Supply Chain Attack Hits AsyncAPI npm Packages

Security researchers from OX Security, SafeDep, Socket, and StepSecurity have flagged four compromised packages in the AsyncAPI npm namespace. The affected packages are @asyncapi/generator-helpers@1.1.1, @asyncapi/generator-components@0.7.1, @asyncapi/generator@3.3.1, and @asyncapi/specs (versions 6.11.2 and 6.11.2-alpha.1).

These packages were not just tampered with — they were weaponized to deliver a multi-stage botnet loader. The attack is a stark reminder that even well-maintained open-source projects can become vectors for malware when credentials fall into the wrong hands.

How the AsyncAPI Malware Worked

The malicious code was hidden in the package tarballs, not in the source repositories. This is a classic supply chain trick: the code looked legitimate on GitHub, but the published npm artifacts contained the payload.

Once installed, the loader executed a multi-stage process. It reached out to a command-and-control server to fetch additional payloads, which could include anything from DDoS bots to credential stealers. The researchers noted that the malware was designed to persist on the infected system, making it particularly dangerous for CI/CD environments where these packages are often installed.

Stage One: The Initial Payload

The first stage involved a script that ran during package installation (the “preinstall” hook). This script decoded a hidden blob and executed it, establishing a foothold on the machine.

Stage Two: Botnet Recruitment

The second stage downloaded a botnet client from a remote server. This client connected to a centralized controller, allowing the attackers to issue commands to all infected machines at once.

Who Was Affected and What to Do

Anyone who installed these specific versions of the AsyncAPI packages between the time of the compromise and the discovery is at risk. The affected versions are:

• @asyncapi/generator-helpers@1.1.1
• @asyncapi/generator-components@0.7.1
• @asyncapi/generator@3.3.1
• @asyncapi/specs@6.11.2 and 6.11.2-alpha.1

If you have any of these in your project, you should immediately audit your lock files, rotate any credentials that might have been exposed, and scan for unusual outbound network connections. The npm registry has since removed the malicious versions, but that doesn’t help if you already pulled them down.

Lessons for Open Source Maintainers

This incident underscores a few hard truths. First, publishing to npm requires more than just strong passwords — two-factor authentication is non-negotiable. Second, maintainers should consider using trusted publishers and signing their packages, so that even if credentials are stolen, the published artifacts can be verified.

For consumers, the lesson is to verify package integrity. Tools like npm audit and lockfile verification can catch some issues, but they won’t catch everything. A healthy dose of skepticism about new versions of critical dependencies is wise.

This isn’t the first time the npm ecosystem has been hit. Similar attacks have targeted other popular packages, and they won’t be the last. The open source supply chain security landscape is evolving, but so are the attackers.

Detection and Response

If you suspect you’ve been compromised, look for these indicators:

• Unexpected network connections from your build servers
• Unusual processes running under the Node.js runtime
• Files created in temporary directories that don’t belong to your project

The researchers have published hashes of the malicious artifacts, so you can cross-check your installed versions. If you find a match, treat it as a security incident — not just a bad dependency.

Botnet malware is particularly nasty because it’s designed to be quiet. It won’t crash your app or throw errors. It just sits there, waiting for commands. That’s why proactive monitoring of your build environment is essential.

In the end, this attack is a reminder that the software supply chain is only as strong as its weakest link. For AsyncAPI users, the fix is straightforward: update to the latest patched versions and review your security posture. For the rest of us, it’s a cautionary tale about the risks we all carry in our node_modules.

The Numbers Behind Oracle’s July 2026 CPU

Oracle has released its July 2026 Critical Patch Update (CPU), and the numbers are staggering. The quarterly security update includes 1,449 patches covering 1,434 unique CVEs across 334 products. It’s one of the largest patch drops in the company’s history.

For context, that’s roughly four times the number of vulnerabilities Oracle typically addresses in a single quarter. And here’s the kicker: a vast majority of these flaws were likely discovered by artificial intelligence, not human researchers.

According to Oracle, only a few dozen of the vulnerabilities were credited to external researchers. The rest were found internally — and the company has been open about its growing reliance on AI to accelerate vulnerability discovery.

What Products Got Patched?

The patch sweep touches nearly every corner of Oracle’s portfolio. Here’s a breakdown of the hardest-hit product families:

• E-Business Suite: 410 patches — the most of any product line
• Fusion Middleware: 355 patches
• Communications: 168 patches
• PeopleSoft: 84 patches

Other products in the update include Database Server, APEX, Autonomous Health Framework, Essbase, GoldenGate, NoSQL Database, SQL Developer, TimesTen In-Memory Database, and a host of industry-specific applications such as Financial Services, HealthCare, Hospitality, Retail, and Utilities.

Java SE, MySQL, JD Edwards, Siebel CRM, and Virtualization products also received fixes. If your organization runs any Oracle software, there’s a good chance it’s in this update.

Remote Exploitation and Critical Severity

Roughly 600 of the patches address vulnerabilities that can be exploited remotely without authentication. That’s a serious concern — these are the flaws attackers can weaponize without needing any prior access to your systems.

Hundreds of the security holes have been assigned a critical severity rating, meaning they could lead to complete system compromise if exploited. The risk profile here is about as high as it gets for an enterprise software vendor.

AI’s Growing Role in Vulnerability Discovery

Oracle revealed earlier this year that it has access to top-tier AI systems, including Anthropic‘s Claude Mythos and OpenAI‘s most capable models. The company is using these tools to speed up and sharpen vulnerability discovery and patching across its own software, Oracle Health, and the open source components it relies on.

The result? A patch cycle where AI found the vast majority of the flaws. This shift has big implications for how we think about software security. If AI can find thousands of vulnerabilities in a single quarter, what does that mean for the attackers using similar tools?

Why You Should Patch Now

Oracle’s patch release isn’t just a routine maintenance task. Threat actors actively exploit Oracle product vulnerabilities in their attacks. Recent examples include the exploitation of a PeopleSoft zero-day and a previously patched EBS vulnerability that hit Estée Lauder.

Organizations running Oracle software should treat this CPU as a priority. Install the patches as soon as possible — ideally within days, not weeks. The window between patch release and exploitation is shrinking, and AI is making it even shorter.

If you’re managing Oracle environments, this is also a good time to review your vulnerability management process and ensure you have visibility into all Oracle products in your infrastructure. The breadth of this update — 334 products — means there’s a real chance something in your environment is affected.

For more on recent Oracle security issues, check out our coverage of Oracle’s monthly security updates and the Zimbra critical vulnerabilities that were patched recently.

Meet LabubaRAT: A Trojan With a Familiar Face

Cybersecurity researchers have uncovered a previously undocumented remote access trojan that borrows a trusted name to get a foot in the door. Dubbed LabubaRAT, this Rust-based malware masquerades as NVIDIA software to blend into Windows environments and avoid raising eyebrows.

Blackpoint Cyber analysts Sam Decker and Nevan Beal published their findings today, describing the tool as something that “creates a reusable foothold for hands-on activity.” Once it lands on a machine, it’s built to profile the host and open a backdoor for attackers to poke around at will.

This isn’t just another piece of commodity malware. It’s a crafted tool with a specific purpose: persistence, stealth, and remote control. And it’s wearing a disguise that could fool even a careful user.

Why the NVIDIA Disguise Is So Slick

The trick is simple but effective. The malware’s files, icons, and even its internal strings are designed to look like legitimate NVIDIA components. Think driver updater utilities, GPU monitoring tools, or installer packages. To the average eye — and to many automated scanners — it looks like the real deal.

Impersonating a well-known vendor like NVIDIA gives the attackers several advantages:

• Trust factor: Users are far less likely to question a pop-up or process that appears to come from a brand they recognize.
• Bypassing filters: Some security tools whitelist signed or known-good software names, which this trojan exploits.
• Muddying the waters: Even if someone spots suspicious activity, the NVIDIA branding makes it harder to report or investigate.

It’s a classic social engineering move, but with a technical twist. The malware doesn’t just claim to be NVIDIA — it’s built to look the part at the binary level.

Inside the Rust-Based RAT’s Arsenal

LabubaRAT is written in Rust, a language that’s become a favorite among malware authors for its performance and memory safety. But the real story is what the trojan does once it’s running.

According to the analysis, the RAT can profile the host system, gathering details about the operating system, hardware, and installed software. That reconnaissance helps attackers tailor their next moves. From there, it establishes a command-and-control channel, giving the operator a persistent way to issue commands.

Blackpoint’s researchers describe it as a “reusable foothold.” That’s a key distinction. Many trojans are one-shot tools — they execute a payload and disappear. LabubaRAT is designed to stick around, letting attackers return to the compromised machine whenever they need to.

What Hands-On Activity Looks Like

Once the backdoor is open, the attacker can do a lot with it. Common follow-ups include:

• Dumping credentials from memory or stored browsers
• Moving laterally across the network to reach other systems
• Deploying additional payloads like ransomware or keyloggers
• Exfiltrating sensitive files before anyone notices

The initial infection vector isn’t fully detailed in the report, but the NVIDIA disguise suggests a few possibilities. Malicious downloads, fake update prompts, or even poisoned search ads could all be in play.

How to Spot LabubaRAT Before It Bites

Detection isn’t impossible, but it requires a shift in mindset. Signature-based antivirus tools will likely miss this one if they haven’t updated their definitions yet. Instead, defenders need to look for behavioral clues.

Watch for processes that claim to be NVIDIA software but behave oddly — unusual network connections, high CPU usage, or attempts to write files to suspicious directories. Pay attention to unsigned binaries that use NVIDIA branding. Legitimate NVIDIA software is digitally signed; if a file fails signature verification, that’s a red flag.

Blackpoint’s researchers recommend monitoring for the specific indicators of compromise (IOCs) they’ve published, which include hashes and network endpoints. For most organizations, though, the simpler advice is to treat any unexpected “NVIDIA” process with suspicion and verify it against an official source.

If you’re looking to harden your defenses, it’s worth reviewing your endpoint security best practices and making sure your team knows how to recognize social engineering attempts. A little skepticism goes a long way.

The Bigger Picture: Brand Impersonation Is on the Rise

LabubaRAT isn’t an isolated incident. Attackers have been increasingly borrowing trusted brand names to slip past defenses. From fake Zoom installers to counterfeit Slack updates, the pattern is consistent: impersonate something familiar, and users will let their guard down.

For security teams, this means staying current on the latest malware trends and threat intelligence. It also means training users to question anything that seems even slightly off, no matter how legitimate it looks.

As for LabubaRAT, the discovery is a reminder that the threat landscape keeps evolving. The tools get craftier, the disguises get better, and the only defense is vigilance.

If you suspect a system is compromised, isolate it immediately and bring in your incident response team. Don’t wait to see what the attacker does next — by then, it’s usually too late.