AI-Powered Application Penetration Testing—Scale Security Without Compromise Learn More

Linux Kernel Exploit, GitHub RCE, and Canvas Cyberattack

This episode explores how every layer of the stack has become an attack surface — from a privilege-escalating Linux kernel flaw and a GitHub infrastructure RCE to a poisoned RubyGems supply chain, a trojanized vendor installer, and a ransomware hit on centralized education infrastructure.

Five stories this week, one thread: the attack surface isn't just at the perimeter. It's every layer your stack inherited and forgot about. The kernel under every Linux box, the pipeline under every code push, the registry behind every build, the installer from a vendor you already trust. Here's what stood out from the operator chair.

A reliable privilege escalation makes every initial access vector more dangerous. A Linux kernel flaw dubbed copy_fail gives any local user root in ten lines of Python, unmodified, across virtually every major distro shipped since 2017. The team's read: this changes how you score risk. A 7.3-severity initial access bug you might have deprioritized becomes a full-compromise chain when paired with something this reliable. The systems most at risk are the ones that can't come down for a patch window: hospitals, critical infrastructure, IoT devices nobody thinks to update. Segment them and assume the race has already started.

A single git push was all it took to reach GitHub's shared storage nodes. Whiz Research found that any authenticated GitHub user could trigger remote code execution on backend infrastructure with a single git push, reaching shared nodes housing millions of public and private repos. GitHub patched and validated in six hours with no evidence of exploitation. However, our team pushes back on treating that as a benchmark. Most organizations aren't wired the same way. When availability isn't the product, patch urgency gets weighed against business disruption, and that calculus looks different for everyone.

Package trust can't be static. Every dependency is a relationship that can change after you adopt it.A sleeper account on RubyGems spent seven months publishing legitimate packages before turning malicious to harvest credentials: environment variables, SSH keys, AWS credentials. You can't catch this with a point-in-time scan. The account name also surfaced a real cybersecurity firm in search results, providing just enough ambient legitimacy to avoid scrutiny. Continuous behavioral monitoring of your dependency tree is no longer optional.

A signed installer is not a safe installer. The official Daemon Tools vendor site distributed a trojanized installer dropping QuickRat, an implant that injects into notepad.exe and conhost.exe and beacons over QUIC. Signature validation failed to catch it; behavioral analysis is the layer that still applies. The telling read: burning a capable RAT on a niche tool suggests the consumer infections were collateral. Government and scientific sector targets in the attribution footprint look like the actual campaign.

Kaspersky found the official Daemon Tools vendor site distributing a trojanized installer dropping QuickRat, an implant that injects into notepad.exe and conhost.exe and beacons over QUIC to blend into normal traffic. Signatures passed. The tell was behavioral. The consumer infections look like collateral: the actual targets were government and scientific sector organizations where Daemon Tools was known to be in use, meaning the delivery vector was chosen for its access, not its volume.

Centralizing education doesn't create convenience — it creates leverage. Shiny Hunters hit Canvas during finals week, threatening to leak data tied to nearly 9,000 institutions. When Canvas went down, school stopped. Not because institutions failed to protect themselves, but because Canvas long ago stopped being a tool and became the infrastructure itself. The grade book is Canvas. The classroom is Canvas. There was no fallback because there was nothing else. Canvas is to education what Salesforce is to enterprise: the single system everything runs through, the one you can't rip out. The difference is that enterprises have the budget, legal exposure, and regulatory pressure to treat that dependency seriously. Education has the same concentration risk and none of the defenses. Compliance requirements exist, but they carry almost no enforcement teeth — which makes it structurally difficult to justify the security investment until something like this happens. Canvas held the data, Canvas was the target, and the schools and students were collateral. That's not a security failure. That's what centralized infrastructure looks like when it gets hit.

A three-day deadline means nothing if the clock started 180 days ago. CISA is weighing a proposal to cut federal KEV remediation deadlines from weeks to three days, driven by AI-accelerated exploitation compressing the gap between disclosure and weaponization. Our team's counterpoint: the three-day clock starts at disclosure, but by then the vulnerability may have been in the wild for 90 to 180 days already. GitHub patching in six hours feels like a gold standard, but finding a bug there is like finding one in hospital. You spot it fast because everything is spotless. Most organizations aren't working in those conditions, and a deadline without enforcement or the infrastructure to meet it is just pressure with nowhere to go.

Security Headlines:


Sean McMillan Headshot

About the speaker, Sean McMillan

Community Manager

Sean McMillan is Community Manager at Bishop Fox, focused on making complex security topics easier to understand and more interesting to follow. He holds a bachelor’s degree in Mass Communication and Media Studies from Arizona State University and brings over a decade of experience in podcasting, live hosting, and audience engagement. As host of Initial Access, he works with practitioners to explore how real-world attacks actually happen.


Dillon Sparks Bio Photo

About the speaker, Dillon Sparks

Senior Operator

Dillon Sparks is a Senior Operator at Bishop Fox, serving on the Threat Enablement Team with a focus on Attack Surface Intelligence and Emerging Threat Analysis. He applies deep expertise in offensive security, network exploitation, and systems analysis to help organizations understand and mitigate real-world risk across complex software and infrastructure environments.


Bfx25 John Untz Author Bio 1

About the speaker, John Untz

Senior Security Engineer, Exploit Developer

John is a Senior Security Engineer, Exploit Developer, where he focuses on reverse engineering emerging threats and developing advanced capabilities to protect our customers' attack surfaces. Prior to joining Bishop Fox, John served in a number of selectively manned US Air Force teams, and is a graduate of the NSA's Computer Network Operations Development Program (CNODP).


Sergio Villegas BF Headshot

About the speaker, Sergio Villegas

Senior Analyst II

Sergio Villegas is a Senior Analyst II in the Attack Surface Intelligence team at Bishop Fox where he is one of the lead researchers. His main areas of focus are emerging threats, attack surface mapping, and tactical lead generation. Sergio has over 11 years of experience in cybersecurity during which he has worked as a researcher and consultant to help companies improve their procedures, technologies, and techniques around threat intelligence and threat hunting.

This site uses cookies to provide you with a great user experience. By continuing to use our website, you consent to the use of cookies. To find out more about the cookies we use, please see our Privacy Policy.