Open to security roles and pentest work

Parth Narula Security researcher and penetration tester

I find the access control, logic and API bugs that scanners miss. 450+ organizations have fixed what I reported, and on 45 Halls of Fame mine is the very first name.

01Record

The numbers, with receipts

Each number links to where you can check it yourself.

Organizations secured
450+
387 listed publicly. Why not all?
Halls of Fame
250+
136 public pages you can open
Times I was the first researcher
45
The first name on their wall
CVEs
3
CVE-2025-56697 published, two pending in Netwrix ADV-2026-018
HackWithIndia 2026, 21st of 10,000
AIR21
See the leaderboard

Why 387 and not 450+?

Not every thank you is mine to publish. The difference is real, reported and fixed work that I can't put my name next to in public yet.

226 of the listed organizations link to a public page or a screenshot right here. Ask me about any of the others.

Check the full list
  • Sub companies and clients of a listed group, credited under the parent's name
  • Private programs and pentest engagements under NDA
  • Programs that haven't approved public disclosure yet
  • Reports already confirmed, waiting for the Hall of Fame to update

02Scope

What I test

Five surfaces, one question every time: where does this system trust something it shouldn't?

Broken access control, business logic and authentication. The bugs that need someone to understand how the app is supposed to work first.

  • IDOR and privilege escalation
  • Business logic and race conditions
  • OTP, password reset and session flaws
  • XSS chains to account takeover
  • Price and checkout manipulation
  • SSRF and open redirects

Certified eWPTX v3CEH v13eJPT v2APIsec ACP Details

03Case files

Reported, fixed, written up

Eight findings that are fixed and public. Replay any of them step by step. Targets stay redacted unless the program allows naming them.

  1. Case 01CriticalUpdate integrity, CWE-494 and CWE-22

    Two flaws in PingCastle's auto updater

    Product
    Netwrix PingCastle 4.0.0 and earlier
    Component
    PingCastleAutoUpdater.exe
    Outcome
    Fixed in 4.0.0.31, advisory ADV-2026-018

    PingCastle audits Active Directory, so it runs with the reach an auditor has. Its updater replaced the installed binaries with whatever it downloaded, without checking the package was really Netwrix's, and the check meant to keep extracted files inside the update folder compared paths by prefix, so a crafted name walked straight past it. Netwrix scored the pair 9.0 and said exploitation could reach the Active Directory environment being assessed. Fixed in 4.0.0.31, with the advisory crediting me by name and the CVE identifiers still pending.

  2. Case 02CriticalExposed remote access

    Recon to shell access at Nokia

    Target
    nokia.com subdomains
    Found with
    Google dorks
    Outcome
    Nokia Hall of Fame, June 2026

    Dorking turned up an Apache Guacamole remote access gateway on a Nokia subdomain that was never meant to face the internet. From there, shell access. Reported, closed, credited.

    Read the writeup
  3. Case 03HighIDOR, PII exposure

    85,000+ emails, one URL away

    Target
    Not disclosed
    Found with
    URL manipulation
    Outcome
    Fixed after report

    Two platforms, the same mistake. Changing one number in a URL returned other people's records. Followed through, the exposure passed 85,000 email addresses.

    Read the writeup
  4. Case 04HighIDOR, account takeover

    The invite token that came back in the response

    Target
    Not disclosed
    Found with
    GraphQL mutation
    Outcome
    Fixed after report

    A chat support platform let me invite teammates. The API answered with the activation token for the invited address, so I could finish the signup myself, for an inbox I never had.

    Read the writeup
  5. Case 05HighMass assignment

    One extra field, a whole account

    Target
    Not disclosed
    Found with
    Request tampering
    Outcome
    Fixed after report

    The profile form let me edit my name and address. The API behind it accepted more than the form ever sent, including the one field the interface refused to show.

    Read the writeup
  6. Case 06MediumLogic flaw, OTP

    One code verified the wrong phone number

    Target
    Not disclosed
    Found with
    Two chained flaws
    Outcome
    Flow removed after report

    Onboarding demanded phone verification and blocked my country in the dropdown. The backend had never heard of either rule, and treated a one time code as valid for any number on the account.

    Read the writeup
  7. Case 07HighXSS chain

    The self XSS everyone closes

    Target
    Not disclosed
    Found with
    Chaining a feature
    Outcome
    Fixed after report

    A payload in my own profile field only ever fired at me, which is where most people stop. One unrelated setting put the same field on a public page.

    Read the writeup
  8. Case 08HighCredential exposure

    Live API keys in the web archive

    Target
    Not disclosed
    Found with
    Archived URLs
    Outcome
    Fixed after report

    The live API was well built. Its history was not. Archived URLs carried API keys in the query string, and some of them still worked.

    Read the writeup
  9. Case 09MediumStored XSS, CWE-79

    CVE-2025-56697

    Product
    Kashipara Computer Based Test 1.0
    Credit
    Parth Narula
    Outcome
    Published CVE

    Input saved by one user ran in the browser of everyone who opened the page. My first published CVE. More findings are confirmed and waiting on their IDs.

Every step, in text

The replay above is the same content. This is here so it works without JavaScript, and so you can read it all at once.

Recon to shell access at Nokia6 steps
  1. 01Dork for the panel

    I was hunting a private program and searched for exposed Apache Guacamole instances. A few queries later a live login panel answered on a Nokia subdomain, on a non standard port.

    https://sub.sub.nokia.com:port/guacamole
  2. 02Try the default pair

    Nothing about the page looked vulnerable. It never does. Guacamole ships with one well known account, and on a production instance it still worked.

    Username: guacadmin
    Password: guacadmin
  3. 03Read the connection list

    The dashboard opened as the admin account. No MFA, no rate limiting. Two folders held the whole internal map: SSH and RDP connections, backend and frontend systems, jumphosts.

    NonProd/
    PROD/
      jumphost, backend, frontend, RDP hosts
  4. 04Measure the blast radius

    History held real usernames, session lengths, internal IP addresses and timestamps. The user list ran to dozens of nokia.com accounts, and the groups were split by Dev, non_prod and infra_access with per connection read, update and delete rights.

  5. 05Open a production shell

    Clicking a PROD jumphost opened a terminal in the browser, already logged in. A single command confirmed it was a real production filesystem. That turned an admin panel takeover into remote shell access on internal infrastructure.

    $ ls
  6. 06Report and close

    Everything stopped there and went into the report: how it was found, what it reached, and the fix. One unchanged default credential pair was the whole chain. Nokia fixed it and credited me in their Hall of Fame.

85,000+ emails, one URL away6 steps
  1. 01Sign up and look at the URL

    Signing up redirected me to the usual confirm your email page. The address carried a plain numeric user ID.

    /user/checkEmail/61250
  2. 02Subtract one

    I changed the ID by one. The page did not break. It rendered another user's confirmation page, with their email address on it.

    /user/checkEmail/61249
  3. 03Measure it

    The IDs were sequential, no login was needed, and nothing rate limited the requests. Administrative addresses showed up in the same range. That alone reached past 60,000 users.

  4. 04Find the same bug wearing a different hat

    On an unrelated platform a profile link led to a printable page that answered page not found. The breadcrumb above the error still printed the account's email address.

    Home > myemail@example.com > Page not found
  5. 05Walk the range again

    Decrementing that ID leaked addresses the same way, about 25,000 more. An error page is still a page: the template rendered the record before deciding it had nothing to show.

    /user/26720/printable  ->  /user/26719/printable
  6. 06Report both

    Both went to their owners with the steps, the ranges I touched and nothing kept. Both were fixed. Authentication is not authorization, and a 404 can still hand over the data.

The invite token that came back in the response6 steps
  1. 01Use the feature as intended

    As the owner of a team I could invite agents by email. One endpoint handled everything, so the whole flow was a single mutation.

    POST /graphql
    
    {"operationName":"InviteAgents",
     "variables":{"emails":["victim@example.com"]}}
  2. 02Read the response, not the screen

    The invite succeeded, as expected. The response also carried an id for the invitation, which is where it stopped being routine.

    {"inviteAgents":{"success":true,
      "invitations":[{"id":"4ffacc07...",
        "email":"victim@example.com"}]}}
  3. 03Recognize the shape

    I had invited myself earlier, so I knew what an activation link looked like. Everything before the token was fixed. Only the token changed.

    /signup/activate/<token>
  4. 04Paste the token from the API

    Dropping the id from the response into that path opened the activation page for an address I did not control. No mailbox, no link, no second check.

    /signup/activate/4ffacc07...
  5. 05Finish someone else's signup

    The form asked for a name and a password. I set both and submitted. The account was created and appeared as an active agent on the team, under the name I chose.

  6. 06Report it

    The chain was short: an identifier that could change account state was returned to whoever asked, and every step downstream trusted it. Reported with the mutation and the response, and fixed.

One extra field, a whole account6 steps
  1. 01Save the profile, watch the request

    Updating my profile sent a plain PUT with the fields the form owns: name, address, phone. Changing the account email was disabled in the interface.

    PUT /api/users/profile
    
    {"about":{"givenName":"..."},
     "contact":{"postalCode":"...",
       "mobile":"..."}}
  2. 02Read what comes back

    The response returned the whole user object, including fields I never touched. The account email sat inside the same contact object I was allowed to edit, next to a confirmed flag.

    201 Created
    
    {"contact":{"email":"me@example.com",
       "confirmed":true, ...}}
  3. 03Send the field the form will not

    So I added email to the object I was already permitted to change, and left everything else identical.

    PUT /api/users/profile
    
    {"contact":{ ...,
       "email":"attacker@example.com"}}
  4. 04Watch it stick

    201 again. The response carried the new address, still marked confirmed. No verification mail, no OTP, no link. Reloading the profile showed the swap had really happened.

    {"contact":{"email":"attacker@example.com",
       "confirmed":true}}
  5. 05Push on the impact

    Pointing the field at an address that already had an account corrupted that account instead: the owner was locked out, and a password reset did not bring them back. That moved it from a settings bug to account takeover.

  6. 06Report it

    Reported with both requests and the lockout case. The interface saying a field is read only means nothing; the API is what decides.

One code verified the wrong phone number6 steps
  1. 01Meet the gate

    No dashboard until a phone number was verified, and the country dropdown did not list mine. The interface was stating a rule.

    POST /api/customer/phone_numbers
    
    {"number":"+54..."}
  2. 02Ask who enforces it

    I filled the form with an allowed country, then changed only that one value in the request to a number in the blocked country. Same token, same headers, same method.

    {"number":"+91..."}
  3. 03The backend never knew

    201 Created. A second verification resource with its own id, and the code arrived on the blocked number. The restriction existed only in the dropdown.

  4. 04Swap the resource

    Now two resources existed, each with its own id. I sent the code that belonged to the second one, against the id of the first.

    Code issued for resource B
    submitted against resource A
  5. 05Watch it pass

    200 OK. The server never checked that the code belonged to the resource in the URL, only that it was a valid code for my account. Onboarding moved to the next step, satisfied by a number that was never verified.

    HTTP/2 200 OK
  6. 06Report it

    Two weaknesses, neither critical alone: client side rules with nothing behind them, and a one time code treated as an account wide token. The program did not patch the endpoint. They removed phone verification from onboarding and rebuilt sign in around email.

The self XSS everyone closes6 steps
  1. 01Test the name field

    Signup had an artist name field. Names get reused everywhere, so it is always worth a probe. A basic HTML payload went in.

    "'<h1>test</h1>
  2. 02Find the first crack

    The confirmation email rendered the HTML instead of escaping it. Only HTML injection in an email, but it proved the value was being trusted in at least one template.

  3. 03Escalate in place

    In profile settings I replaced it with a script payload. It executed when I opened my own settings page. Stored, but firing only at me. Self XSS, low impact, usually closed.

    <img/src/x onerror=...>
  4. 04Check the public view

    The public profile did not execute it. That is the point where the finding looks finished and most reports get written up as low severity.

  5. 05Turn on a feature

    Settings had an Open to Collaboration option, off by default. Switching it on changed what the public profile renders.

    {"changeType":"PROFILE",
     "fullname":"<payload>",
     "collaborationStatusId":"REMOTE_OR_IN_PERSON"}
  6. 06Fire it at everyone else

    A new icon appeared on the public profile, and clicking it ran the payload. Anyone visiting the profile could trigger it. Same payload, same field, new reach: stored XSS. Reported with both states and fixed.

Live API keys in the web archive6 steps
  1. 01Read the documentation first

    A Swagger interface laid out endpoints for deployments, logs and more. The implementation itself looked solid, so the attack surface had to come from somewhere else.

  2. 02Search the archive

    I went through archived URLs looking for the usual markers. Several of them carried credentials directly in the query string.

    ...?api_key=...
  3. 03Separate dead from live

    Most answered 401, 403, trial expired or invalid key, which is what you expect from old credentials. A few answered normally, which is not.

  4. 04Confirm read access

    Authorizing Swagger with a working key and calling a read endpoint returned real account data. At that point it stopped being historical disclosure and became an active credential.

    GET /v3/logs?api_key=<redacted>
  5. 05Check what else it allows

    A write endpoint answered 201 Created, and reading back showed the new entry. The key was not read only.

    POST /v3/logs?api_key=<redacted>
  6. 06Report it

    Reported with the archived sources and the exact permissions proven, so the keys could be rotated rather than just removed from the index. A secret that leaked years ago is still a secret that works until someone revokes it.

04In their words

What they wrote next to my name

Most Halls of Fame list a name and a date. These programs also published what was found, so this is their wording, not mine.

05First in

45 walls where my name came first

Some programs had never thanked anyone before my report. On these Halls of Fame, I'm entry number one. Select any card to open its public page.

38 countries, every report sent from India. Drag to turn it.

06Where the reports went

From WHO to the city hall

Global enterprises, governments, universities, banks and the startups in between.

  1. NLNetherlands71
  2. USUnited States57
  3. DEGermany51
  4. AUAustralia30
  5. GBUnited Kingdom19
  6. INIndia18
  7. SESweden15
Browse all 387 organizations

Some of the organizations that recognized my work

  • Apple
  • World Health Organization
  • UNESCO
  • European Union
  • Boeing
  • Nokia
  • BASF
  • Thales
  • Philips
  • Red Hat
  • TP-Link
  • Michelin
  • MediaTek
  • BBC
  • Pirelli
  • Lloyd's of London
  • Commerzbank
  • Deutsche Börse
  • Proton
  • Wiz
  • Swedish Police Authority
  • City of Berlin
  • Australian Department of Home Affairs
  • Drexel University
  • Cambridge University Press
  • TU Delft

07Proof

What showed up at my door

Letters on university letterhead, certificates, swag from 30 programs, and a trophy that did not arrive by post. Pick them up and move them around.

172 Hall of Fame screenshots 11 letters and certificates 22 swag photos

Open the archive

08Background

Where the experience comes from

Experience

  1. Security Researcher

    Nov 2023 to now

    Independent, through disclosure and bug bounty programs

    • Reported vulnerabilities to 450+ organizations, from Apple, NVIDIA and the United Nations to Nokia, Boeing, Red Hat and the BBC.
    • Credited on 250+ Halls of Fame, including Nokia and CERT-EU, and the first researcher on 45 of them. Thanked in writing by Drexel University's CISO and GeeksforGeeks, with a letter of appreciation from Thales PSIRT and one from TP-Link on the way.
    • CVE-2025-56697 for a stored XSS, published on NVD and CVE.org. Two more are pending for Netwrix PingCastle, fixed in 4.0.0.31 and published as advisory ADV-2026-018 at CVSS 9.0.
  2. Mentor, Bug Bounty and Web Security

    Jun 2025 to now
    Unihackers
    • Mentor students and new hunters, turning my own reports into practical lessons.
    • Write guides on API security, request manipulation, target selection and prompt injection.
  3. Founder and Lead Pentester

    2023 to now
    ScriptJacker LLP
    • Manual security assessments across web, API, Android and network for 150+ clients, with 2,500+ bugs reported and resolved.
  4. Penetration Tester Intern

    Aug to Sep 2023

    Forensic Academy

    • Burp Suite, Wireshark, Cisco Packet Tracer and malware analysis tools. Built and tested an enterprise network and a live web application, graded A+.

Talks and teaching

Web application pentesting masterclassInvited by the Cyber Phoenix Club, Guru Jambheshwar University of Science and Technology, Hisar. Around 150 students, covering web fundamentals, Burp Suite, the OWASP Top 10 and bug bounty basics.Oct 2025. Photos and press coverage
Learn AI security: attack it, defend it, learn itA session I wrote and delivered on where AI systems actually break: the LLM, agentic and MCP attack surface, the OWASP list behind each one, and two prompt injection demos I built to run live.2026. Open all 57 slides
How to start in cybersecurityGuest session for cybersources on getting into the field: what to learn in what order, where the practice comes from, and how to make the first reports count.Watch the recording
Mentor and author, UnihackersGuides on API security testing, request manipulation, target selection and prompt injection.Jun 2025 to now. About Unihackers
Bash scripting course, Ginius Cyber Security CouncilWrote and taught the council's bash scripting course, and stayed on as a mentor.Certificate
Guest speaker, web application penetration testingWebinar for Forensic Academy, the training centre where I had interned the month before.Sep 2023. Certificate

In the room rather than on the stage: BSides Ahmedabad 2025. Photo

Parth Narula

Parth NarulaSecurity researcher and penetration tester, India

Competitions

  • 1stCapture the Flag at TechFusion 2026, Vaish College of Engineering, Rohtak, October 2026. Team ScriptJackers, 9 of 9 challenges and 1900 points. Scoreboard and certificate
  • 2ndData Detectives at TechFusion 2026, on AI attacks and prompt injection. Certificate
  • 2ndDrift and Draft at TechFusion 2026, on UI and UX design and web development. Certificate
  • AIR 21HackWithIndia 2026 Web Hacking Competition, 21st of 10,000. Public leaderboard and certificate
  • 1stRoot Breach, national level CTF at UTKANSH '26, NIT Jalandhar, March 2026. Certificate
  • 1stCyber War 2026, national level CTF by Quick Heal and CompTIA. Scoreboard and photo
  • 1stRCS 2026, national level CTF, Republic of Cyber Sentinels. Press coverage
  • 1stCyberthon 2026 by Play2Unite and Qubicsquare, April 2026. Certificate and trophy
  • 1stEncrypted CTF on ThunderCipher, November 2025. Certificate
  • 2ndXploit404 CTF, April 2026. Certificate
  • 3rdTrivarna 2.0, international CTF by Astitwam. Event certificate

CTFs as team lead of ScriptJackers.

Certifications

  • eWPTX v3Web Application Penetration Tester eXtreme, INE SecurityDec 2025
  • CEH v13Certified Ethical Hacker, EC-Council. No. ECC7154890236. CertificateNov 2025
  • eJPT v2Junior Penetration Tester, INE Security. ID 171442005. CertificateJan 2026
  • ACPAPIsec Certified Practitioner, APIsec University. Verify on CredlyMar 2026
  • Ethical HackerCisco Networking Academy. Verify on CredlyNov 2025
  • EHEEthical Hacking Essentials, EC-Council. No. 152849. CertificateFeb 2022
Every certificate in the archive

Education

BCA, CybersecurityLovely Professional University, PhagwaraAug 2025 to now. CGPA 9.75
Senior Secondary, CommerceVishvas Senior Secondary School, Hisar2023 to 2025. 86%

What I work with

Offensive security
VAPTWeb application pentestingAPI security testingAndroid pentestingNetwork pentestingApplication securityBroken access controlBusiness logic flawsBug bounty
AI and LLM security
AI red teamingLLM application securityPrompt injectionAgentic AIRAG pipelinesMCPAgent driven pentestingTool and output abuse
Cloud and systems
Cloud securityAWSAzureGCPIAMLinuxApache and NginxNetworking
Recon
OSINTAttack surface mappingGoogle dorkingAsset discoveryDark web researchCustom wordlists
Tools
Burp SuitePostmanWiresharkKali LinuxAndroid StudioGraphQLREST APIs
Code
PythonBashCC++PHPSQLJavaScriptHTMLRuby

Building

  1. Autonomous recon and pentesting, run by agents

    Started Jun 2026, in progress

    The replacement for the tool below, and a much larger idea. Several layers of models orchestrate the work, and instead of marching through one fixed methodology they decide what this target deserves. The judgement is mine, written down from years of reports; the agents call my own Python tooling and wordlists when they need them. I expect it to take six to twelve months.

    Private until it is worth reading. Happy to walk through the architecture in an interview.

  2. ScriptJacker, recon and scanning

    Python

    My first recon tool: 21+ modules that sweep a target and flag the obvious before I start looking properly. Still on GitHub, and the honest reason the project above exists.

    See the repository
  3. find_logs_leak

    Python

    Static analysis for Android reviews. It walks a decompiled app for logging calls that put sensitive values somewhere anyone with adb can read them.

    See the repository
  4. Reports and articles index

    Open collection

    Public reports, writeups and notes for web, API, Android, AI, network and cloud, sorted by bug class so you can read ten takes on one vulnerability in a row. The credit goes to the researchers who wrote them.

    See the collection
  5. CVE

    Record

    The CVEs I have been credited on, with the advisory and the proof of concept for each, kept in one place so nobody has to take my word for the count.

    See the repository
Everything else on GitHub

09Replies

What came back after the report

Collected through a feedback form after each engagement, or written publicly on LinkedIn. Only the people who gave permission are shown here.

Aleksandr TischenkoCEO, Lamantine SoftwareRe: Security reports

We've recently received several security reports from Parth Narula. The reports were accurate, and the discussions were constructive and fair. We're completely satisfied with the service and happy to have our security improved as a result.

Piotr KarczemskiBeatportRe: Bug bounty

Parth Narula has identified and reported multiple vulnerabilities in our services. I highly recommend him for his expertise, reliability and commitment.

Sutty teamCooperativa de Trabajo SuttyRe: Pro bono test

ScriptJacker has been a really nice hacker from the start. They reached out to us kindly asking to pentest our platform, and to our surprise, came back very shortly after our confirmation with three different and detailed reports. Our platform and users are safer thanks to ScriptJacker!

Jurgen Gaeremynpubliq vzwRe: Disclosure and project

Parth approached us after finding a vulnerability in one of our systems. His mail stood out by its clarity and detailed instructions to reproduce the issue. During a follow up project, he has gone the extra mile finding vulnerabilities in our systems. His capability to triage issues and clearly explain the business risk has earned our deep respect.

Richard AllenLeaveWizardRe: Security reports

We've received several security reports from ScriptJacker. The reports were accurate, and the discussions were constructive and fair. Great stuff, highly recommended.

Clement P.Co founder, DLX MediaRe: Project

We had a great experience working together. Communication was quick and thorough, with all of the information we needed to triage and get to fixing the bugs as soon as possible.

D. FeskensManaged hosting companyRe: Bug bounty

In our line of work we get daily reports of bug hunters using the same automations to detect the obvious. Parth went a step further and was able to detect a small security issue with one of our Git repositories. He disclosed this information to us in a professional manner and we were able to correct our mistake.

Dave LoweFounder and CEO, isiFederalRe: Proactive report

Parth was proactive in reaching out to us about some web security issues that he found. He gave us a thorough report.

Lain.laInfrastructure teamRe: Bug bounty

Parth is certainly persistent, I'll give him that. And sometimes, that's enough to stand out from the crowd, from those who get discouraged and give up chasing their dreams.

João Otávio StiviFounder and CEO, Square CloudOn LinkedIn

I highly recommend Parth as a talented and ethical security researcher. He approached Square Cloud via our responsible disclosure program to report vulnerabilities. Beyond his sharp technical skills in identifying and documenting the issue, what stood out was his exceptional professionalism and communication. He was thorough in his reporting, patient throughout the remediation process, and a pleasure to collaborate with. Any company looking to strengthen their security posture would be lucky to work with him.

Oliver J. MaicherCompliance and pentest consultantOn LinkedIn

Parth is one of my trusted experts in cybersecurity. He's highly recommended. Looking forward to working with you again.

Shridhar PatelAM (Systems), GCMMFOn LinkedIn

Thank you, your findings helped us. Keep doing good work!

Neyts ZupMayet RXRe: Contract work

Good pre start communication, promptly delivered, well made report.

11Contact

Let's find it before someone else does.

Hiring for a security role, planning a pentest or running a bug bounty program? Email me or book a 30 minute call.