Critical Red Hat Quay Flaws Allow Private Image Overwrites

Dominic Jainy stands at the forefront of modern infrastructure security, bringing a wealth of knowledge from the worlds of artificial intelligence and blockchain to the critical realm of container management. As systems become increasingly automated, the integrity of the container registry has emerged as the linchpin of the entire software supply chain. Today, he joins us to dissect the recent, alarming discoveries within Red Hat Quay—vulnerabilities that threaten the very foundation of private image security. We explore the technical nuances of these flaws, ranging from compromised cryptographic trust to subtle logic errors in database queries, and discuss what this means for the future of secure deployments in an era of rapid CI/CD integration.

In this discussion, we delve into the mechanics of two critical vulnerabilities that allow unauthorized access to private repositories, the dangerous persistence of planted security keys, and the ways in which simple character patterns in database queries can lead to massive data leaks. We also examine the broader implications for OpenShift environments and the specific, multi-layered defense strategies organizations must adopt to shield their internal registries from these sophisticated exploitation paths.

The discovery of CVE-2026-123710 has sent shockwaves through the community due to its staggering 9.8 CVSS score. Could you walk us through how an attacker can manipulate a registry to trust their own keys indefinitely, and why this bypasses the standard authentication barriers we rely on?

The gravity of this flaw stems from a fundamental breakdown in how the registry validates administrative requests for key rotation. Essentially, Red Hat Quay uses specific trusted keys to sign tokens for image pushes and pulls, but the key server fails to verify if a request to add a new key is actually intended for a legitimate rotation. An attacker can leverage the /v2/auth endpoint to obtain an anonymous token—which Quay signs with a trusted key—and then present that same token back to the key server to insert their own malicious key. This new, untrusted key can be set with no expiration date, meaning it effectively grants the attacker a permanent “skeleton key” to the entire kingdom. Once this key is in the system, it survives server restarts and allows the attacker to generate valid tokens for any repository, enabling them to read or overwrite private images at will. It is a chilling scenario because even if you change your passwords or rotate your robot tokens, that planted key remains active and lethal until someone manually identifies and purges it.

Beyond the key insertion, there is a second critical flaw involving robot account tokens that carries an even higher CVSS score of 9.9. How does a seemingly minor detail like an unescaped underscore in a namespace name allow one tenant to compromise another’s entire container environment?

This vulnerability, tracked as CVE-2026-110697, is a classic example of how a tiny oversight in database logic can lead to a catastrophic security failure. In the code responsible for listing robot accounts, the system builds an SQL LIKE pattern using the namespace name without properly escaping wildcard characters. In the world of SQL, an underscore is not just a character; it is a wildcard that matches any single character, which creates a massive loophole for tenant isolation. An attacker with the ability to create a namespace can strategically place an underscore so that their search results include the robot accounts of a completely different tenant. This allows them to steal decrypted tokens that were never meant for their eyes, giving them the power to push or overwrite images in a victim’s private space. The most disturbing part is that the victim does not have to take any action for this to happen; their data is exposed simply because the attacker chose a clever name for a new organization or namespace.

When we talk about attackers overwriting private container images, the consequences sound devastating for any modern CI/CD pipeline. What does the real-world impact look like when these malicious changes begin to propagate through an organization’s infrastructure?

The impact of an image overwrite is a nightmare for any DevOps team because it poisons the well from which every deployment draws. When an attacker replaces a legitimate image with a compromised one, that change is automatically pulled into every system, testing environment, and production server that relies on that specific tag. If Quay is being used as the integrated registry for OpenShift, these stolen credentials or hijacked images can be used to inject malware or backdoors directly into the heart of the CI/CD pipeline. We are not just talking about losing data; we are talking about the potential for attackers to harvest secrets, credentials, and sensitive proprietary code stored within those images. Furthermore, an attacker can use their access to delete the legitimate keys used by running Quay instances, which causes all normal push and pull operations to fail with authentication errors, effectively launching a denial-of-service attack that brings development to a grinding halt.

Given that these vulnerabilities affect every version of Quay 3 and Quay 5, and that even disabling anonymous access doesn’t provide full protection, what immediate actions must IT administrators take to secure their registries?

Administrators must act with extreme urgency because any instance reachable by untrusted clients is currently exposed, including those tucked away inside internal networks. The first line of defense is at the networking layer; you need to go to your proxy or load balancer and immediately block the /keys endpoint for every entity except for your most trusted internal components. Next, you must become a digital detective by scouring your audit logs for service_key_rotate and service_key_delete events, specifically looking for any entry where the expiration_date is listed as null. If you find any service keys that you do not recognize or that lack an expiry, they must be removed instantly to stop the ongoing unauthorized access. While these steps mitigate the immediate risk, they are temporary band-aids; the only true resolution is to install the fixed releases for the affected branches, as the attacker can simply re-add their keys until the patched logic is actually running.

What is your forecast for the evolution of container registry security over the next few years?

I expect we will see a massive shift toward “zero-trust” architectures for registry internals, where the assumption that an internal request or an authenticated user is “safe” is completely discarded. In the coming years, from 2026 to 2028, we will likely see registries move away from simple token-based trust and toward more robust, hardware-backed attestation and automated, short-lived cryptographic identities. The flaws we are seeing today in Red Hat Quay demonstrate that even mature products can harbor logic errors for years—sometimes since their first public commit—so the industry will likely lean harder on automated security auditing tools that can catch SQL injection patterns and key-management oversights before they reach production. Ultimately, the registry will no longer be viewed as just a storage bucket, but as a high-security gateway that requires the same level of scrutiny as a financial transaction system or a primary identity provider.

Explore more

How Does KB5132061 Affect Windows 11 Security and Performance?

Dominic Jainy stands at the forefront of the modern IT landscape, possessing a deep technical understanding of how artificial intelligence and blockchain are reshaping traditional operating systems. As an expert in systems administration, he has spent years navigating the complexities of Windows deployment and security hardening for enterprise environments. With the recent rollout of the November Patch Tuesday updates, Dominic

Why Was the Nvidia RTX 5080 Super Reportedly Cancelled?

Reports originating from the Board Channels trade platform indicate that Nvidia has officially notified manufacturing partners that the RTX 5080 Super project is no longer moving forward. This revelation has sent shockwaves through the enthusiast community, which had been eagerly anticipating a high-performance bridge between the standard 50-series offerings and the nearly unattainable flagship models. The high-end PC gaming landscape

How Can Enterprises Successfully Scale AI Beyond the Pilot?

Maintaining a representative set of test cases is essential because new software releases and evolving user behaviors can rapidly degrade model performance. While a compelling artificial intelligence demonstration can be built in just a few weeks, an enterprise-grade capability must earn the trust of its stakeholders every day through consistent accuracy and reliability. The fundamental shift from isolated experimentation to

How is the Holiday Hiring Model Changing in 2026?

The once-standard practice of flooding store aisles with temporary seasonal help has finally given way to a sophisticated, data-driven labor architecture that prioritizes long-term retention over short-term volume. In 2026, the retail industry is distancing itself from the familiar rhythm of the past, opting instead for a more calculated evolution in how companies view human capital during the year’s most

Security Implications of Distributed Cloud Networking

The sophisticated architecture of a modern digital enterprise resembles a sprawling, borderless city where data flows through invisible veins, bypassing the once-mighty castle walls of traditional security perimeters. In this highly interconnected landscape, the binary concept of “inside” versus “outside” has dissolved into a complex web of microservices, edge nodes, and multicloud environments. As organizations race to minimize latency and

Fill out the form below