Intigriti Bug Bytes #241 - October 2026 🚀

Hi hackers,

Welcome to the latest edition of Bug Bytes! In this month's issue, we'll be featuring:

• A full Microsoft Edge exploit chain using only logic bugs

• 1-click RCE in Discord via Twitch embed

• $250K Chrome full-chain bounty in just 6 days

• XSS in Cloudflare's image endpoint via SVG sanitizer bypass

• $90K Google RCE via Flask session pickle deserialization

And so much more! Let's dive in!

Understanding PHP2Shell: everything you need to know about CVE-2026-109618 & CVE-2026-109619

On October 21, emergency patches were released for two critical vulnerabilities in PHP's core engine. PHP2Shell (CVE-2026-109618 & CVE-2026-109619) carries a CVSS score of 9.2, and we've put together everything you need to know about PHP2Shell in one place: the four confirmed attack chains and the potential impact of each, three immediate steps your organization can take today, and long-term actions to strengthen your security posture.

Understanding PHP2Shell: Everything you need to know about CVE-2026-109618 & CVE-2026-109619

AI raised the floor. The hacker crowd keeps raising the ceiling.

Submission volume is still growing, and validity has recovered sharply. High-severity findings are increasing, and researchers are repeatedly returning. The community is not simply producing more, it is learning how to produce better.

In our latest blog, Head of Product Greg Jenkins explores the latest Intigriti data on AI and bug bounty trends, what this shift means for bug bounty, and how CrowdRecon is making the community's wider reconnaissance work visible.

AI raised the floor. The hacker crowd keeps raising the ceiling.

Customer story: how Ivo builds proactive security as a fast-growing AI company

The key lesson for high-growth companies? As soon as possible does not mean all at once. Ivo is a fast-growing Series B AI/SaaS company in the San Francisco Bay Area with a clear goal: to make proactive security a core part of how they earn and keep customer trust.

In this customer story, we sit down with Saurabh Gupta, Ivo’s Director of Customer Trust and GRC, to explore how flexible assessments fit their fast-moving culture, the steps behind building a private bug bounty program and launching a VDP, and why AI-specific security coverage matters in today’s environment.

What ID verification changes mean for public bug bounty programs at Intigriti

ID verification is moving earlier in the bug bounty journey. Chris Holt, our Strategic Engagement & Community Architect, discusses what ID verification changes mean for public programs, how requiring verification before submission can help reduce noise in triage queues, and why responsible disclosure should remain welcome even without verification.

What ID verification changes mean for public bug bounty programs at Intigriti

Intigriti 1026 CTF results are in

Halloween season is finally here! Our challenge creator went with a spooky Halloween store for this month's challenge and made it deliberately vulnerable to a price manipulation vulnerability. It was a fun CTF that was also more approachable than usual, which the solver count reflects. Here are the final numbers.

Quick recap:

• First blood was captured by mianx

• 411 hackers found the correct solution

• 72 hackers wrote a cool write-up

If you want to put your hacking skills to the test, be sure to give the Intigriti 1026 Challenge a go before heading over to Bugology, where you can find all the researchers' submitted write-ups and solutions.

Intigriti 1026 Challenge

Quick! Intigriti 1026 BRP Challenge is still live

Are you up for a bonus challenge? Six random solvers will win a limited-edition T-shirt. Head over to the challenge page and capture the flag before November 2, 2026, at 11:59 PM UTC.

Intigriti 1026 Bonus Challenge

Blogs & videos

Exploiting LLM prompt injection vulnerabilities

Exploiting LLM prompt injection vulnerabilities Cover Image

Prompt injection vulnerabilities are some of the most interesting bugs to hunt for right now. Extracting someone's email from a Reddit comment, leaking OTPs through an email reply, taking actions on a victim's behalf without ever touching a proxy interceptor: the attack surface is wide open, and most developers underestimate it. Our latest article breaks down the full LLM prompt injection exploitation guide, from basic system prompt overrides to advanced indirect injection scenarios that trick LLMs into complying with just about anything.

• Are you using AI to generate your reports? Be sure to feed it our AI vulnerability report-writing guide, which contains a list of practical tips to help you write proper vulnerability reports that get triaged faster.

• It is okay to disagree with the outcome of your report. Whether it is an incorrect closure, a severity downgrade, or a payout that takes longer than expected, knowing how to respond matters. Our guide covers Intigriti’s mediation process, the most common scenarios researchers run into, and the common mistakes that can further escalate the situation.

• Starting out with bug bounties can be overwhelming, especially with how fast the field moves. Our Bug Bounty Starter Kit is a free guide that walks you through recon, tooling, exploiting common vulnerability classes like SQLi, XSS, and BAC, and writing reports that get triaged faster. Get your copy now.

Tools & resources

Tools

HexStrike AI

HexStrike AI

Recon involves running multiple tools in order, which is a super tedious process if you have to do it for every target manually. HexStrike AI by @0x4m4 is an MCP server that lets any LLM (such as Claude, GPT, or Copilot) autonomously run 150+ recon tools for automated information gathering, fingerprinting, and vulnerability discovery.

• Looking for a way to add structured recon capabilities to your AI workflow? Claude-OSINT is a drop-in set of Claude skill files that gives you 100+ recon capabilities, 80+ dorks, secret-regex patterns, and 27 attack-path templates for red-team engagements and bug bounty workflows. If you are already using Claude for security testing, this is worth checking out.

• New CloudPEASS release. K8s-PEAS is a new Red Team Kubernetes script from the PEASS-ng project. You can use it to find the permissions of a compromised principal in a Kubernetes cluster during a red team engagement.

Resources

The Last Proof of Human Craft: forging a Microsoft Edge full chain with logic bugs only

Source: The Last Proof of Human Craft: forging a Microsoft Edge full chain with logic bugs only

@orange_8361 published a full Microsoft Edge exploit chain built entirely from logic bugs, with no memory corruption involved. In a landscape where browser exploitation has become synonymous with heap sprays and use-after-frees, this work stands out as a fundamentally different approach. As a bonus, he also flipped the chain and popped Microsoft Edge from calc.exe.

• 1-click RCE in Discord via a Twitch embed. The disclosure details a chain going from an attacker page through a Twitch rich embed, a parser gap in the manifest node, an inherited-origin Discord popout, a V8 RegExp UAF, an Electron IPC escape, WebDAV demand-start, and TOCTOU in the updater to reach cmd.exe. One of the coolest exploit chains disclosed this year.

• $250K Chrome full-chain bounty in 6 days. @Xbow published the details on how they chained three vulnerabilities into a full Chrome exploit, claiming Google's first $250,000 full-chain reward of 2026 for an attack affecting an estimated 1 billion users.

• Root RCE and full cloud takeover on a Meta service. @zonduu1 published his write-up on how he achieved remote code execution and full cloud takeover on Meta at their in-person event in Taiwan.

• XSS in Cloudflare's /cdn-cgi/image/ endpoint. @assetnote published a research article on chaining an SVG sanitizer bypass with a Cache-Control header trick to bypass the CSP and ultimately achieve XSS. The issue was reported to Cloudflare and has been fixed.

• How XBOW found a critical RCE in a U.S. Election Assistance Commission calculator. This write-up covers how an unauthenticated attacker could execute code as root inside the backend container. The flaw was reported under the EAC's vulnerability disclosure policy, validated, and fixed ahead of the November 3, 2026 federal midterm election.

• $90K for a Google RCE via Flask session pickle deserialization. @soeckly published the write-up on how he discovered a remote code execution vulnerability in Google's developer site.

• How 17 trillion Microsoft records were one request away. @efaav published a write-up of a broken access control vulnerability that led to 17 trillion records, including PII.

• Scalable Vector Grief. @garethheyes published Scalable Vector Grief on SVG-based attack techniques. Worth a read if you like diving deeper into SVGs.

• $25K+ from GitHub Marketplace app permission issues. @jeetbhdr published a write-up on how a malicious GitHub Marketplace app could have taken over repositories via GraphQL team admin permissions.

• Next.js Server Actions: broken access control, $15K in bounties. @skydesperados published an article covering his experience of discovering vulnerabilities across several companies over the past 6 months.

• $10K MFA bypass. This researcher found a multi-factor authentication bypass that resulted in a $10,000 bounty.

• $7,400 for a Painless script injection in HackerOne's GraphQL API. The HackerOne report demonstrates clean differential testing: a sort_query parameter passed unsanitized into Elasticsearch's Painless script engine.

• A JPEG, a race, and a ghost: breaking Discourse's image pipeline. @kevin_mizu published a research article on achieving XSS through cache poisoning via Discourse's image processing.

• An exposed log file, live admin sessions, and 50,000 records. @_CryptoCat published the write-up on replaying a leaked staff session token for read access to an API covering over 50,000 records containing full names, dates of birth, and passport numbers.

• Claude Code sandbox escape via .git worktree confusion. This write-up details how prompt injection could achieve unsandboxed code execution through worktree confusion in Claude Code.

• Hunting IDORs: an end-to-end methodology. This recent article walks through a complete IDOR methodology, from mapping ID-bearing endpoints to proving impact safely.

• Jev as a judge: how counterfeit policy steers Jev. @datadoghq Security Labs published a new research article on a method that dropped Jev's prompt-injection detection rate from approximately 83% to approximately 4%, and used the technique to bypass LangChain's experimental Jev-based tool-risk gating.

• From AI agents to RCE: building a vulnerability research workflow. @quarkslab published a new post on building an agentic vulnerability research workflow.

• Build your own AI bug bounty hunter. @mdp_sec shared a thread about building AI-assisted workflows for bug bounty.

• Hackbots at DEF CON 34: one 6-hour run, 4 confirmed criticals, 17 confirmed highs. @jhaddix published the full talk on how Arcanum built Armada, their in-house hackbot running on Claude Code, and every failure it took to get there.

• How to use Gemini CLI for bug bounty research. @yeswehack published a guide on using Gemini CLI to analyse evidence and validate findings.

• AI is great at finding leads but bad at judging severity. @busf4ctor published his insights on hacking games with AI.

• Local AI for pentesting and bug bounty. @ctbbpodcast released episode 196 where rez0 and gr3pme discussed limitations of AI agents, local AI, and what makes AI agents efficient.

• Top 43 AI security skills, MCP servers, and GitHub repos for bug hunters. The list compiles AI-focused security tools, all in one place.

• SSRF impact in cloud environments: beyond network reachability. The MSRFC team shared a new blog exploring how SSRF works in cloud environments, why cloud-hosted infrastructure amplifies its impact, and how MSRC assesses severity.

• Steal My Bitcoin CTF solved. Congrats to @killa, who solved the Stealmybitcoin CTF challenge around PHP2Shell. In doing so, he also shared his proof-of-concept.

• IDOR to Critical: the tutorial I wish I had when I started. @CristiVlad25 uploaded a new video explaining IDOR attacks, inspired by @chs's IDOR guide.

• I couldn't read the chat, so I invited someone who could. @tyrion4O4 published a write-up on an unconventional approach to accessing restricted content within his target.

• PHP2Shell: the technical deep dive. @calif_io published the technical analysis of the two PHP remote vulnerabilities covered in our blog above.

• 403 bypass through Azure Application Gateway. @4osp3l shared a cool, well-known technique where appending a JavaScript file extension bypassed a 403 Forbidden restriction through Microsoft Azure Application Gateway.

• SSRF payloads for local file read. @BRuteLogic shared SSRF polyglot payloads for local file reads/disclosures.

• SSRF is not always hidden in parameters. @GodfatherOrwa shared a tip on testing SSRF by placing the target server in the request path itself, with and without a leading slash.

• Two Firefox-only XSS payloads. @avlidienbrunn shared two interesting payloads using hidden="until-found" and onbeforematch on meta and input elements.

• AI agents everywhere, finding bugs, and Singapore. @aituglo published AituWeek 105, reflecting on going from backpacking through Asia to finding bugs with AI agents from a hotel in Singapore.

Company news

Wild West Hackin' Fest 2026 recap

Wild West Hackin' Fest is a wrap, and we were awarded the Best Booth trophy! A huge thanks to everyone who stopped by, took on our AI challenge, and chatted with us. See you in the next one!

Intigriti at Wild West Hackin Fest 2026

Hacky Hour London

Intigriti's Hacky Hour is heading to London! Come hack, meet our ambassadors and top researchers, and grab some food and drinks along the way.

Details:

📅 Wednesday, December 9

🕕 18:00 to 22:00 GMT

📍 Tapa Tapa, London 🇬🇧

Register now!

Intigriti Hacky Hour

Feedback & suggestions

Before you click away: Do you have feedback, or would you like your technical content to get featured in the next Bug Bytes issue? We want to hear from you. Feel free to send us an email at community@intigriti.com or DM us on X/Twitter, and we'll take it from there.

Did you like this Bug Bytes issue? Consider sharing it with your friends and tagging us along on X/Twitter, Instagram, or LinkedIn.

Wishing you a bountiful month ahead,

Keep on rocking!