Five Eyes Warn of Exploited Cisco Catalyst SD-WAN Flaws; Claude Code RCE and API Key Theft Patched; CrowdStrike 2026 Global Threat Report

Cybersecurity agencies in the Five Eyes countries, Australia, Canada, New Zealand, the UK, and the US, are urging users to patch two actively exploited vulnerabilities in Cisco Catalyst SD-WAN. Attacks were detected by the Australian Signals Directorate (ASD), which noted that the threat actors "are compromising SD-WANs to add a malicious rogue peer and then conduct a range of follow-on actions to achieve root access and maintain persistent access to the SD-WAN." The first vulnerability is a high-severity path traversal flaw in the CLI of Cisco SD-WAN software that could lead to privilege escalation. The vulnerability was disclosed in September 2022, and at that time, Cisco released software updates to address the issue. The second vulnerability is a critical improper authentication flaw in the peering authentication in Cisco Catalyst SD-WAN Controller (formerly SD-WAN vSmart) and Cisco Catalyst SD-WAN Manager (formerly SD-WAN vManage), that was disclosed on February 25; Cisco has released updates to address this vulnerability. Both CVEs have been added to the US Cybersecurity and Infrastructure Security Agency Known Exploited Vulnerabilities (CISA KEV) catalog with a mitigation deadline of Friday, February 25 for Federal Civilian Executive Branch (FCEB) agencies. CISA has published an Emergency Directive (ED 26-03: Mitigate Vulnerabilities in Cisco SD-WAN Systems) that requires FCEBs to identify all in-scope Cisco SD-WAN systems within their networks and provide that information to CISA by 11:59 PM ET on Thursday, February 26, 2026. The directive also includes a list of required actions to take within the same time frame, and requires agencies to apply relevant Cisco updates by 5:00 PM ET on February 27, 2026. The directive lists additional mitigation and reporting requirements.

Anthropic has patched three vulnerabilities in its agentic coding tool, Claude Code, which were discovered and disclosed by researchers from Check Point in 2025. Two of the flaws stem from the way the platform manages project configuration files, including possible abuse of a "Hooks" feature that conditionally triggers user-defined commands without user confirmation, and a means of circumventing user warnings when executing commands contained in Model Context Protocol (MCP) configurations. These vulnerabilities are both tracked under CVE-2025-59536, CVSS score 8.7, which allows an attacker to trick Claude Code into executing code before a user accepts the startup trust dialog, when the tool is started in an untrusted directory. This CVE affects Claude Code before 1.0.111. The other flaw, CVE-2026-21852, carries CVSS score 5.3 and affects Claude Code before 2.0.65: an attacker could steal a user's plaintext API keys by manipulating a configuration file to direct Claude Code's API communication to a malicious endpoint, also bypassing a trust prompt. Check Point notes that these flaws introduce risk of supply chain attacks due to likely inherent trust in project configuration files. Anthropic added more explicit warnings to the user permission dialog, and ensured that neither MCP servers nor API requests can initiate before users confirm the trust dialog: "Claude Code now defers all network operations until after explicit user consent." Users should update to the most current version of Claude Code, inspect configuration directories, heed tool warnings, audit configuration changes, and exercise caution with unusual project setup requirements.

CrowdStrike has published their 2026 Global Threat Report, analyzing trends and statistics observed in threat actor behavior in 2025. Among their findings is the steady acceleration of attacker "breakout time" — meaning time from initial access to lateral movement, a critical window for defenders — speeding up by an average of just over 17 minutes every year, from 98 minutes in 2021 to 29 minutes in 2025, with the fastest breakout observed at just 27 seconds; CrowdStrike contends that "speed is now the defining characteristic of intrusion." The research also highlights six "key adversary themes" in 2025, all of which CrowdStrike expects to continue: 1. *AI systems* were increasingly targeted by attackers and also integrated into attack techniques. 2. *Ransomware attacks* focused on "cross-domain" intrusions to avoid EDR tools, instead pursuing multiple attack surfaces including unmanaged assets such as VPNs, firewalls, personal accounts and devices, unauthorized applications, and third-party systems. 3. *Network perimeter devices* were targeted in 40% of attacks by PRC-based threat actors, often exploiting vulnerabilities soon after disclosure, with the aim of establishing covert persistence in strategically important systems. 4. *Supply chain attacks* increased, targeting both software providers' update mechanisms and developer credentials for public repositories; the report predicts supply chain attacks focusing on SaaS and cloud systems in 2026. 5. *Zero-day exploitation* increased 42% year-over-year, and CrowdStrike urges an in-depth defensive strategy in response, including "penetration testing, establishing a robust vulnerability management program, conducting red teaming exercises, and executing tabletop drills." 6. *Cloud intrusions* rose 37% year-over-year, with a 266% increase specifically by "named state-nexus threat actors," and CrowdStrike predicts hybrid identity solutions will be targeted in 2026. The report concludes with recommendations to secure and govern AI, to treat identity and SaaS as primary attack surfaces, to eliminate cross-domain blind spots, to harden the software supply chain, to prioritize patching and monitoring edge devices, to hunt proactively for threats and intelligence, and to train for secure human decision making.

Taiwanese network technology company Zyxel has published a security advisory addressing seven security flaws in their products. The most severe is CVE-2025-13942, CVSS score 9.8, which allows a remote attacker to execute operating system commands by performing command injection using specially crafted UPnP SOAP requests, if both WAN access and the vulnerable UPnP function have been enabled on the firmware of certain 4G LTE/5G NR CPE, DSL/Ethernet CPE, Fiber ONTs, and Wireless Extenders. WAN is disabled by default on the affected devices. The advisory also covers two high-severity post-authentication command injection vulnerabilities enabling unauthenticated OS commands, and four medium-severity null pointer dereference vulnerabilities enabling a denial-of-service (DoS) condition. Zyxel urges users to consult the table of affected devices to select the appropriate patch, and then contact their Zyxel sales representative or support team to obtain the file. Users who received a Zyxel device from their ISP should reach out to the ISP directly. ShadowServer's data collection shows approximately 120,000 internet-exposed Zyxel devices worldwide.

SolarWinds has released updates to address four critical vulnerabilities in the company's Serv-U file transfer software. An attacker could exploit the flaws to execute code with root privileges; however, all four flaws require administrative privileges to exploit. CVE-2025-40538 is a broken access control vulnerability, CVE-2025-40539 and CVE-2025-40540 are type confusion vulnerabilities, and CVE-2025-40541 is an insecure direct object reference (IDOR) vulnerability. Users are advised to update to Serv-U version 15.5.4 or later. The vulnerabilities do not appear to have been exploited; however, SolarWinds vulnerabilities and file transfer software have both often been targeted by attackers.

Google Threat Intelligence Group (GTIG), Mandiant, and other partners have disrupted a cyber espionage campaign emanating from China. The campaign targeted dozens of telecommunications companies and government organizations in Asia, Africa, and North and South America. GTIG has identified the threat actor behind the attacks as a group tracked since 2017, believed to be acting in the interests of the People's Republic of China. GTIG writes that "the attacker was using API calls to communicate with SaaS apps as command-and-control infrastructure to disguise their malicious traffic as benign." Disruption of the group's activity included "terminating all Google Cloud Projects controlled by the attacker, effectively severing their persistent access to environments compromised by the novel GRIDTIDE backdoor; identifying and disabling all known UNC2814 infrastructure; disabling attacker accounts and revoked access to the Google Sheets API calls leveraged by the actor for command-and-control (C2) purposes; and releasing a set of IOCs linked to UNC2814 infrastructure active since at least 2023."

Peter Williams, former executive at a US defense contractor, has been sentenced to 87 months in prison for selling trade secrets to a Russian broker. Williams, who is Australian, was employed by US defense contractor L3Harris as general manager of its cyber branch, Trenchant. He admitted to stealing eight exploits over a three year period, and pleaded guilty to the charges in October 2025. Williams was ordered to forfeit USD 1.3 million in cash along with cryptocurrency, properties, and other items purchased with the funds obtained from the sale of the information. "According to court documents, the trade secrets were comprised of national-security focused software that included at least eight sensitive and protected cyber-exploit components." In a related story, the US Treasury Department has sanctioned Sergey Sergeyevich Zelenyuk (Zelenyuk) and his company, Matrix LLC (doing business as Operation Zero), as well as five associated individuals and entities for allegedly acquiring and distributing cyber tools that harmed US national security.

A settlement of US $17.25 million has been reached in an unprecedented class-action lawsuit representing over ten million students whose communications were allegedly unlawfully wiretapped and eavesdropped upon during use of school-mandated software. The suit was brought against educational software company PowerSchool, its subsidiary Hobsons Inc., data analytics company Heap Inc., and Chicago Public Schools (CPS), contending that "PowerSchool Holdings and Hobsons allegedly aided and conspired with Heap and other third parties to unlawfully intercept without consent the confidential and sensitive communications of Settlement Class Members while using the Naviance Platform ... [including] sensitive and confidential statutorily-protected education and school student records." In addition to the monetary settlement (of which any uncashed funds will be donated to the Electronic Frontier Foundation), PowerSchool is directed to create a web governance committee for oversight of analytics and advertising technology within Naviance, to publicly disclose such technology within its platforms, to request vendors delete data from the class period, and to prominently announce its privacy principles. CPS is directed to modify its vendor contracts to include transparency and compliance measures surrounding handling of student data, and Heap is directed to delete all primary and backup data of class members. The defendants are released from liability by the settlement.

Massachusetts-based medical device manufacturer UFP technologies has disclosed that a cyberattack in mid-February compromised the company's IT systems and data. The company detected suspicious activity on its network on February 14. UFP then took steps to isolate affected systems and began an investigation with the help of an outside team of cybersecurity specialists. According to the filing, "The incident appears to have impacted many but not all of the Company’s IT systems and affected functions such as billing and label making for customer deliveries. Certain Company or Company-related data appear to have been stolen or destroyed." FTP says it believes the company's primary systems are operational.