June 2025 Summaries

The Langflow visual framework for building AI applications has a critical unauthenticated remote code execution (RCE) vulnerability, CVE-2025-3248, affecting versions prior to 1.3.0. Researchers at Trend Micro discovered that attackers can exploit this vulnerability by sending specially crafted POST requests to the /api/v1/validate/code endpoint, allowing arbitrary code execution. The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added CVE-2025-3248 to its Known Exploited Vulnerabilities catalog, indicating confirmed exploitation in the wild. Cloudsmith Enterprise Policy Management can help quarantine affected packages by detecting vulnerable versions with a CVSS score above 6 and addressing them before they reach production environments. Users can check if they are affected by searching for publicly exposed Langflow servers using tools like Shodan or GitHub Search queries. The vulnerability is exploited by attackers who use public proof-of-concept code to gain remote shell access, run bash commands, install malware, and launch DDoS attacks. This highlights the importance of auditing and securing everything that comes from public upstreams, such as PyPI packages, using upstream proxying and caching.

The text highlights the importance of securing system configurations in Continuous Integration and Continuous Delivery (CI/CD) pipelines, as misconfigured systems can create entry points for attackers. The focus is on hardening systems across the pipeline, not just credential hygiene. Cloudsmith's approach to secure configuration by default mitigates risks such as outdated Jenkins or GitLab runners with known vulnerabilities, artifact repositories exposing ports or APIs publicly, and poor logging setups. Cloudsmith's immutable infrastructure, fine-grained access controls, secure defaults, and audit trails provide a secure foundation for managing artifacts. The text emphasizes the need to scan infrastructure, review configurations, and consider using a hardened platform like Cloudsmith to harden CI/CD pipelines end-to-end.

CI/CD systems heavily rely on credentials to operate, but poorly managed or overly permissive credentials can create a complex web of access points that become prime targets for attackers. Common risks associated with insufficient credential hygiene include secrets being accidentally committed to code, overly-permissive credentials in pipelines, secrets embedded in container image layers, secrets printed to build logs, and unrotated, long-lived credentials. Real-world breaches caused by poor credential hygiene have been documented, such as the 2021 Travis CI security issue and Uber's two major breaches tied to credential mismanagement. To mitigate these risks, organizations must adopt a proactive and layered approach to secrets management, including mapping the credential landscape, classifying secrets by sensitivity and exposure risk, preferring ephemeral credentials, restricting credential usage context, preventing secret leaks in code, securing console output, and cleaning artifacts thoroughly.

Pulling Docker images from private registries like Cloudsmith presents a security challenge due to authentication management, network access, and trust across distributed systems. To address this, OpenID Connect (OIDC) can be used when pulling Docker images into Kubernetes, providing a layer of OAuth 2.0 that identifies who is making the request and verifying that requests come from a trusted source. This enhances security and simplifies credential management by using short-lived access tokens instead of long-lived credentials. An OIDC-based solution involves creating an Image Pull Secret with a Cloudsmith token obtained through OIDC, which can be automated using a Kubernetes CronJob to run at regular intervals and perform tasks such as retrieving the Kubernetes service account token, exchanging it for a Cloudsmith token via OIDC, and creating or updating an Image Pull Secret. This approach offers benefits including automatic rotation, simplified management, and alignment with security best practices.

The OWASP Top 10 for CI/CD Security Risks highlights Pipeline-Based Access Controls (PBAC) as a critical vulnerability, which refers to fine-grained permissions tied to the context in which pipelines execute. Insufficient PBAC can lead to lateral movement, data exfiltration, or malicious artifact injection, causing severe consequences such as exposing sensitive data or deploying compromised software directly into production. To mitigate these risks, modern access control tools like Open Policy Agent (OPA) and Rego are leveraged to enforce granular, well-scoped PBAC policies, which extend beyond user-level RBAC to encompass the entire execution environment, including secrets, network boundaries, and software artifacts. Policy-as-Code (PaC) approaches using Rego enable teams to define and enforce detailed access policies programmatically, ensuring consistency, auditability, automation, versioning, and scalability, with tools like OPA providing a scalable and auditable way to enforce access controls across infrastructure and CI/CD pipelines.

The software supply chain is under increasing attack, with attackers targeting earlier stages of the development process, particularly build pipelines. This has led to a critical shift in how security measures are implemented, with a focus on securing the artifact lifecycle from development through deployment. Key strategies include using Software Bills of Materials (SBOMs), digital signatures, and provenance data to provide visibility and control over dependencies. Docker Hardened Images offer a straightforward way to integrate trusted, verifiable container artifacts into modern workflows without slowing down development. Centralized artifact management platforms like Cloudsmith become essential in configuring repositories to proxy public registries, performing key security checks automatically, and providing enforcement points for policy. Ultimately, the goal is to create easy paths to security that are accessible, automated, and non-disruptive to gain widespread adoption.

The npm, PyPI, and RubyGems package repositories have been targeted by malicious actors, with over thirty packages affected across the three repositories. The attacks aim to steal cryptocurrency, source code, and other sensitive data, often through typosquatting or brandjacking tactics. Some of the specific targets include packages related to blockchain platforms such as BSC and Ethereum, Excel to JSON converters, and AI services like Alibaba's AI labs. Cloudsmith has implemented a strong line of defense against these attacks, including policy management tools that can quarantine malicious packages before they infect software supply chains. Despite this, some customers may still be affected if they do not follow recommended setups or if the attacker finds a way to bypass security controls.

Kubernetes, the de facto platform for orchestrating containers, benefits from open standards that complement its implementation, ensuring vendor neutrality and system efficiency. The Open Container Initiative (OCI), a Linux Foundation project, focuses on standardized container formats and runtimes, with two core specifications: `runC` and the OCI Specification. OCI-compliant registries have become essential for storing and distributing Helm chart packages, with Cloudsmith being one such registry that supports OCI compliance. ORAS (OCI Registry As Storage) is a tool that enables pushing, pulling, and managing non-container artifacts in OCI-compliant registries, transforming registries into general-purpose artifact stores. Cloudsmith has announced full support for Helm OCI, enabling users to push and pull Helm charts as OCI artifacts, with improved display limitations removed. The company's support now allows users to push via legacy means and pull via OCI, offering greater flexibility and smoother transitions.

The webinar discussed the importance of securing public Kubernetes Helm charts, which are increasingly being used by teams without a full understanding of their security risks. The session covered practical techniques for analyzing and scanning Helm charts, including using tools like Trivy and helm-diff, as well as implementing proper upgrade strategies and understanding runtime behavior. Maintaining and upgrading Helm charts after installation is essential to keep Kubernetes workloads secure, stable, and compatible with the evolving ecosystem. Regular upgrades should be considered every few weeks for critical applications or quarterly for lower-risk workloads, while security patches should be applied immediately when relevant. The webinar also emphasized the need for vigilance in operators' behavior as much as chart design in securing Helm usage. Various tools were mentioned to scan Helm charts, including Trivy, KubeSec, Datree, Pluto, and helm lint. A combination of static analysis and runtime controls is recommended to verify that a Helm chart doesn't allow or initiate egress calls to unwanted external sources. The official Kubernetes CVE feed provides an auto-refreshing list of known vulnerabilities, while tools like Cilium with Hubble and Calico with Flow Logs can monitor real network traffic at the pod level. Deploying an API management server is also recommended to safeguard Kubernetes pods, especially for public or partner-facing services where access control and visibility are critical. The end goal of these sessions is to raise awareness around securing software packages throughout their lifecycle, providing practical education, actionable best practices, and tooling insights to support the community in building safer pipelines with confidence.