02/18/2026-02/25/2026 SolarWinds Patches 4 Critical Serv-U 15.5 Flaws, Cline CLI 2.3.0 Supply Chain Attack And More.

1. SolarWinds Patches 4 Critical Serv-U 15.5 Flaws Allowing Root Code Execution

SolarWinds has released updates for Serv-U file transfer software to fix four critical vulnerabilities that could allow remote code execution. All are rated 9.1 on the CVSS scale:

• CVE-2025-40538: Broken access control letting attackers create admin users and run code as root;
  -CVE-2025-40539 & CVE-2025-40540: Type confusion flaws enabling execution of native code as root;
  -CVE-2025-40541: Insecure direct object reference (IDOR) allowing native code execution as root.

Exploitation requires administrative privileges, though risk is medium on Windows, as services often run under less-privileged accounts. These issues affect Serv-U version 15.5 and are fixed in 15.5.4. SolarWinds hasn’t reported active exploitation, but past Serv-U flaws (e. g., CVE-2021-35211, CVE-2021-35247, CVE-2024-28995) were targeted by hackers, including China-based group Storm-0322 (formerly DEV-0322).

2. Malicious npm Packages Harvest Crypto Keys, CI Secrets, and API Tokens

Cybersecurity researchers have disclosed an active supply chain worm campaign, dubbed SANDWORM_MODE, leveraging at least 19 malicious npm packages to steal credentials and cryptocurrency keys. The malware exfiltrates system information, tokens, environment secrets, and API keys, propagating via stolen npm and GitHub identities. Core features include a polymorphic engine, hook-based persistence, USB and SSH propagation fallbacks, and an “McpInject” module that targets AI coding assistants (Claude, Cursor, VS Code) to harvest SSH keys, environment files, and LLM API keys from providers like OpenAI, Anthropic, and Cohere. The attack unfolds in two stages, with a delayed secondary stage performing deeper harvesting, worm-like spread, and full exfiltration. Some packages include sleeper components or kill switches, which remain off by default. Users are urged to remove affected packages, rotate tokens and CI secrets, and review workflows. The campaign mirrors recent malicious npm activity, including buildrunner-dev and eslint-verify-plugin, which deliver RATs and agents targeting Windows, macOS, and Linux.

3. Cline CLI 2.3.0 Supply Chain Attack Installed OpenClaw on Developer Systems

On February 17, 2026, a supply chain attack targeted the Cline CLI open-source package, installing OpenClaw—an AI agent—on developer and CI/CD systems via the malicious [email protected] release. The attacker exploited a prompt injection vulnerability in the Cline GitHub Actions workflow, stealing a long-lived npm publish token to publish the compromised version. The post-install script silently installed OpenClaw globally, giving it system-level permissions, persistent presence, and potential access to credentials. The package was downloaded roughly 4,000 times over an eight-hour window. No evidence of data exfiltration or additional payloads was found, but OpenClaw’s unauthorized installation posed serious security risks, particularly in CI/CD environments. The attack was mitigated by deprecating [email protected], revoking the token, and releasing [email protected]. The incident highlights critical supply chain security weaknesses and the dangers of AI-driven automation in software workflows. Users are advised to remove OpenClaw and rotate any exposed credentials.

4. Wormable XMRig Campaign Leverages BYOVD and Timed Kill Switch For Stealth

Researchers uncovered a wormable cryptojacking campaign spreading via pirated software to deploy a custom XMRig miner. The malware uses a BYOVD exploit (Bring Your Own Vulnerable Driver) and a time-based logic bomb to evade detection and boost Monero mining efficiency by 15–50%. At the core is Explorer.exe, a persistent state machine that switches roles—installer, watchdog, payload manager, cleaner—based on command-line arguments. Payloads, including the miner, watchdogs, and a vulnerable driver (WinRing0x64.sys), are embedded in the binary, decompressed to hidden files, and disguised as legitimate software. A circular watchdog ensures the miner restarts if terminated, even killing Windows Explorer to maintain activity. The malware also spreads via USB drives, copying itself and creating malicious shortcuts. A kill switch set for December 23, 2025, triggers cleanup, suggesting a limited operational window. The campaign highlights evolving malware tactics, combining social engineering, worm-like propagation, kernel-level exploitation, and AI-like persistence to create a resilient, high-performance cryptojacking botnet.

02/11/2026-02/18/2026 Malicious npm and PyPI Packages Linked to Lazarus APT, Ivanti EPMM Exploit And More.

1. CISA Flags Four Security Flaws Under Active Exploitation in Latest KEV Update

CISA has added four actively exploited flaws to its KEV catalog. The vulnerabilities include CVE-2026-2441, a use-after-free bug in Google Chrome that can enable heap corruption via a crafted webpage; CVE-2024-7694, an arbitrary file upload flaw in TeamT5 ThreatSonar Anti-Ransomware; CVE-2020-7796, an SSRF issue in Zimbra Collaboration Suite; and CVE-2008-0015, a buffer overflow in Microsoft Windows Video ActiveX Control that allows remote code execution.

Google recently confirmed in-the-wild exploitation of CVE-2026-2441. Meanwhile, threat researchers previously observed large-scale abuse of the Zimbra flaw, and Microsoft warned the 2008 bug has been used to spread malware such as the Dogkild worm.

Federal agencies must apply patches by March 10, 2026, to mitigate risks.

2. Patch Immediately: BeyondTrust Remote Code Execution Flaw Exploited in the Wild

BeyondTrust has released urgent updates to fix a critical remote code execution vulnerability (CVE-2026-1731, CVSS 9.9) affecting its Remote Support (RS) and Privileged Remote Access (PRA) products, with evidence of active exploitation. The flaw allows unauthenticated attackers to execute arbitrary system commands through specially crafted requests, potentially leading to full compromise, data theft, or service disruption.

The issue was discovered by Hacktron AI using AI-driven variant analysis and disclosed in January 2026. Internet scans from Shodan indicate about 11,000 exposed instances, many of them on-premise deployments that remain vulnerable until patched.

BeyondTrust automatically updated SaaS environments, but on-premise customers must apply patches or upgrade manually. Administrators are urged to verify systems quickly, as the vulnerability is easy to exploit and attractive to attackers seeking ransomware or lateral movement within enterprise networks.

3. Malicious npm and PyPI Packages Linked to Lazarus APT Fake Recruiter Campaign

Researchers from ReversingLabs have uncovered malicious npm and PyPI packages tied to a fake recruitment campaign attributed to the North Korea-linked Lazarus Group. The operation, dubbed “graphalgo,” has been active since May 2025 and targets JavaScript and Python developers with fraudulent cryptocurrency-related job offers. Attackers pose as recruiters on LinkedIn, Facebook, and Reddit, directing victims to GitHub “interview tasks” that secretly depend on malicious packages. Some packages, such as bigmathutils, built trust and gained thousands of downloads before being updated to deliver malware.

The campaign uses a multi-stage approach: creating fake companies and websites, distributing poisoned open-source dependencies, and installing remote-access trojans capable of executing commands, accessing files, and searching for cryptocurrency wallets.

Researchers say the activity shows the hallmarks of Lazarus operations, including staged payloads, delayed malicious updates, encrypted communications, and a modular design that allows attackers to rotate front-end infrastructure while maintaining the same backend systems.

4. 83 % of Ivanti EPMM Exploits Linked to Single IP on Bulletproof Hosting Infrastructure

Most exploitation attempts against a critical flaw in Ivanti Endpoint Manager Mobile (EPMM) have been traced to a single IP address hosted on infrastructure run by PROSPERO. According to GreyNoise, 417 exploitation sessions were recorded between February 1 and 9, 2026, with 83% originating from one source. The activity targets CVE-2026-1281 (CVSS 9.8) and CVE-2026-1340, vulnerabilities that enable unauthenticated remote code execution. Ivanti confirmed limited real-world compromises, and several European organizations—including the Dutch Data Protection Authority and the European Commission—reported targeting attempts. Researchers also observed the same host exploiting unrelated flaws in Oracle WebLogic, GNU InetUtils, and GLPI, suggesting automated scanning. About 85% of probes used DNS callbacks to verify vulnerable systems without deploying malware, consistent with initial-access operations. Security experts advise patching immediately, auditing internet-facing MDM systems, reviewing DNS logs, and blocking PROSPERO’s network ranges to reduce risk.

5. Dell RecoverPoint for VMs Zero-Day CVE-2026-22769 Exploited Since Mid-2024

A critical zero-day vulnerability (CVE-2026-22769, CVSS 10.0) in Dell RecoverPoint for Virtual Machines has been exploited since mid-2024 by a suspected China-linked threat cluster, UNC6201, according to researchers from Google Mandiant and Google Threat Intelligence Group. The flaw involves hard-coded credentials that allow unauthenticated attackers to gain root-level access, deploy web shells, and install backdoors such as BRICKSTORM and its newer variant GRIMBOLT. Attacks have primarily targeted North American organizations and appliances that often lack endpoint detection tools, enabling long-term persistence. Investigators observed techniques such as temporary “Ghost NICs” to move laterally and erase evidence. UNC6201 shares tactics with another China-linked cluster, UNC5221, though they are considered distinct. Separately, Dragos reported activity by Volt Typhoon compromising cellular gateways in energy sectors, then pivoting into operational technology networks, highlighting growing risks to industrial systems.

02/04/2026-02/11/2026 Critical SQLi Flaw, Microsoft Patches 59 Vulnerabilities, Critical n8n Flaws Disclosed Along With Public Exploits And More.

1. Fortinet Patches Critical SQLi Flaw Enabling Unauthenticated Code Execution

Fortinet has released security updates to fix a critical vulnerability in FortiClientEMS (CVE-2026-21643), rated 9.1 in severity, that could allow unauthenticated attackers to execute arbitrary code. The flaw is an SQL injection issue that can be exploited through specially crafted HTTP requests. The vulnerability affects FortiClientEMS 7.4.4, and users are advised to upgrade to version 7.4.5 or later. Versions 7.2 and 8.0 are not affected. Although there are no reports of active exploitation, Fortinet recommends applying patches as soon as possible. The flaw was discovered and reported by a member of Fortinet’s Product Security team. Separately, Fortinet recently addressed another critical vulnerability (CVE-2026-24858) affecting several products, including FortiOS and FortiManager. That issue, rated 9.4, has been actively exploited by attackers to create persistent admin accounts, modify configurations to enable VPN access, and exfiltrate firewall configuration data.

2. Microsoft Patches 59 Vulnerabilities Including Six Actively Exploited Zero-Days

Microsoft released security updates addressing 59 vulnerabilities, including six actively exploited flaws. Five issues are rated Critical, 52 Important, and two Moderate. The vulnerabilities include privilege escalation, remote code execution, spoofing, information disclosure, security feature bypass, denial-of-service, and cross-site scripting. The six exploited flaws affect components such as Windows Shell, MSHTML, Microsoft Word, Desktop Window Manager, Remote Access Connection Manager, and Remote Desktop. Some allow attackers to bypass security prompts or elevate privileges after gaining access to a system, potentially enabling malware deployment or credential theft. Researchers from Microsoft, Google Threat Intelligence Group, and others reported several of the issues, though details of exploitation remain limited. CISA has added all six vulnerabilities to its Known Exploited Vulnerabilities catalog, requiring U.S. federal agencies to patch them by March 3, 2026. Microsoft also issued Edge browser fixes, updated Secure Boot certificates, and introduced new security features aimed at strengthening default protections and improving user transparency and consent.

3. Threat Actors Publish Malicious dYdX Packages to npm and PyPI Repositories

Cybersecurity firm Socket has uncovered a supply-chain attack in which threat actors published malicious versions of dYdX client libraries to both npm and PyPI, targeting developers building cryptocurrency trading tools. The incident, detected on January 27, 2026, likely involved a compromised maintainer account. Affected packages included several versions of @dydxprotocol/v4-client-js and the PyPI package dydx-v4-client.

The malware was hidden in core files and executed during normal use. In npm packages, tampered functions stole seed phrases and device fingerprints, sending them to a typosquatted domain. The PyPI version was more severe, installing a remote access trojan that periodically contacted a command-and-control server and could run arbitrary code, enabling theft of credentials, source code, and other sensitive data.

dYdX warned users to isolate systems and rotate credentials. Developers are advised to audit dependencies, upgrade to safe versions, block known indicators, and monitor environments, as compromised packages could lead to wallet theft or full system compromise.

4. Open Source Security Gets AI Boost As Claude Detects 500+ Critical Issues

Anthropic says its latest model, Claude Opus 4.6, has demonstrated the ability to autonomously audit open-source software, discovering more than 500 previously unknown high-severity vulnerabilities in widely used libraries such as Ghostscript, OpenSC, and CGIF. All reported flaws were confirmed as real and have since been patched by maintainers.

In testing, the model operated in a virtualised environment with access to tools like debuggers and fuzzers but no detailed instructions. Researchers found it used reasoning similar to human security analysts, identifying patterns in code and past fixes to uncover subtle weaknesses that traditional fuzzing sometimes missed. Examples included memory-handling errors and overflows that could cause crashes or allow exploitation.

Anthropic is deploying the system to help maintainers find and fix vulnerabilities, highlighting AI’s growing role as a defensive security tool that can complement manual review. However, the company also warned that similar capabilities could be misused and said it is adding safeguards to reduce risks.

5. Critical n8n Flaws Disclosed Along With Public Exploits

Multiple critical vulnerabilities in the open-source workflow automation platform n8n, tracked as CVE-2026-25049, allowed authenticated users with permission to create or edit workflows to escape the sandbox and execute arbitrary code on the host server. Researchers found the flaws stemmed from weak sanitization and incomplete sandboxing of user-written JavaScript, enabling attackers to run system commands, access files, and steal credentials, API keys, and configuration data. In multi-tenant environments, the issue could also allow lateral movement to other tenants or connected cloud services.

Several security firms independently identified bypasses, showing that earlier patches were incomplete. The vulnerabilities were fixed in n8n versions 2.5.2 and 1.123.17, and users are urged to update, rotate encryption keys, and review workflows.

Although no active exploitation has been confirmed, researchers have observed large-scale scanning of exposed n8n systems, suggesting growing attacker interest and highlighting the importance of prompt patching and access controls.