Aggregator

Tenable Research discovered a privilege-escalation vulnerability in Google Cloud Platform (GCP) that is now fixed and which we dubbed ConfusedComposer. The vulnerability could have allowed an identity with permission (composer.environments.update) to edit a Cloud Composer environment to escalate privileges to the default Cloud Build service account. The default Cloud Build service account includes permissions to Cloud Build itself, as well as to Cloud Storage, Artifact Registry, and more.

Cloud Composer is a fully managed workflow-orchestration service in GCP based on Apache Airflow that is used for scheduling and automating data pipelines.

Cloud Build is a fully managed continuous integration and delivery (CI/CD) service in GCP that builds, tests and deploys applications and containers at scale.

Cloud Composer uses Cloud Build to build packages, and that is exactly where attackers could have abused the process to escalate privileges.

Cloud Composer allows users to install custom PyPI packages in their environments. However, this functionality introduced a privilege escalation vulnerability due to how Composer interacts with Cloud Build. When a user specifies a custom PyPI package, Composer initiates a behind-the-scenes build process, and the Cloud Composer service account automatically provisions a Cloud Build instance in the user's project. This instance is attached to the default Cloud Build service account, a highly privileged identity with broad permissions to GCP services including to Cloud Build itself, as well as to Cloud Storage, Artifact Registry or Container Registry, and more. (Click here to learn more about the default Cloud Build service account permissions).

An attacker with the composer.environments.update permission could have abused the Cloud Composer service orchestration process to escalate privileges. The attack would have been executed by injecting an attacker-controlled malicious PyPI package into the victim’s Composer custom-package configuration: 

When Cloud Build installs this package in an attempt to build the environment, it uses Pip. 

But how would one have executed remote code by adding a package to the Composer service? Turns out that Pip automatically runs pre- and post-package installation scripts. This would have allowed an attacker to execute arbitrary code within the correlated Cloud Build environment by using installation scripts inside their malicious package, despite lacking direct control over Composer’s underlying service account. 

The privilege escalation would have occurred when an attacker injected code that accessed the Cloud Build’s metadata API. Because the build instance runs with the default Cloud Build service account, an attacker could have extracted and exfiltrated its token. With this token, the attacker would have gained control over a privileged service account, allowing further escalation across the victim’s GCP project. This attack was particularly dangerous because the attacker did not need direct access to the Composer’s service account or to Cloud Build’s service account—only the ability to update a Composer environment. By simply adding a PyPI package to Composer, they could have manipulated the trusted automation pipeline to escalate privileges beyond their original access level. To clarify the impact of the now-fixed vulnerability: gaining full ownership of the project from the default Cloud Build service account was well within reach. 

Previously, during update operations to perform PyPI module installations, Composer used the Cloud Build service account, which might have had broader permissions than the user performing the operation. After implementing the fix, Composer stopped using the Cloud Build service account and instead will use the Composer environment service account for performing PyPI module installations.

The fix has been rolled out to new Composer instances already (rel. notes), and existing instances should be updated to exhibit this behavior by April 2025 (rel. notes).

In addition, our findings led GCP to update parts of Composer’s documentation, such as the sections on Access Control, Installing Python Dependencies and Accessing the Airflow CLI.

The ConfusedComposer privilege-escalation vulnerability in GCP builds upon a broader attack class of vulnerabilities in cloud services that we call "Jenga®" . This attack vector is a variant of ConfusedFunction, another GCP privilege-escalation vulnerability we discovered last year, and exploits the somewhat-hidden cloud provider misconfigurations related to cloud services permissions to escalate privileges beyond intended access levels. This variant highlights how attackers can abuse interconnected services the cloud provider automatically deploys behind the scenes, as part of a service-orchestration process.

(JENGA® is a registered trademark owned by Pokonobe Associates.)

Threat actors are increasingly exploiting mavinject.exe, a legitimate Microsoft utility, to bypass security controls and compromise systems.  This sophisticated attack technique allows hackers to hide malicious activity behind trusted Windows processes. Mavinject.exe is the Microsoft Application Virtualization Injector, designed to inject code into external processes as part of Microsoft’s App-V environment.  Included by default since […]

The post Hackers Leverage Windows MS Utility Tool to Inject Malicious DLL Payload appeared first on Cyber Security News.

Cybersecurity researchers have recently uncovered a sophisticated new variant of the notorious Lumma InfoStealer malware, featuring advanced code flow obfuscation techniques designed to evade detection by security solutions. This latest iteration represents a significant evolution in the malware’s capabilities, with threat actors implementing multiple layers of obfuscation to conceal the malicious code’s true purpose and […]

The post Researchers Uncovered Latest Version of Lumma InfoStealer with Code Flow Obfuscation appeared first on Cyber Security News.

In a sophisticated onslaught targeting the open-source ecosystem, reports have emerged detailing several malicious npm packages that are nefariously exploiting the Telegram Bot API to install backdoors on unsuspecting developers’ Linux systems. This alarming trend has escalated concerns over the integrity of software supply chains, particularly in light of the platform’s open development architecture. Cybercriminals […]

The post Malicious npm Packages Target Linux Developers with SSH Backdoor Attacks appeared first on GBHackers Security | #1 Globally Trusted Cyber Security News Platform.

Proofpoint has unveiled the global availability of Proofpoint Prime Threat Protection, the human-centric cybersecurity solution that brings together previously disparate critical threat defense capabilities—protection against multistage attacks across digital channels, impersonation protection, and risk-based employee guidance and education—in a single integrated solution. As organizations face an overwhelming array of fragmented, siloed and reactive cybersecurity tools, Proofpoint Prime is the integrated solution that unifies threat defense and human risk management into seamless workflows that span the … More →

The post Proofpoint Prime unifies multistage attack protection across digital channels appeared first on Help Net Security.

The post New Pacu Module: <br>Secret Enumeration in Elastic Beanstalk appeared first on Rhino Security Labs.

A critical security vulnerability in Samsung’s One UI system has been discovered, exposing millions of users’ sensitive information through the clipboard functionality.  Security researchers have identified that Samsung devices running Android 9 or later store all clipboard content—including passwords, banking details, and personal messages in plain text indefinitely with no automatic deletion mechanism. Clipboard Data […]

The post Samsung One UI Security Flaw Exposes Users Data in Plain Text With No Expiration! appeared first on Cyber Security News.

A glaring vulnerability has come to light within Samsung’s One UI interface: the clipboard history function stores all copied text, including sensitive data like passwords and personal information, in plain text and retains it indefinitely, unless users manually delete it. For countless smartphone users, copying and pasting is a daily activity. Complex passwords, banking information, […]

The post Samsung One UI Vulnerability Leaks Sensitive Data in Plain Text With No Expiration! appeared first on GBHackers Security | #1 Globally Trusted Cyber Security News Platform.

There are now several public proof-of-concept (PoC) exploits for a maximum-severity vulnerability in the Erlang/OTP SSH server (CVE-2025-32433) unveiled last week. “All users running an SSH server based on the Erlang/OTP SSH library are likely to be affected by this vulnerability. If your application uses Erlang/OTP SSH to provide remote access, assume you are affected,” Ruhr University Bochum researchers, who discovered and reported the flaw, said. About CVE-2025-32433 Erlang/OTP SSH is a set of libraries … More →

The post PoC exploit for critical Erlang/OTP SSH bug is public (CVE-2025-32433) appeared first on Help Net Security.

Phishing attacks are not only more frequent but also more sophisticated, leveraging AI to craft highly convincing messages that bypass traditional security measures.​ 

The post Beyond Firewalls: Why Phishing Demands a People-First, Trust-Centric Response  appeared first on Security Boulevard.