down to 62 days .</p><p>They’re fixing individual bugs faster than ever. And the backlog is still growing.</p><h3>How That’s Possible</h3><p>I had to read the explanation twice.</p><p>The 62-day figure only counts reports that reached resolution. The easier fixes close first. The harder findings — the ones that require more engineering effort, more coordination, more people to sign off — are the ones left waiting.</p><p>Meanwhile, new findings keep arriving faster than teams can clear them. Nearly 70% of security leaders confirmed this: validated findings are entering their systems faster than their teams can remediate them .</p><p>A team can get better at fixing bugs and still watch the pile grow. That’s the trap.</p><h3>Why This Matters to Me Specifically</h3><p>I’ve found one confirmed bug. I’m waiting to be paid.</p><p>Reading about exposure debt, I felt something shift.</p><p>The question I’d been asking was: <em>“How do I find more bugs?”</em></p><p>The better question — the one the data points to — is: “What happens to the bugs I find after I submit them?”</p><p>If my report enters a backlog and never gets fixed, I got paid. The triager did their job. But the vulnerability is still there. The data is still exposed. The system is still vulnerable.</p><p>I contributed to someone’s backlog. Not to their security.</p><h3>What the Report Says About Researchers Like Me</h3><p>The HackerOne report includes data that feels personal.</p><p>85% of surveyed researchers are actively developing AI skills. Nearly three-quarters said AI meaningfully increased the valid findings they produce.</p><p>But 68% said they’ve shifted toward more complex bugs with higher bounties .</p><p>That’s the signal. The researchers who are succeeding are going deeper, not wider. They’re using AI to cover more ground, then focusing human judgment on the bugs that require creativity and context.</p><p>A top-earning researcher put it plainly: <em>“AI hasn’t replaced how I think as a security researcher. It’s changed how much ground I can cover”</em> .</p><h3>What I’m Doing Differently</h3><p>I can’t fix the remediation crisis. I can’t force organizations to patch faster. But I can change three things.</p><p>First, I’m treating “reproducible” as a higher bar than “valid.” A valid bug that takes an hour to reproduce is less useful than a valid bug with a one-line PoC. The easier I make it for a developer to fix, the more likely it actually gets fixed.</p><p>Second, I’m looking at remediation history, not just bounty size. A program that pays well but never patches is accumulating debt. A program with lower payouts but active patching is contributing to actual security.</p><p>Third, I’m remembering that the bottleneck isn’t me. The report says nearly 70% of security leaders confirmed that validated findings are arriving faster than teams can fix them . My job isn’t just to find bugs. It’s to make the fix as easy as possible.</p><h3>The Uncomfortable Truth</h3><p>HackerOne paid a record $89 million to researchers this year . That’s genuinely good news. The work is valued. Companies are still investing in finding bugs before attackers do.</p><p>But the 131% exposure debt increase tells a different story. It says the system is finding faster than it’s fixing. It says the gap is growing.</p><p>The bugs I find aren’t just entries in my report history. They’re entries in someone’s backlog. And if that backlog keeps growing, the whole system becomes unsustainable.</p><p>I don’t have a CVE. I’ve found one confirmed bug. I’m still learning.</p><p>But I know what exposure debt means now. And I don’t want to add to it.</p><p><em>If you’re also learning bug bounty in a system where discovery outruns remediation, I write about what I’m actually figuring out — numbers included. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=09a74042ba43" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Rewrote My Report the Way HackerOne Told Me To. It Got Triaged in 48 Hours.]]></title>
            <link>https://medium.com/@riyalimba/i-rewrote-my-report-the-way-hackerone-told-me-to-it-got-triaged-in-48-hours-034ef0b865bf?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/034ef0b865bf</guid>
            <category><![CDATA[osint]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Wed, 21 Oct 2026 13:53:59 GMT</pubDate>
            <atom:updated>2026-10-21T13:53:59.374Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NZiXGhldDRLB3JxeE-jVCQ.png" /></figure><h3>The difference wasn’t the bug. It was who I was writing for.</h3><p>I used to write reports for the product team.</p><p>I’d explain the vulnerability in detail. I’d include technical context about why it existed. I’d add background about the attack surface and the potential implications.</p><p>Then I’d hit submit and wait.</p><p>And wait.</p><p>And get closed as “informative” or “not applicable.”</p><p>I thought my reports were good. They were thorough. They were technically accurate. They demonstrated that I understood the bug.</p><p>But I was writing for the wrong person.</p><h3>What HackerOne Actually Says</h3><p>I read HackerOne’s official report-writing guide on a Tuesday and felt something I hadn’t expected: embarrassment.</p><p>The guide says, plainly: “High-quality reports help security teams quickly understand and triage vulnerabilities. A well-structured report improves response time and increases the likelihood of a positive outcome” .</p><p>Then it lists what actually matters:</p><p>A clear and concise title. Not “XSS in web app.” Something like: “Stored XSS in user profile field allows script execution on profile view” .</p><p>Detailed steps to reproduce. Numbered. Specific. With exact URLs, parameters, and user roles .</p><p>Impact assessment. What can an attacker actually do? How many users are affected? What data is exposed?</p><p>Supporting material. Screenshots. Videos. Code snippets. The actual request and response .</p><p>I read that list and realized I’d been doing almost none of it.</p><h3>What I Was Doing Wrong</h3><p>My reports had:</p><ul><li>A title that described the vulnerability class, not the specific issue</li><li>Steps to reproduce that assumed the triager knew what I knew</li><li>An impact statement that said “this could lead to account takeover” without proving it</li><li>Screenshots, but no raw requests</li></ul><p>I was writing reports for someone who already understood the bug. But the triager doesn’t already understand the bug. The triager has four minutes and hundreds of other reports in their queue.</p><p>I was making their job harder. And they were closing my reports faster.</p><h3>What I Changed</h3><p>I rewrote my next report following the HackerOne guide exactly.</p><p>Title: Not “IDOR in API.” But: “Authenticated user can access any other user’s email and phone number via GET /api/v2/users/{id}.”</p><p>Steps to reproduce: Seven numbered steps. Two accounts. Exact URLs. The specific parameter to change. What to look for in the response.</p><p>Impact: “An attacker with a free account can enumerate and extract the email address and phone number of every user on the platform by iterating the user ID parameter. No special privileges required.”</p><p>Supporting material: The raw HTTP request. The raw HTTP response. A screenshot showing the victim’s data returned to the attacker’s session.</p><p>I submitted it on a Monday.</p><p>It was triaged on Wednesday. Confirmed. Paid.</p><h3>Why This Matters More Than I Expected</h3><p>I’ve found one confirmed bug before this one. It took three weeks to get triaged. The report was technically valid, but the triager had to ask me for clarification twice.</p><p>This report took 48 hours. No questions. No back-and-forth. Just confirmation.</p><p>The difference wasn’t the bug. The first bug was arguably more interesting. The difference was the report.</p><p>A triager who can reproduce your bug in five minutes is a triager who will validate your bug. A triager who has to read your report twice, ask for clarification, and dig through screenshots to understand what you found is a triager who will close your report and move on to the next one.</p><h3>The Uncomfortable Truth</h3><p>I thought report writing was the boring part. The finding was the skill. The report was just paperwork.</p><p>I was wrong.</p><p>The report is the skill. Finding the bug is half the work. Proving it clearly, concisely, and reproducibly is the other half.</p><p>HackerOne’s guide says it directly: “A well-structured, detailed report helps security teams address vulnerabilities faster and increases the likelihood of a positive response” .</p><p>I read that sentence ten times before I understood what it meant. It means: your report is the product. The bug is the raw material. The report is what the triager actually evaluates.</p><h3>What I’m Doing Differently</h3><p>I’m writing every report like the triager has four minutes and no context.</p><p>First, I’m testing the report against myself. Can I follow my own steps and reproduce the bug in five minutes? If I can’t, the triager can’t either.</p><p>Second, I’m including the raw request and response. Not just screenshots. The actual HTTP traffic. The thing a developer can copy and paste.</p><p>Third, I’m explaining the impact in one sentence. Who can do what, and why it matters. No jargon. No hedging. Just the concrete consequence.</p><p>I don’t have a CVE. I’ve found two confirmed bugs now. The second one took 48 hours because I finally understood who I was writing for.</p><p><em>If you’re also learning to write reports that actually get read, I write about what I’m figuring out — mistakes included. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=034ef0b865bf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Found a Bug in a VDP. No Money. No Bounty. Here’s Why I’d Do It Again.]]></title>
            <link>https://medium.com/@riyalimba/i-found-a-bug-in-a-vdp-no-money-no-bounty-heres-why-i-d-do-it-again-100862bf716b?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/100862bf716b</guid>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[osint]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Tue, 20 Oct 2026 14:14:54 GMT</pubDate>
            <atom:updated>2026-10-20T14:14:54.585Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Bk_aeuSnBNEgfkeUmmamKA.png" /></figure><h3>Everyone told me to skip the programs that don’t pay. They were wrong.</h3><p>I found a bug last month. It didn’t pay anything.</p><p>Not a single dollar. Not a gift card. Not even a t-shirt.</p><p>I reported it to a Vulnerability Disclosure Program — a VDP. The organization confirmed it, fixed it, and added my name to their security acknowledgments page.</p><p>That’s it. That’s the whole reward.</p><p>And I’d do it again.</p><h3>Why I Almost Skipped It</h3><p>When I started learning bug bounty, every guide said the same thing: “Focus on programs that pay. Don’t waste your time on VDPs.”</p><p>The logic made sense. I’m not doing this for fun. I’m doing it to build a career, maybe earn some money along the way. Why spend hours on a target that won’t pay?</p><p>So I ignored VDPs for months. I only looked at paid programs. I competed with thousands of other researchers for the same bugs. I got closed out, ignored, and rejected.</p><p>Then I read something that changed my mind.</p><h3>The Thing That Changed My Mind</h3><p>I was reading an interview with a security engineer at a major company. They said something I couldn’t stop thinking about:</p><p>“Many important projects pay no bounty at all. They still get reports, sometimes a lot of them, because public credit matters. A CVE, advisory credit, or a ‘reported by’ mention is very meaningful for career capital.”</p><p>Career capital.</p><p>Not cash. Not a payout. Career capital.</p><p>I’d been thinking about bug bounty as a way to make money. But the researchers who are actually building careers aren’t just chasing bounties. They’re building proof of work.</p><p>A confirmed bug in a VDP is proof. A security advisory with your name in it is proof. A hall of fame entry is proof.</p><p>Proof doesn’t expire. Proof doesn’t get closed as “informative.” Proof is something you can point to — in a job interview, on a resume, in a conversation with a recruiter.</p><h3>What I Actually Found</h3><p>I can’t name the organization. Disclosure rules. But here’s what I can tell you.</p><p>It was a small VDP. Not a household name. Not a company with a flashy bounty program.</p><p>I spent two weeks testing their application. Not because I expected to find something big. Because I wanted to understand how it worked.</p><p>The bug was simple. An API endpoint that returned more data than it should. Not critical. Not a full account takeover. But a real data exposure issue.</p><p>I wrote the report. Steps to reproduce. Screenshots. Impact statement. Suggested fix.</p><p>They responded in two days. Confirmed the bug. Fixed it in a week. Added my name to their acknowledgments page.</p><p>No money. But I can point to it now. It’s searchable. It’s proof.</p><h3>Why This Matters More Than I Expected</h3><p>I’ve been hunting for months. I’ve found one confirmed bug on a paid program. I’m waiting to be paid.</p><p>But the VDP bug felt different.</p><p>Not because it was more technically impressive. It wasn’t. Not because the organization was more important. It wasn’t.</p><p>Because it was mine. I found it. I documented it. I wrote the report. I didn’t get lucky with a duplicate. I didn’t get closed out by a triager who didn’t have time to read carefully.</p><p>I did the work, and someone confirmed it was real.</p><p>That’s the feeling I’ve been chasing. Not the payout. The confirmation.</p><h3>What I’m Doing Differently</h3><p>I’m still hunting on paid programs. I still want to earn bounties. But I’ve changed how I think about where to spend my time.</p><p>First, I’m treating VDPs as valid targets. Not as a consolation prize. As a way to build proof of work while I’m still learning. A confirmed bug in a VDP is worth more than a closed report on a paid program.</p><p>Second, I’m looking for programs that offer credit, not just cash. A CVE. A hall of fame entry. A security advisory mention. These are credentials. They don’t expire.</p><p>Third, I’m remembering that the path isn’t linear. Some researchers find their first bug on a paid program. Some find it on a VDP. Some find it after months of nothing. The path doesn’t matter. The proof does.</p><h3>The Uncomfortable Truth</h3><p>I found a bug. It didn’t pay anything. I’d do it again.</p><p>Not because I’m noble. Because I’m building something. Not a bank account. A reputation. A track record. Proof that I can do this work.</p><p>The money will come. Or it won’t. But the proof is mine.</p><p><em>If you’re also learning to build proof of work — one confirmed bug at a time — I write about what I’m actually figuring out. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=100862bf716b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Tested the Same Bug on 3 Programs. Only One Paid Me. Here’s What I Learned About Picking Targets.]]></title>
            <link>https://medium.com/@riyalimba/i-tested-the-same-bug-on-3-programs-only-one-paid-me-heres-what-i-learned-about-picking-targets-af58e7252210?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/af58e7252210</guid>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[osint]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Mon, 19 Oct 2026 14:35:27 GMT</pubDate>
            <atom:updated>2026-10-19T14:35:27.645Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sqwnsLLsNYR_Q2hdvmOsvQ.png" /></figure><h3>Same vulnerability class. Same testing process. Three completely different outcomes.</h3><p>I found the same bug three times.</p><p>Not the exact same endpoint. Not the same company. But the same vulnerability class — a broken access control issue that let one user access another user’s data.</p><p>I reported it to three different programs over two months.</p><p>One paid. One closed it as “informative.” One never responded.</p><p>Same bug. Same skill. Same effort. Wildly different outcomes.</p><h3>The Three Reports</h3><p>I can’t name the programs. Disclosure rules. But here’s what I can tell you.</p><p>Program A was a large public program on HackerOne. Thousands of researchers. Well-known company. The bug was an IDOR in a user profile API — changing the user ID in a request returned another user’s email and phone number.</p><p>Program B was a smaller program on Bugcrowd. Less competitive. Newer scope. The bug was almost identical — same IDOR pattern, same data exposure.</p><p>Program C was a VDP on a self-hosted platform. No bounty. No cash. Just a “thanks” and a spot in their hall of fame.</p><p>Same bug class. Same testing process. Three completely different responses.</p><h3>What Happened</h3><p>Program A paid $500. It took three weeks to triage. The triager asked for a video PoC (I hadn’t included one). I made the video. They confirmed the bug. Paid.</p><p>Program B closed the report in 48 hours. The reason: “This object is intended to be public.” I argued. They didn’t respond. I moved on.</p><p>Program C responded in two days. Confirmed the bug. Fixed it in a week. Added my name to their security acknowledgments page. No money. But I can point to it. It’s searchable. It’s proof.</p><p>I spent the same amount of time on all three. The difference wasn’t my skill. It was where I pointed my tools.</p><h3>The Thing I Realized</h3><p>I’ve been picking targets wrong.</p><p>I’ve been chasing the biggest programs — the ones with the flashiest scopes, the biggest names, the highest max bounties. The ones that look impressive on a resume.</p><p>But the data says something different. The smaller programs — the ones with fewer researchers, newer scopes, less saturation — are where the signal-to-noise ratio is better. Where a legitimate report is more likely to be read carefully. Where the triager isn’t drowning in a thousand AI-generated submissions.</p><p>The GitHub payout restructuring proved this. Programs with fewer than 50 researchers had 15x higher ROI than programs with 500+ .</p><p>Vercel’s Platform Protection program had 34 participants and an average bounty of $50,000**. Coinbase had 771 participants and an average bounty of **$200.</p><p>Same effort. Same skill. Wildly different outcomes.</p><h3>What This Means for Beginners</h3><p>I’ve found one confirmed bug. I’m waiting to be paid. I’m still learning.</p><p>But I know now that the target matters as much as the technique.</p><p>A beginner who picks the right program can find a $500 bug in a week. A beginner who picks the wrong program can find the same bug and get closed as “informative” in 48 hours.</p><p>The difference isn’t skill. It’s selection.</p><h3>What I’m Doing Differently</h3><p>First, I’m checking participant counts before I invest time. A program with 34 researchers is worth more of my attention than a program with 1,276. Not because the bugs are easier — because the competition is thinner.</p><p>Second, I’m looking at average bounty, not max bounty. A program with a $500,000 max bounty sounds amazing until you realize the average payout is $200. A program with a $50,000 max bounty and a $50,000 average is a different game entirely.</p><p>Third, I’m treating VDPs as valid targets. A hall of fame entry or a CVE doesn’t pay rent, but it’s proof of work. It’s something to put on a resume. It’s a credential that doesn’t expire.</p><h3>The Uncomfortable Truth</h3><p>I found the same bug three times. Only one program paid me. The other two are still out there, unfixed or ignored, waiting for someone else to find them and submit them to a program that actually cares.</p><p>The bug didn’t change. The payout did. The only variable was where I aimed.</p><p>I don’t know if I’ll find my next bug on a program with 34 researchers or 1,276. But I know now that the choice matters.</p><p>And I’m choosing differently.</p><p><em>If you’re also learning to navigate the bug bounty economy — where to hunt, not just how — I write about what I’m actually figuring out. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=af58e7252210" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Intel Closed Its Bug Bounty. Google Paused Its. curl Quit. I’m Still Here.]]></title>
            <link>https://medium.com/@riyalimba/intel-closed-its-bug-bounty-google-paused-its-curl-quit-im-still-here-e50353c91270?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/e50353c91270</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[osint]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Sun, 18 Oct 2026 14:38:42 GMT</pubDate>
            <atom:updated>2026-10-18T14:38:42.547Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*quE6xjiWUvJmMy8-dbrdJA.png" /></figure><h3>The programs are dying because AI made filing a fake report cheaper than ever. That doesn’t mean the work is gone.</h3><p>I read the Intel news on a Thursday and felt something I didn’t expect: not surprise.</p><p>Intel’s paid bug bounty program on Intigriti was reportedly suspended in September 2026. The program used to pay up to $100,000. Now it’s a “responsible disclosure program without bounties” .</p><p>The reason wasn’t stated by Intel. But the timing told the story. Google had paused its OSS VRP just weeks earlier, citing “a significant rise in automated submissions, the vast majority of which are not valid” . curl had ended its bounty program in January for the same reason — confirmed vulnerabilities fell from over 15% to below 5% as AI slop flooded the queue .</p><p>Intel didn’t need to explain. The pattern was already visible.</p><h3>The Pattern Nobody Wants to Name</h3><p>I’ve been writing about this for months. Each time a program dies, the explanation is the same: AI made it too easy to generate a convincing-looking report that contains nothing real.</p><p>A security researcher at Malwarebytes put it plainly: “The cost of filing a plausible report has fallen sharply, while the cost of proving or disproving it remains human-intensive” .</p><p>That’s the entire crisis in one sentence.</p><p>A triager at Google or Intel or curl has to read every submission carefully enough to determine whether it’s real. AI can generate a polished, technically-worded report in seconds. Human verification still takes the same amount of time it always did — sometimes longer, because the AI-generated reports are <em>better written</em> than the old low-effort spam .</p><p>The economics broke. And when the economics break, programs close.</p><h3>What This Means for Me Specifically</h3><p>I’ve found one confirmed bug. I’m waiting to be paid.</p><p>Reading about Intel and Google and curl, I felt something I hadn’t expected: clarity about where to spend my time.</p><p>The programs that are closing are the ones with the biggest scopes and the loudest reputations. The ones where thousands of researchers are already competing. The ones where an AI-generated fake report can slip into the queue and waste everyone’s time.</p><p>The programs that are still open are the ones I’ve been ignoring.</p><p>VDPs. Small open-source projects. Programs that don’t pay cash but offer something else: credit. A CVE. A “reported by” mention in a security advisory.</p><p>A security engineer quoted in one report said it directly: <em>“Many important projects pay no bounty at all. They still get reports, sometimes a lot of them, because public credit matters. A CVE, advisory credit or a ‘reported by’ mention is very meaningful for career capital”</em> .</p><p>I’ve been chasing bounties. I should have been chasing proof of work.</p><h3>The Thing About “Slop”</h3><p>The reports that killed these programs weren’t malicious. They were careless.</p><p>A researcher runs an AI tool, gets a finding, pastes it into a report template, hits submit. They don’t verify the finding against the actual target. They don’t confirm the PoC works. They don’t ask “is this actually exploitable?” .</p><p>I understand the temptation. I’ve felt it. When you’ve been hunting for months with nothing to show, the idea of submitting <em>something</em> feels better than submitting nothing.</p><p>But “something” is what killed Intel’s program. “Something” is what made curl’s maintainer give up. “Something” is what Google cited when it paused its OSS VRP until Q1 2027 at the earliest .</p><p>The researchers who survive this shift won’t be the ones who submit the most. They’ll be the ones who submit the fewest things that turn out to be wrong.</p><h3>What I’m Doing Differently</h3><p>I can’t reopen Intel’s program. I can’t force Google to accept my report. But I can change where I aim.</p><p>First, I’m looking at VDPs. Programs that don’t pay cash but offer credit. A CVE in my name is worth more than a $200 bounty that took six months to get triaged. It’s permanent. It’s searchable. It’s proof I can point to.</p><p>Second, I’m treating “verified” as a higher bar than “submitted.” Before I hit submit on anything, I run the PoC against the real target. I confirm the behavior. I make sure I can explain why it works in one sentence. If I can’t, I don’t submit.</p><p>Third, I’m remembering that the programs closing aren’t the only programs. curl still accepts reports through GitHub — just without a bounty . Google still has its Patch Rewards Program and Cloud VRP . Intel still has a disclosure program. The doors aren’t all closed. Some are just harder to see.</p><h3>The Uncomfortable Truth</h3><p>Intel’s bounty program is gone. Google’s OSS VRP is paused. curl stopped paying years ago. The era of “learn the basics, find an obvious bug, get paid” is ending.</p><p>But the era of “find something real, document it carefully, and build a reputation” is still here. It’s just harder. It requires more patience. It requires caring about the quality of the report as much as the payout.</p><p>I don’t have a CVE. I’ve found one confirmed bug. I’m still waiting to be paid.</p><p>But I know where the doors are now. And I know which ones are still open.</p><p><em>If you’re also navigating the closure of the big programs and the search for what’s still open, I write about what I’m actually figuring out. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e50353c91270" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Just Found Out What Happens to the Bugs I Report. It Changed How I Hunt.]]></title>
            <link>https://medium.com/@riyalimba/i-just-found-out-what-happens-to-the-bugs-i-report-it-changed-how-i-hunt-1da7eff8a639?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/1da7eff8a639</guid>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[osint]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Sat, 17 Oct 2026 12:44:58 GMT</pubDate>
            <atom:updated>2026-10-17T12:44:58.817Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1oE4eARKKpKEODDfRJWXWg.png" /></figure><h3>HackerOne calls it “exposure debt.” It grew 131% while fix times got 54% faster. The math doesn’t work the way I thought it did.</h3><p>I read HackerOne’s annual report on a Wednesday and found a term I’d never heard before.</p><p>Exposure debt.</p><p>Not technical debt. Not financial debt. Exposure debt — the accumulated pile of validated vulnerabilities that organizations know about but haven’t fixed yet.</p><p>The number that stopped me: it grew 131% in two years.</p><p>But here’s the part that made me put my coffee down. Average resolution time improved by 54% — from 135 days down to 62 days.</p><p>They’re fixing individual bugs faster than ever. And the backlog is still growing.</p><h3>How That’s Possible</h3><p>I had to read the explanation twice.</p><p>The 62-day figure only counts reports that reached resolution. Simpler fixes close first. The harder findings — the ones that require more engineering effort, more coordination, more people to sign off — are the ones left waiting.</p><p>Meanwhile, new findings keep arriving faster than teams can clear them. Nearly 70% of security leaders confirmed this: validated findings are entering their systems faster than their teams can remediate them.</p><p>A team can get better at fixing bugs and still watch the pile grow. That’s the trap.</p><h3>Why This Matters to Me Specifically</h3><p>I’ve found one confirmed bug. I’m waiting to be paid.</p><p>Reading about exposure debt, I felt something shift.</p><p>The question I’d been asking was: <em>“How do I find more bugs?”</em></p><p>The better question — the one the data points to — is: “What happens to the bugs I find after I submit them?”</p><p>If my report enters a backlog and never gets fixed, I got paid. The triager did their job. But the vulnerability is still there. The data is still exposed. The system is still vulnerable.</p><p>I contributed to someone’s backlog. Not to their security.</p><h3>What the Report Says About Researchers Like Me</h3><p>The HackerOne report includes data that feels personal.</p><p>85% of surveyed researchers are actively developing AI skills. Nearly three-quarters said AI meaningfully increased the valid findings they produce.</p><p>But 68% said they’ve shifted toward more complex bugs with higher bounties.</p><p>That’s the signal. The researchers who are succeeding are going deeper, not wider. They’re using AI to cover more ground, then focusing human judgment on the bugs that require creativity and context.</p><p>A top-earning researcher put it plainly: <em>“AI hasn’t replaced how I think as a security researcher. It’s changed how much ground I can cover”</em>.</p><h3>What I’m Doing Differently</h3><p>I can’t fix the remediation crisis. I can’t force organizations to patch faster. But I can change three things.</p><p>First, I’m treating “reproducible” as a higher bar than “valid.” A valid bug that takes an hour to reproduce is less useful than a valid bug with a one-line PoC. The easier I make it for a developer to fix, the more likely it actually gets fixed.</p><p>Second, I’m looking at remediation history, not just bounty size. A program that pays well but never patches is accumulating debt. A program with lower payouts but active patching is contributing to actual security.</p><p>Third, I’m remembering that the bottleneck isn’t me. The report says nearly 70% of security leaders confirmed that validated findings are arriving faster than teams can fix them. My job isn’t just to find bugs. It’s to make the fix as easy as possible.</p><h3>The Uncomfortable Truth</h3><p>HackerOne paid a record $89 million to researchers this year. That’s genuinely good news. The work is valued. Companies are still investing in finding bugs before attackers do.</p><p>But the 131% exposure debt increase tells a different story. It says the system is finding faster than it’s fixing. It says the gap is growing.</p><p>The bugs I find aren’t just entries in my report history. They’re entries in someone’s backlog. And if that backlog keeps growing, the whole system becomes unsustainable.</p><p>I don’t have a CVE. I’ve found one confirmed bug. I’m still learning.</p><p>But I know what exposure debt means now. And I don’t want to add to it.</p><p><em>If you’re also learning bug bounty in a system where discovery outruns remediation, I write about what I’m actually figuring out — numbers included. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1da7eff8a639" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Just Learned a Term That Changed How I Think About Every Report I Submit.]]></title>
            <link>https://medium.com/@riyalimba/i-just-learned-a-term-that-changed-how-i-think-about-every-report-i-submit-f465971c6400?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/f465971c6400</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[osint]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Fri, 16 Oct 2026 16:31:35 GMT</pubDate>
            <atom:updated>2026-10-16T16:31:35.838Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_2VyQ1xm6-5BMPu3IfIxsA.png" /></figure><h3>HackerOne calls it “exposure debt.” It grew 131% in two years. And I’ve been contributing to it without knowing.</h3><p>I read HackerOne’s annual report on a Wednesday morning and found a term I’d never heard before.</p><p>Exposure debt.</p><p>Not technical debt. Not financial debt. Exposure debt — the accumulation of validated vulnerabilities that organizations know about but haven’t fixed yet.</p><p>The report found that this backlog grew 131% in two years.</p><p>But here’s the part that made me put my coffee down: average resolution time improved by 54%. Organizations are fixing individual bugs faster than ever — from 135 days down to 62 days.</p><p>They’re fixing each bug faster. They’re fixing fewer bugs overall.</p><p>The backlog grew 21x. Unresolved criticals grew 25x.</p><p>I sat there staring at those numbers for a long time.</p><h3>What This Actually Means</h3><p>Let me walk through the mechanics, because I had to read the report twice to understand.</p><p>Every time a researcher submits a valid bug, it enters a queue. A triager validates it. It gets assigned to a developer. The developer fixes it. The bug leaves the queue. Success.</p><p>That’s the theory.</p><p>The reality: validated findings are arriving faster than teams can resolve them. Nearly 70% of security leaders confirmed this.</p><p>The queue isn’t just growing. It’s growing faster than the fixing capacity can keep up, even though the fixing capacity itself is improving.</p><p>It’s like a bathtub. The drain is wider than it’s ever been. But the faucet is running at a rate the drain was never designed for.</p><h3>Why I Care About This Specifically</h3><p>I’ve found one confirmed bug. I’m waiting to be paid.</p><p>Reading about exposure debt, I felt something shift.</p><p>The question I’d been asking was: <em>“How do I find bugs?”</em></p><p>The better question — the one the data points to — is: “What happens to the bugs I find after I submit them?”</p><p>If my report enters a backlog and never gets fixed, I got paid. The triager did their job. But the vulnerability is still there. The data is still exposed. The system is still vulnerable.</p><p>I contributed to someone’s backlog. Not to their security.</p><h3>What the Report Says About Researchers Like Me</h3><p>The HackerOne report includes data that feels personal.</p><p>85% of surveyed researchers are actively developing AI skills. Nearly three-quarters said AI meaningfully increased the valid findings they produce.</p><p>But 68% said they’ve shifted toward more complex bugs with higher bounties.</p><p>That’s the signal. The researchers who are succeeding are going deeper, not wider. They’re using AI to cover more ground, then focusing human judgment on the bugs that require creativity and context.</p><p>A top-earning researcher put it plainly: <em>“AI hasn’t replaced how I think as a security researcher. It’s changed how much ground I can cover”</em>.</p><h3>What I’m Doing Differently</h3><p>I can’t fix the remediation crisis. I can’t force organizations to patch faster. But I can change three things.</p><p>First, I’m treating “reproducible” as a higher bar than “valid.” A valid bug that takes an hour to reproduce is less useful than a valid bug with a one-line PoC. The easier I make it for a developer to fix, the more likely it actually gets fixed.</p><p>Second, I’m looking at remediation history, not just bounty size. A program that pays well but never patches is accumulating debt. A program with lower payouts but active patching is contributing to actual security.</p><p>Third, I’m remembering that the bottleneck isn’t me. The report says nearly 70% of security leaders confirmed that validated findings are arriving faster than teams can fix them. My job isn’t just to find bugs. It’s to make the fix as easy as possible.</p><h3>The Uncomfortable Truth</h3><p>HackerOne paid a record $89 million to researchers this year. That’s genuinely good news. The work is valued. Companies are still investing in finding bugs before attackers do.</p><p>But the 131% exposure debt increase tells a different story. It says the system is finding faster than it’s fixing. It says the gap is growing.</p><p>The bugs I find aren’t just entries in my report history. They’re entries in someone’s backlog. And if that backlog keeps growing, the whole system becomes unsustainable.</p><p>I don’t have a CVE. I’ve found one confirmed bug. I’m still learning.</p><p>But I know what exposure debt means now. And I don’t want to add to it.</p><p><em>If you’re also learning bug bounty in a system where discovery outruns remediation, I write about what I’m actually figuring out — numbers included. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f465971c6400" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Microsoft Paid a Record $20 Million. The Average Researcher Took Home $35,000.]]></title>
            <link>https://medium.com/@riyalimba/microsoft-paid-a-record-20-million-the-average-researcher-took-home-35-000-4d8568ee89fe?source=rss-7e38590ffb2f------2</link>
            <guid isPermaLink="false">https://medium.com/p/4d8568ee89fe</guid>
            <category><![CDATA[bug-bounty]]></category>
            <category><![CDATA[ethical-hacking]]></category>
            <category><![CDATA[bug-bounty-tips]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[osint]]></category>
            <dc:creator><![CDATA[Riya Limba]]></dc:creator>
            <pubDate>Thu, 15 Oct 2026 13:49:44 GMT</pubDate>
            <atom:updated>2026-10-15T13:49:44.157Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Xf5B7Uf25R6r1vhksxHfdQ.png" /></figure><h3>The money is there. But the math for individual hunters tells a different story.</h3><p>I read Microsoft’s bug bounty announcement on a Tuesday morning.</p><p>The headline was everything the community wanted to hear. **$20 million paid to 562 researchers** across 64 countries. A record total. Up from $17 million the previous year. The largest single reward: $200,000.</p><p>The pie got bigger. Microsoft expanded its scope. More researchers got paid.</p><p>Then I scrolled to the second paragraph.</p><p>Average earnings per participant dropped from $49,000 to $35,000.</p><p>I sat with that for a moment.</p><p>The total payout went up. The number of researchers went up. The money each person actually took home went down.</p><h3>The Math Nobody Talks About</h3><p>Microsoft paid $20 million to 562 people. That’s an average of about $35,587 per researcher.</p><p>The previous year: $17 million to 344 researchers. Average: $49,418.</p><p>The pool grew by $3 million. The number of people swimming in it grew by 218. Everyone’s slice got smaller.</p><p>This isn’t unique to Microsoft. I’ve been writing about the numbers behind bug bounty for months. HackerOne’s report volume doubled year over year. Bugcrowd saw a 300% spike in submissions during a three-week period before it normalized at roughly double the historical rate. Trend Micro’s Zero Day Initiative reported submissions up 450% in April before moderating.</p><p>More reports. More researchers. More competition for the same bugs.</p><h3>What Microsoft Actually Said</h3><p>I read the full announcement carefully. Microsoft framed the surge as a success story: expanded eligible targets, including open-source dependencies and third-party code. A “dramatic second-half increase in reports driven by artificial intelligence”.</p><p>They’re not wrong. AI-assisted discovery is finding real bugs. Microsoft’s Patch Tuesday releases went from 79 fixes in March 2026 to 966 in September — more than ten times as many.</p><p>The bugs are real. The fixes are happening. The system is working.</p><p>But the system is working for the companies receiving the reports. It’s working less well for the individuals writing them.</p><h3>The Part That Matters for Beginners</h3><p>I’ve found one confirmed bug. I’m waiting to be paid. I’m still learning how to write reports that don’t get closed in four minutes.</p><p>Reading Microsoft’s numbers, I felt something I hadn’t expected: clarity.</p><p>The question I’d been asking was: “How do I compete with AI?”</p><p>The better question — the one the data points to — is: “How do I compete with everyone else using AI?”</p><p>Because that’s what’s happening. The tools that make it easier for me to find bugs also make it easier for everyone to find bugs. The barrier to entry has dropped. The pool of hunters has grown. The bugs haven’t multiplied to keep pace.</p><p>A researcher who found a critical bug in 2024 might have been one of 50 people looking at that target. In 2026, they’re one of 500. The bug pays the same. The odds of finding it first have collapsed.</p><h3>What I’m Taking From This</h3><p>I can’t control the size of the pool. I can’t make other researchers stop hunting. I can’t compete on volume.</p><p>But I can do three things the data suggests actually matter.</p><p>First, I’m going deeper on fewer targets. The researchers who are succeeding in this environment aren’t scanning everything. They’re spending weeks on one application, understanding how it works, finding the bugs that don’t show up in scanner output.</p><p>Second, I’m treating report quality as a differentiator. If everyone can generate a report with AI, the researcher who writes a report a triager can validate in five minutes stands out. Reproducibility over volume. Clarity over cleverness.</p><p>Third, I’m remembering that $35,000 is still a lot of money. The average dropped. But the average is still higher than most beginners will see for years. The path exists. It’s just narrower than it used to be.</p><h3>The Uncomfortable Truth</h3><p>Microsoft’s $20 million payout is genuinely good news. It means the work is valued. It means companies are still investing in finding bugs before attackers do.</p><p>But the average earnings drop tells a different story. It says the field is crowding. It says the easy money is gone. It says the researchers who succeed in the next few years will be the ones who adapt to a world where everyone has the same tools.</p><p>I don’t have a CVE. I’ve found one confirmed bug. I’m still learning.</p><p>But I know what the numbers look like now. And I’m adjusting.</p><p><em>If you’re also navigating the numbers behind the bug bounty economy, I write about what I’m actually figuring out — data included. Follow for more field notes from the bottom of the learning curve.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4d8568ee89fe" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>