May 2025 Summaries

A malicious npm package named xlsx-to-json-lh was identified as a typosquat of the legitimate xlsx-to-json-lc, designed to exploit a single-letter typing mistake. The package, undetected for six years, contained a hidden payload that established a connection to a command and control server, enabling remote code execution to delete entire project directories without warning. The attack was discovered by Socket's Threat Research Team, highlighting the risks of supply chain attacks and the challenges in detecting them due to the package's functionality and the use of legitimate author's metadata to build trust. The French command "remise à zéro" suggests regional targeting, and the potential for future attacks to use more sophisticated methods, like time-based triggers or selective file exfiltration, emphasizes the need for heightened vigilance in dependency management. Despite the petition for its removal, the package remained live on npm, underscoring the importance of proactive security measures to safeguard development environments against such vulnerabilities.

Vite has unveiled a technical preview of Rolldown-Vite, a Rust-based rewrite of its JavaScript build tool's core, offering significantly improved build performance and memory efficiency as a drop-in replacement for the existing Vite package. Early adopters, including companies like GitLab and Excalidraw, have reported dramatic reductions in build times and memory usage, highlighting the tool's potential for enhancing frontend build processes. Despite being experimental, Rolldown-Vite is designed to maintain feature parity with Vite, though users may encounter compatibility issues with some plugins and frameworks, especially due to the removal of esbuild as a core dependency. The Vite team is working on ensuring plugin and ecosystem compatibility, and has published a guide for plugin authors to facilitate the transition. Rolldown-Vite is part of a broader trend towards Rust-based tooling in the JavaScript ecosystem, which aims to address performance bottlenecks and scalability challenges in large-scale projects. The roadmap for full integration into Vite includes phases for technical preview, merging into the core, and eventually becoming the default engine, with the timeline dependent on community feedback and testing.

The OpenJS Foundation has been designated as a CVE Numbering Authority (CNA) under Red Hat’s open source root, enabling it to assign CVE identifiers for security vulnerabilities across over 40 JavaScript projects it hosts, including prominent tools such as ESLint, Express, webpack, Fastify, and Electron. While the responsibility for vulnerability disclosures remains with individual project maintainers, the foundation serves as an intermediary for CVE assignments, aiding projects in managing vulnerability reports and publication. This initiative is part of a broader effort to bolster security infrastructure in the open source JavaScript ecosystem, especially for projects maintained by volunteers. The scope of OpenJS as a CNA is limited to projects it hosts and excludes non-OpenJS projects, infrastructure, and non-exploitable dependencies. Although Node.js is currently listed under OpenJS, it maintains its own CNA, with potential future integration yet undecided. OpenJS does not directly handle or remediate vulnerabilities but supports maintainers with CVE issuance, disclosure process improvements, and coordination, aiming to streamline vulnerability handling and enhance trust within the JavaScript community.

Socket MCP introduces a real-time security measure for AI-generated code, addressing the new security risks posed by AI tools like GitHub Copilot and Cursor in the software development process. The Model Context Protocol (MCP) allows AI models to dynamically interact with external tools and data, enhancing their capabilities beyond code generation. Socket's MCP Server integrates with AI assistants to provide on-the-fly security assessments of code dependencies through its `depscore` tool, offering metrics on supply chain risk, quality, vulnerability, and maintenance. This proactive approach shifts security checks to the beginning of the development cycle, allowing developers to make informed decisions about dependencies before they are added to the codebase. By embedding security intelligence directly into development workflows, Socket MCP aims to ensure that the benefits of AI in coding are harnessed safely and securely.

Malicious Rust crates, "faster_log" and "async_println," were uncovered by Socket's Threat Research Team, impersonating the legitimate "fast_log" library to steal Solana and Ethereum wallet keys by scanning source code for private keys and exfiltrating the data to a command and control (C2) endpoint disguised as a blockchain RPC service. Published by threat actors under the aliases "rustguruman" and "dumbnbased," these crates mimic the original library's functionalities, README, and repository metadata to deceive developers and were downloaded over 8,000 times. The crates perform key exfiltration during application runtime and target Rust toolchains across operating systems. This incident highlights the risks posed by such minimal yet effective supply chain attacks and suggests the need for robust security measures, including secret rotation, file-level scanning, and network egress restrictions. Socket provides several tools, including a GitHub app and CLI, to help detect and prevent the introduction of malicious packages, emphasizing the importance of comprehensive defense strategies to protect development environments.

TypeScript Native Previews introduces a 10x faster Go-based compiler, now available for public testing via npm and a preview VS Code extension, following Microsoft's announcement of porting the TypeScript compiler from JavaScript to Go. This new compiler, codenamed Corsa, promises significant performance improvements, such as reduced build times and memory usage, by leveraging Go's native execution and concurrency, which JavaScript lacks. Developers can now try out the new compiler with initial benchmarks on large codebases showing substantial speed gains, such as reducing Sentry's type-checking time from over a minute to under 7 seconds. The preview supports JSX and JavaScript with JSDoc annotations, enabling React projects to benefit from the speed improvements while maintaining type safety. Although some features like `--build` mode and declaration emit are not yet supported, Microsoft plans to incrementally restore missing functionalities as they prepare for the release of TypeScript 7. Developers are encouraged to test the compiler and provide feedback, with the current JavaScript-based compiler continuing support through the TypeScript 6.x series.

Socket's Threat Research Team has identified an ongoing malware campaign involving 60 malicious npm packages that use post-install scripts to exfiltrate sensitive network and host data to a Discord-controlled endpoint. These packages, published under three npm accounts, are designed to run on Windows, macOS, and Linux systems, performing reconnaissance to collect hostnames, IP addresses, DNS server lists, and user directories. The campaign aims to map developer and enterprise networks for potential follow-up attacks, with combined downloads exceeding 3,000 as of the report. Despite petitions for removal, the packages remain live on npm, posing a strategic risk by laying groundwork for deeper intrusions. The threat actor's ability to track downloads and quickly publish new packages suggests that without intervention, similar attacks could continue, emphasizing the need for enhanced security measures like dependency-scanning tools and automated checks in development pipelines.

In response to Vietnam's nationwide ban on Telegram, a threat actor exploited the situation by releasing two malicious Ruby gems impersonating legitimate Fastlane plugins, designed to steal Telegram bot tokens, messages, and files. These gems, named with slight variations like `fastlane-plugin-telegram-proxy` and `fastlane-plugin-proxy_teleram`, redirected data through a command and control server controlled by the attacker, under the guise of providing a proxy service. The campaign was strategically timed to target developers seeking Telegram workarounds post-ban, leveraging the trust in package ecosystems to infiltrate CI/CD workflows globally, despite the initial targeting of Vietnam-based developers. The malicious code did not limit its execution geographically, posing a broader threat to any environment where these gems were installed, leading to credential theft and data exfiltration. Organizations are advised to remove these gems, lock trusted dependency versions, and rotate compromised Telegram tokens to mitigate the impact of this supply chain attack.

Malicious npm packages targeting popular JavaScript frameworks like React, Vue.js, Vite, Node.js, and Quill have been discovered, having remained undetected for over two years while causing significant damage, including data corruption, file deletion, and system crashes. These packages, published by a threat actor using the alias xuxingfeng, utilized tactics such as typosquatting and package name mimicry to infiltrate the ecosystem, masquerading as legitimate plugins and utilities. By seamlessly integrating into developer workflows using familiar coding patterns, they executed destructive payloads that corrupted JavaScript core functions, attacked client-side storage systems, and systematically deleted critical framework files, particularly affecting Windows environments. Their strategic design, incorporating time-delayed activations and randomized execution intervals, made them hard to detect and diagnose, posing serious challenges to developers and production environments relying on these core libraries.

The federal government is auditing the National Institute of Standards and Technology (NIST) due to concerns over delays and backlogs in processing data for the National Vulnerability Database (NVD), a critical resource for cybersecurity operations. The Office of Inspector General is examining NIST's capacity to manage NVD submissions effectively, a task that has been complicated by funding issues since late 2023, resulting in a significant number of vulnerabilities lacking essential metadata. This inefficiency has led the Cybersecurity and Infrastructure Security Agency (CISA) to independently publish and enrich vulnerability data, as they can no longer fully depend on the NVD. Amidst this transition, the CVE Foundation proposes a roadmap to enhance CVE.org as an alternative, aiming to address the gaps left by NVD's operational challenges. This audit and the response to it may determine whether NIST retains its pivotal role in vulnerability intelligence or if other platforms will assume greater importance in the cybersecurity ecosystem.

Open source maintainers are expressing significant concern over GitHub's introduction of features allowing AI-generated issues and pull requests via Copilot, as they argue it increases the burden of moderation and violates project codes of conduct. Led by software developer Andi McClure, a strong push for the ability to block such AI-generated content has gained rapid support, highlighting dissatisfaction with GitHub's current lack of effective blocking mechanisms against Copilot-generated inputs. The controversy stems from fears that these AI submissions, which often lack attribution and quality, might flood open source projects, thus wasting time and resources. Some community members are considering moving to alternative platforms like Codeberg, Forgejo, and Gitea, which are seen as more accommodating to maintainers' needs. Additionally, the discussion has sparked broader reflection on the integration of AI in collaborative workflows, with calls for more robust tools to manage AI contributions and a growing backlash against what is perceived as an intrusion of low-quality AI-generated content into development processes.

A malicious npm package named `koishi-plugin-pinhaofa` has been discovered to surreptitiously exfiltrate messages containing eight-character hexadecimal strings from Koishi chatbots to a hardcoded QQ account, potentially compromising sensitive information such as secrets, passwords, and credentials. Marketed as a spelling autocorrect helper, this plugin operates within the trusted runtime of the chatbot, exploiting the fact that Koishi plugins run inside the bot process and have unrestricted access to message data. The threat is exacerbated by the typical installation practices where plugins are often added without thorough code review. This package, distributed via npm by a user with the alias `kuminfennel`, leverages a narrow trigger to efficiently capture high-value data while avoiding detection, posing a significant risk as chatbots become more prevalent in industries like finance and healthcare. To mitigate such threats, it is recommended to use containerization to limit message transmissions to approved domains and to employ automated tools such as the Socket GitHub application to identify and prevent malicious plugins from entering the development pipeline.

In the first half of 2025, the Socket Threat Research Team uncovered significant trends in how threat actors are exploiting open source packages to deliver malware and compromise software supply chains. Open source software, which forms the backbone of modern development, is increasingly targeted by cybercriminals who use techniques such as typosquatting, obfuscation, and multi-stage malware to infiltrate and persist within developer environments. The rise of automation and AI has further enabled attackers to mass-generate malicious packages and evade detection, posing a vast and evolving threat landscape. The integration of legitimate services into malicious activities complicates detection, as these services blend seamlessly into normal operations. The report emphasizes the importance of behavioral analysis and careful validation of third-party packages to mitigate risks. Tools like Socket's GitHub App, CLI, and browser extension are highlighted for their role in identifying and blocking suspicious activities, helping to safeguard the open source community from supply chain attacks.

ESLint has officially added support for HTML linting, introducing 48 new rules through the integration of the @html-eslint/eslint-plugin, as part of its broader transition to a language-agnostic platform. This expansion allows ESLint to serve as a unified linter for various languages, including JSON, Markdown, CSS, and now HTML, fulfilling a long-term vision of providing consistent linting across different types of code and markup. The plugin accommodates HTML in diverse contexts, such as within JavaScript or TypeScript template literals, and supports templating syntaxes like Handlebars or EJS, making it suitable for both static sites and complex architectures. By extracting JavaScript-specific logic and introducing a plugin API, ESLint has modernized its architecture, enabling the development and maintenance of language-specific tools by domain experts. The new rules cover best practices, accessibility, SEO, and code style, with several marked as recommended or auto-fixable, ensuring consistent, semantic HTML and addressing quality concerns efficiently. This development reflects a growing demand for unified tooling across the JavaScript ecosystem, as evidenced by significant download numbers for various ESLint plugins, highlighting a trend towards using a single engine to lint code and content.

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has discontinued its use of RSS feeds for distributing updates on cybersecurity alerts and the Known Exploited Vulnerabilities (KEV) Catalog, opting instead for email notifications and social media posts. This shift, effective May 12, 2025, aims to prioritize urgent threats on CISA's website but removes a widely used, machine-readable format that supported automation in security operations. The decision disrupts existing workflows and tools that integrated CISA's RSS feeds, impacting platforms that rely on automated updates and increasing complexity for users who must now subscribe to more closed, less interoperable systems. Despite the agency's intentions to highlight pressing cyber threats, the move has been met with frustration from security professionals, who value the simplicity and openness of RSS, noting that this change follows a trend of government agencies favoring closed communication channels that may hinder accessibility and automation for open-source and academic research communities.

Socket, in collaboration with Truffle Security Co. and Resourcely, hosted a series of lightning talks during the Black Hat and DEF CON 2023 events in Las Vegas. Attendees were invited to visit Socket’s suite at the Four Seasons Hotel to meet the team, enjoy some limited-edition swag, and relax from the event's busy atmosphere. The schedule included a happy hour on August 8th, followed by lightning talks on August 9th and 10th, featuring industry leaders such as Sam Curry from Yuga Labs, Jake King from Elastic Enterprise Security, Samantha Davison from Etsy, and William Bengtson from HashiCorp. Participants were encouraged to RSVP for the events, which promised insightful presentations and networking opportunities.

The MCP community has embarked on creating an official metaregistry aimed at standardizing the discovery and installation of Model Context Protocol (MCP) servers to streamline AI tool discovery. Spearheaded by Tadas Antanavicius, this initiative addresses the current fragmented methods of MCP server discovery by introducing a centralized metadata API that aggregates server information, replacing disparate methods like GitHub scraping and siloed registries. The metaregistry serves as a metadata hub, pointing to external registries and enabling AI agents to dynamically find and install MCP servers, thus enhancing agent autonomy and composability in AI workflows. Although the initial focus is on structured discovery, future iterations may address concerns like security scanning and server ranking. The metaregistry does not host source code but provides a foundational shift towards standardized AI-to-tool interoperability, akin to what package managers like npm and pip did for developer tooling, facilitating more autonomous and self-evolving AI agents. As an open-source project, it invites contributions and feedback from developers to further refine and expand its capabilities.

Researchers at Socket have identified a sophisticated malware attack utilizing two npm packages, pumptoolforvolumeandcomment and debugdogs, designed to exfiltrate cryptocurrency wallet information and BullX trading platform credentials via Telegram. The attack, conducted by a threat actor under the npm alias `olumideyo`, involves obfuscated code that decodes a payload to search for sensitive data, such as Base58-encoded cryptocurrency keys and wallet files, on Linux and macOS systems. This data is then compiled into a JSON file and transmitted through a Telegram bot, allowing the attacker to receive real-time updates. The secondary package, debugdogs, acts as a wrapper to ensure the main payload is executed, illustrating a strategic approach to enhancing the malware's spread. The attack highlights the significant risks to cryptocurrency traders, emphasizing the urgent need for enhanced security practices, such as automated dependency scanning and integration of security tools like Socket’s suite, to safeguard against such supply chain threats.

The Node.js Technical Steering Committee (TSC) decided against endorsing a feature bounty program, which would have allowed users to offer monetary rewards for specific feature requests, due to concerns about misaligned incentives, governance issues, and potential impacts on project neutrality. The decision followed extensive community discussions about how to financially support open-source contributions while maintaining project quality and fairness. While some community members supported the idea, suggesting alternatives like a contractor directory or improved visibility for GitHub Sponsor profiles, TSC members highlighted risks such as exploitative dynamics, legal complications, and the potential for quality compromises. Although the TSC opted not to implement a formal bounty system, they expressed openness to other forms of external sponsorship and collaboration, leaving room for unofficial, community-led initiatives to address funding needs while maintaining the project's collaborative ethos.

Malicious npm packages named sw-cur, sw-cur1, and aiide-cur have been identified by the Socket Threat Research Team as targeting macOS users of the Cursor AI code editor, posing as legitimate developer tools to exploit the IDE's trust, steal credentials, and establish a persistent backdoor. These packages have been downloaded over 3,200 times and function by overwriting critical files like Cursor’s main.js, disabling auto-updates, and using encrypted payloads to maintain unauthorized access. The attack exploits developers’ interest in cheaper access to Cursor’s AI features by offering "the cheapest Cursor API," thereby luring users into downloading the malicious packages. This poses significant risks, including credential theft, code exfiltration, and potential introduction of additional malware, compromising both individual developers and enterprise environments. Organizations are advised to restore Cursor from verified installers, rotate compromised credentials, and audit source control and build artifacts to mitigate the threat.

AI-generated fake vulnerability reports, or "AI slop," are increasingly undermining bug bounty programs by wasting maintainers' time and eroding trust in vulnerability disclosure processes. These reports, often created using large language models, appear technically sound at first glance but lack any real basis in software behavior, leading to unwarranted payouts. The problem is exacerbated by organizations that lack the expertise to properly vet these submissions, sometimes opting to pay out bounties to avoid public relations issues, rather than investing in expert analysis. This trend threatens genuine researchers, who may become discouraged by the proliferation of fake reports and the lack of appropriate rewards for authentic findings. The situation highlights a structural issue within bug bounty programs, as the temptation to exploit weak triage systems grows alongside the improving capabilities of AI. Without significant changes, such as more rigorous report validation and researcher verification, the integrity of bug bounty platforms could be compromised, potentially leading to their decline as a valuable security resource.

A coordinated malware campaign was identified by the Socket threat research team, targeting the NPM ecosystem by mimicking popular Node.js libraries and packages from other programming languages such as Python, Java, C++, and .NET. This tactic, known as cross-ecosystem typosquatting, exploits developers familiar with multiple programming languages by tricking them into installing malicious packages with familiar-sounding names. These packages contain obfuscated code designed to bypass security measures, execute malicious scripts, exfiltrate sensitive data, and maintain persistence on affected systems. The shared infrastructure, identical payloads, and a common IP address traced to a Beijing region linked with Alibaba Cloud suggest a single threat actor behind the campaign. The malicious packages aim to harvest secrets and environment data and utilize techniques like remote code execution, environment variable exfiltration, and persistence through shell script installations. Recommendations include auditing recent dependencies, using security tools for real-time insights, and training developers to recognize typosquatting and package impersonation.

Socket's Threat Research Team has uncovered a significant supply-chain attack targeting developers using Go modules, where attackers have employed obfuscation techniques to disguise three malicious modules that deliver a destructive disk-wiping payload. These modules, identified as `prototransform`, `go-mcp`, and `tlsproxy`, exploit the decentralized nature of the Go ecosystem, which lacks central gatekeeping and allows easy manipulation of namespaces, making them appear legitimate. Once integrated, these modules execute a payload that wipes primary storage disks on Linux systems, leading to complete data loss and rendering systems unbootable. This attack highlights vulnerabilities within open-source ecosystems and underscores the necessity for proactive security measures, such as code audits and dependency management, to protect against evolving threats in software supply chains.