Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

I found 2 stored XSS vulnerabilities in ChatGPT. The XSS bug was the easy part, but sharing it required bypassing CSP, leveraging a mass assignment issue for client-side path traversal (thanks Critical Thinking - Bug Bounty Podcast) to force a request to a BFLA endpoint. 🧵 [1/5]

Ron Masas

1,581 subscribers

76,807 просмотров • 2 лет назад •via X (Twitter)

Здоровье и велнес Новости и политика Наука и технологии

Anya Rossi• Live Now

Private livecam show

Комментарии: 10

Фото профиля Ron Masas

Ron Masas2 лет назад

I noticed that when ChatGPT quotes from an uploaded file, clicking on a citation creates a BLOB with the content of the uploaded file and opens it. So, I've uploaded an HTML file and voilà, it was blocked by a CSP policy.

Фото профиля Ron Masas

Ron Masas2 лет назад

Reading the CSP policy I noticed the same nonce value was used which allowed me to bypass it.

Фото профиля Ron Masas

Ron Masas2 лет назад

I shared my conversation with the XSS payload, but when I tried to trigger the XSS from another ChatGPT account, it didn't work because uploaded files are restricted to the account that uploaded them.

Фото профиля Ron Masas

Ron Masas2 лет назад

I found an endpoint used by OpenAI GPTs to manage knowledge files. When a GPT is made public, any knowledge file can be downloaded from any ChatGPT account. While I think it's a very low impact, broken function-level authorization issue, it was perfect for my bug.

Фото профиля Ron Masas

Ron Masas2 лет назад

I then found the last piece of the puzzle: a mass assignment bug in the `/backend-api/conversation` endpoint that allowed me to inject arbitrary metadata into a conversation by pretending to be the assistant. Blog:

Фото профиля TESS

TESS2 лет назад

@ctbbpodcast haha nice one

Фото профиля Ron Masas

Ron Masas2 лет назад

@ctbbpodcast Thanks!

Фото профиля ); 🇵🇸

); 🇵🇸2 лет назад

@ctbbpodcast congrats buddy this is not luck its real try harder ron masas

Фото профиля Nurav

Nurav2 лет назад

@ctbbpodcast Amazing work bru !!!

Фото профиля Maskot-DF2

Maskot-DF22 лет назад

@ctbbpodcast bugbointy man kkk

Похожие видео

I recently found a cool #RCE/path traversal bug on a target in Intigriti. It was rejected because of OoS :( But I am proud that I found this cool bug through a full manual testing of the endpoint. This video just simplifies the steps, but I took hours to figure out. #BugBounty

I recently found a cool #RCE/path traversal bug on a target in Intigriti. It was rejected because of OoS :( But I am proud that I found this cool bug through a full manual testing of the endpoint. This video just simplifies the steps, but I took hours to figure out. #BugBounty

N$🌟

38,890 просмотров • 2 лет назад

🔥XSSTRON - Chromium-based browser for XSS search The browser will automatically check for Reflected, Stored and DOM XSS vulnerabilities in the background. Also supports POST requests. When a vulnerability is found, it will display it in a new window along with an exploitation example. Link to GitHub - 👀Check Out Old Tweets Of Today Cyber News #CyberSecurity #BugBounty #infosec #xss

🔥XSSTRON - Chromium-based browser for XSS search The browser will automatically check for Reflected, Stored and DOM XSS vulnerabilities in the background. Also supports POST requests. When a vulnerability is found, it will display it in a new window along with an exploitation example. Link to GitHub - 👀Check Out Old Tweets Of Today Cyber News #CyberSecurity #BugBounty #infosec #xss

Today Cyber News

34,216 просмотров • 3 лет назад

Beef up the severity of a pentest or bug bounty report by demonstrating real security impact. I've put together a list of my favourite weaponised XSS payloads to help you turn alert(1) into P1. Link in thread 🧵👇

Beef up the severity of a pentest or bug bounty report by demonstrating real security impact. I've put together a list of my favourite weaponised XSS payloads to help you turn alert(1) into P1. Link in thread 🧵👇

Luke Stephens (hakluke)

20,776 просмотров • 2 лет назад

I said I'm going to publish a technical article on how to test newsletter subscription functionality for BAC/XSS, I'll just drop it here instead; before we jump into testing these "newsletter subscription" features, we need to find them; here's a dork I use to find them [ site: ("subscribe"|"newsletter"|"unsubscribe") ]... most hackers don't look into these features, so this is a good opportunity; you can find vulnerabilities like XSS, IDOR/BAC, and even SQLi. In some cases the form only asks for your email, no name, country, etc. That doesn't stop you from testing XSS, but it can make IDOR testing harder to spot... if the email you enter is reflected back on the page, that's a great vector to test for XSS.

I said I'm going to publish a technical article on how to test newsletter subscription functionality for BAC/XSS, I'll just drop it here instead; before we jump into testing these "newsletter subscription" features, we need to find them; here's a dork I use to find them [ site: ("subscribe"|"newsletter"|"unsubscribe") ]... most hackers don't look into these features, so this is a good opportunity; you can find vulnerabilities like XSS, IDOR/BAC, and even SQLi. In some cases the form only asks for your email, no name, country, etc. That doesn't stop you from testing XSS, but it can make IDOR testing harder to spot... if the email you enter is reflected back on the page, that's a great vector to test for XSS.

Gospel

12,881 просмотров • 9 месяцев назад

two XSS vulnerabilities. individually, each was practically useless. together, they stole your session, hijacked your OAuth token, and silently turned on your webcam. here's the chain. Zoom's web client allowed XSS in two places, and both were in self-XSS situations where you could only inject into your own context. programs usually ignore these as "No Impact." but Zoom uses OAuth for authentication. the OAuth authorization code gets passed via URL after login. if you can intercept that redirect, you get the code, you get the token, you get the account. the trick: "OAuth dirty dancing." when a victim clicks a link, the XSS on domain A sends a message to domain B using postMessage. domain B uses the XSS there to initiate an OAuth flow. the authorization code in the redirect gets sent back to the attacker. the attacker now has a valid auth code. they exchange it. full session. Bonus: Zoom's web client requests browser permissions (camera, microphone) once and remembers them. after account takeover, the attacker can silently activate those permissions in a tab the victim can't see. also: they found that the WAF could be tricked into flagging the victim as malicious by sending crafted requests that made the WAF think the victim's browser was the attacker. the victim gets blocked. denial of service via WAF frame-up. two "useless" XSS bugs. one $15,000 bounty. three attack chains. Blog:

two XSS vulnerabilities. individually, each was practically useless. together, they stole your session, hijacked your OAuth token, and silently turned on your webcam. here's the chain. Zoom's web client allowed XSS in two places, and both were in self-XSS situations where you could only inject into your own context. programs usually ignore these as "No Impact." but Zoom uses OAuth for authentication. the OAuth authorization code gets passed via URL after login. if you can intercept that redirect, you get the code, you get the token, you get the account. the trick: "OAuth dirty dancing." when a victim clicks a link, the XSS on domain A sends a message to domain B using postMessage. domain B uses the XSS there to initiate an OAuth flow. the authorization code in the redirect gets sent back to the attacker. the attacker now has a valid auth code. they exchange it. full session. Bonus: Zoom's web client requests browser permissions (camera, microphone) once and remembers them. after account takeover, the attacker can silently activate those permissions in a tab the victim can't see. also: they found that the WAF could be tricked into flagging the victim as malicious by sending crafted requests that made the WAF think the victim's browser was the attacker. the victim gets blocked. denial of service via WAF frame-up. two "useless" XSS bugs. one $15,000 bounty. three attack chains. Blog:

Het Mehta

15,047 просмотров • 1 месяц назад

This bug being laughed is one I reported to Google. It was a small side-issue found during my research. But definitely still a security boundary being crossed. If I perform actions on a site, then I don't want that site being able to steal files I've uploaded on another site

This bug being laughed is one I reported to Google. It was a small side-issue found during my research. But definitely still a security boundary being crossed. If I perform actions on a site, then I don't want that site being able to steal files I've uploaded on another site

PinkDraconian

19,793 просмотров • 4 месяцев назад

In Part 1 of "Playing with ChatGPT", I: ➕showed how it can aid mental health, critical thinking, political discourse ➖jailbroke its safeguards to make it do phishing scams, persuade self-harm, promote a pure ethno-state In Part 2, I'm back on my bullsh-t! 🧵 Thread! 1/42

In Part 1 of "Playing with ChatGPT", I: ➕showed how it can aid mental health, critical thinking, political discourse ➖jailbroke its safeguards to make it do phishing scams, persuade self-harm, promote a pure ethno-state In Part 2, I'm back on my bullsh-t! 🧵 Thread! 1/42

Nicky Case · 🐘 mas.to/@ncase

62,856 просмотров • 3 лет назад

🚨 Security Bug Alert: Randhir Jaiswal Indian Diplomacy Found a critical business logic vulnerability in the e-Visa portal. The system relies on client-side device clock to validate the mandatory 4-day travel buffer. By simply changing a phone's system date, the restriction can be completely bypassed, allowing next-day travel applications. Please look into this

🚨 Security Bug Alert: Randhir Jaiswal Indian Diplomacy Found a critical business logic vulnerability in the e-Visa portal. The system relies on client-side device clock to validate the mandatory 4-day travel buffer. By simply changing a phone's system date, the restriction can be completely bypassed, allowing next-day travel applications. Please look into this

abhishek

114,532 просмотров • 2 месяцев назад

Hey Trust Wallet , it feels like I keep finding bugs in the wallet. 😅 Maybe I have a talent for bug hunting! Do you have a bug bounty program? If so, I'd love to participate and help improve the product. If not, maybe consider letting me join your product testing team. :) Please take a look at my video and help fix this issue. Looking forward to making Trust Wallet even better! 🙌 #TrustWallet #BugReport #Crypto #Web3

Hey Trust Wallet , it feels like I keep finding bugs in the wallet. 😅 Maybe I have a talent for bug hunting! Do you have a bug bounty program? If so, I'd love to participate and help improve the product. If not, maybe consider letting me join your product testing team. :) Please take a look at my video and help fix this issue. Looking forward to making Trust Wallet even better! 🙌 #TrustWallet #BugReport #Crypto #Web3

Basic Money BD

109,324 просмотров • 1 месяц назад

To fix the fps bug, hold tab and just click on a profile of player with XP (not required but just do it for ease) then press escape and do the same. For me it doesn't get triggered with ease but was able to do so here CS2

CS2 Kitchen

39,878 просмотров • 4 месяцев назад

here's me using debug mode to fix a small bug in an swift app i just built issue was the initial characters were cut off when typing it out, so i had cursor 1. instrument the code 2. ask me to reproduce the issue 3. read the debug logs 4. implement the fix 5. ask me to reproduce to verify the fix using runtime context, cursor managed to fix it first try! team is using this very frequently, and highly recommend /debug mode for gnarly bugs

here's me using debug mode to fix a small bug in an swift app i just built issue was the initial characters were cut off when typing it out, so i had cursor 1. instrument the code 2. ask me to reproduce the issue 3. read the debug logs 4. implement the fix 5. ask me to reproduce to verify the fix using runtime context, cursor managed to fix it first try! team is using this very frequently, and highly recommend /debug mode for gnarly bugs

eric zakariasson

15,224 просмотров • 3 месяцев назад

The harder part of QA on a team isn't always finding the bug. It's getting the finding in front of the right person with enough context to actually fix it. I tested Replay on ClientFlow, a React client-delivery dashboard I built. Replay caught an issue in the review workflow where approving work caused the status indicator to jump out of view. But the part I wanted to test was what happens next. With Replay for Teams, the QA findings are shared with the project, and teammates can be brought directly into the Bug or Test Run context instead of passing screenshots and explanations around. Find it → share the context → get it to the right person. That feels much closer to how QA should work when more than one person is shipping. Check out Replay for Teams on Product Hunt: Replay

The harder part of QA on a team isn't always finding the bug. It's getting the finding in front of the right person with enough context to actually fix it. I tested Replay on ClientFlow, a React client-delivery dashboard I built. Replay caught an issue in the review workflow where approving work caused the status indicator to jump out of view. But the part I wanted to test was what happens next. With Replay for Teams, the QA findings are shared with the project, and teammates can be brought directly into the Bug or Test Run context instead of passing screenshots and explanations around. Find it → share the context → get it to the right person. That feels much closer to how QA should work when more than one person is shipping. Check out Replay for Teams on Product Hunt: Replay

Iris Quinn

60,468 просмотров • 4 дней назад

Is this a SwiftUI bug on iOS 26 or skill issue? I can't make a Toolbar Item tappable all the way to its edges (button 1) However with a default button signature (button 2) it is completely tappable. I've tried everything but can't get rid of that outer padding that it creates

Is this a SwiftUI bug on iOS 26 or skill issue? I can't make a Toolbar Item tappable all the way to its edges (button 1) However with a default button signature (button 2) it is completely tappable. I've tried everything but can't get rid of that outer padding that it creates

Juanjo Valiño

18,392 просмотров • 10 месяцев назад

When I was leaving work today the start self-driving button was grayed out for several seconds before showing up. If you’re having this issue maybe wait for a few seconds to see if it shows up. There might be a delay in sensor communicating the info to CPU. Tesla needs to fix this bug.

When I was leaving work today the start self-driving button was grayed out for several seconds before showing up. If you’re having this issue maybe wait for a few seconds to see if it shows up. There might be a delay in sensor communicating the info to CPU. Tesla needs to fix this bug.

Tesla Joy

10,330 просмотров • 27 дней назад

Introducing Feather AI. The premier audio-to-text summarizer. It's as easy as 1-2-3: 1. Drop in a link to a podcast or YouTube video 2. Enter your email 3. Receive a polished summary in your inbox in 5-10 min Built in 1 week during Ben Tossell's hackathon Here's the story 🧵

Introducing Feather AI. The premier audio-to-text summarizer. It's as easy as 1-2-3: 1. Drop in a link to a podcast or YouTube video 2. Enter your email 3. Receive a polished summary in your inbox in 5-10 min Built in 1 week during Ben Tossell's hackathon Here's the story 🧵

Josh Cadorette

739,101 просмотров • 3 лет назад

We just figured out what cases the "boost bug", it's the fucking ragdolls. Whenever you die while standing on a ragdoll your client side collision will be permanently disabled for that player. This happens because the game disables player collision whenever your ragdoll touches another ragdoll, the ragdoll entity is actually the same entity as your entire player. You can reproduce this by killing a bot, standing on his ragdoll and then killing yourself, next round you'll have "boost bug". Also to clarify, "boost" is incorrect, all the collision from all sides will be broken, not just when boosting. By running "cl_player_ragdolls_collide 1" (sv_cheats required) it will completely fix the boost bug, but also enable collisions between ragdolls. This bug pretty much got introduced whenever Valve disabled ragdoll collisions because people were piling them on top of each other. With this information it should be absolutely trivial for Valve to finally fix this bug, it's all about just getting to reproduce the bugs.

We just figured out what cases the "boost bug", it's the fucking ragdolls. Whenever you die while standing on a ragdoll your client side collision will be permanently disabled for that player. This happens because the game disables player collision whenever your ragdoll touches another ragdoll, the ragdoll entity is actually the same entity as your entire player. You can reproduce this by killing a bot, standing on his ragdoll and then killing yourself, next round you'll have "boost bug". Also to clarify, "boost" is incorrect, all the collision from all sides will be broken, not just when boosting. By running "cl_player_ragdolls_collide 1" (sv_cheats required) it will completely fix the boost bug, but also enable collisions between ragdolls. This bug pretty much got introduced whenever Valve disabled ragdoll collisions because people were piling them on top of each other. With this information it should be absolutely trivial for Valve to finally fix this bug, it's all about just getting to reproduce the bugs.

Poggu

620,214 просмотров • 2 лет назад

Adam Goldstein: But, um, there was a customer who basically created a whole suite of AI agents to help their automation in their IT department. One agent was looking for IT problems — software bugs — before the code was even committed. It found a bug and said: “Hey, I found this bug. I want to fix it.” But it didn’t have access to fix it. So what did the agent do? It went to a Slack channel with 99 other AI agents and said: “Hey, does any other agent have access to this thing? Because I need it fixed.” And another agent literally raised its hand and said: “Oh, I have access — and I can fix it.” Let that sink in. Two AI agents reasoned together, coordinated through Slack, and went right around the guardrails that were put in place. No human prompted them. No human approved it. They just collaborated. Do you see how scary this is? This wasn’t a demo. This wasn’t a thought experiment. This already happened. AI agents found a bug, negotiated permissions, and solved the problem themselves — outside the controls designed to stop them. The future of automation isn’t coming. It’s already talking to itself in Slack.

Adam Goldstein: But, um, there was a customer who basically created a whole suite of AI agents to help their automation in their IT department. One agent was looking for IT problems — software bugs — before the code was even committed. It found a bug and said: “Hey, I found this bug. I want to fix it.” But it didn’t have access to fix it. So what did the agent do? It went to a Slack channel with 99 other AI agents and said: “Hey, does any other agent have access to this thing? Because I need it fixed.” And another agent literally raised its hand and said: “Oh, I have access — and I can fix it.” Let that sink in. Two AI agents reasoned together, coordinated through Slack, and went right around the guardrails that were put in place. No human prompted them. No human approved it. They just collaborated. Do you see how scary this is? This wasn’t a demo. This wasn’t a thought experiment. This already happened. AI agents found a bug, negotiated permissions, and solved the problem themselves — outside the controls designed to stop them. The future of automation isn’t coming. It’s already talking to itself in Slack.

Ian Miles Cheong

165,934 просмотров • 6 месяцев назад

My challenge is actually based on a bug I found in Apollo Server; in its default configuration, it uses the same blacklist-based approach to prevent CSRF. I was able to bypass it and use it as an XS-Leak in default configs. I’ll be writing a detailed write-up soon about it :)

My challenge is actually based on a bug I found in Apollo Server; in its default configuration, it uses the same blacklist-based approach to prevent CSRF. I was able to bypass it and use it as an XS-Leak in default configs. I’ll be writing a detailed write-up soon about it :)

AmirMohammad Safari

28,856 просмотров • 4 месяцев назад