Malware delivery increasingly leveraged malvertising and social engineering campaigns, with the InstallFix operation distributing payloads through fake Claude AI installer pages promoted via Google Ads. Victims executed malicious PowerShell commands that triggered mshta-based loaders, AMSI bypass, and staged payload retrieval from attacker infrastructure. A parallel phishing campaign impersonated the US Social Security Administration deployed signed remote monitoring tools to establish persistence and continuous surveillance across compromised systems.

Multiple remote code execution vulnerabilities saw active exploitation across widely deployed platforms, including Weaver E-cology CVE-2026-22679 and MetInfo CMS CVE-2026-29014. Both flaws enable unauthenticated attackers to execute arbitrary commands via crafted requests, with observed exploitation involving PowerShell loaders and automated scanning of exposed systems. Apache HTTP Server CVE-2026-23918 introduced a double-free flaw in HTTP/2 handling that can trigger denial-of-service conditions and potential remote code execution in default configurations.

Supply chain attacks continued to affect both software distribution and development ecosystems, with DAEMON Tools installers trojanised to deliver persistent malware through signed binaries distributed via official channels. A separate compromise of a Yanbian gaming platform enabled ScarCruft to distribute BirdCall and RokRAT backdoors across Android and Windows environments, targeting users for espionage data collection. Additional activity involving compromised CI pipelines demonstrated credential exfiltration from Jenkins and AWS environments through malicious GitHub pull request workflows.

State-linked threat activity remained active, with UAT-8302 conducting intrusions against government entities across South America and southeastern Europe using multiple malware families including NetDraft and CloudSorcerer. Post-compromise actions involved credential harvesting, lateral movement, and network reconnaissance using PowerShell scripts, Impacket tooling, and scheduled tasks. Overlapping toolsets with other China-aligned clusters included SNOWLIGHT and SNOWRUST loaders, indicating shared development or operational resources.

Mobile and cross-device threats expanded with new malware capabilities and vulnerabilities, including CloudZ RAT abusing Microsoft Phone Link to intercept SMS and one-time passwords without infecting mobile devices directly. WhatsApp disclosed vulnerabilities affecting Windows, Android, and iOS that enable file spoofing and arbitrary URL handling through crafted messages and attachments. Separately, the DarkSword exploit chain leveraged multiple iOS zero-day vulnerabilities to deploy surveillance malware across targeted regions using full-chain compromise techniques.

Germany’s federal cabinet has advanced a legislative package allowing law enforcement automated biometric image matching of publicly available internet data using AI systems, enabling police to upload facial images and search the internet for matching depictions, replacing manual searches of online sources. Civil society organisations and some lawmakers oppose the measures over mass surveillance and privacy risks, while noyb has filed a lawsuit against the Hamburg data protection authority over alleged failure to enforce EU rules against the PimEyes facial recognition search engine despite an earlier illegality ruling.

CVE-2026-22679 affects Weaver E-cology 10.0 and enables unauthenticated remote code execution through the /papi/esearch/data/devops/dubboApi/debug/method endpoint using crafted POST parameters interfaceName and methodName fields. Attackers exploited the flaw for remote command execution, deploying PowerShell loaders, MSI payload fanwei0324.msi, and running discovery commands including whoami, ipconfig, tasklist. Vega Research Team identified exploitation from 17 March 2026, earlier than Shadowserver reporting on 31 March, with activity spanning post-patch targeting.

WhatsApp has disclosed CVE-2026-23863 affecting Windows and CVE-2026-23866 affecting iOS and Android, introducing attachment spoofing and arbitrary URL processing issues in 2026 advisories. Windows flaw allows malicious attachments with embedded NUL bytes in filenames to appear as benign files but execute when opened in WhatsApp application. iOS and Android issue involves incomplete validation of AI rich response messages for Instagram Reels, enabling arbitrary URL processing and OS URL scheme triggering.

ESET Research identified a ScarCruft supply-chain attack compromising a Yanbian gaming platform, trojanising Windows and Android games with RokRAT and BirdCall backdoors. The Android BirdCall variant in trojanised APKs collects contacts, SMS, call logs, files, screenshots and audio, exfiltrating data via cloud storage C2 channels. Windows infection chain used malicious updates and mono.dll trojanisation delivering RokRAT and BirdCall, targeting Yanbian users for espionage since late 2024 campaign.

Cisco Talos disclosed UAT-8302, a China-nexus APT targeting government entities in South America and southeastern Europe, deploying NetDraft, CloudSorcerer v3 and VSHELL. Post-compromise activity included information collection, credential extraction and network proliferation using Impacket, PowerShell scripts, scheduled tasks and reconnaissance commands across infected endpoints. Tooling overlap with other China-nexus clusters included Draculoader, SNOWLIGHT, SNOWRUST and ZingDoor, alongside proxying tools, AD enumeration, and credential harvesting capabilities.

MetInfo CMS versions 7.9 through 8.1 are affected by CVE-2026-29014, an unauthenticated PHP code injection vulnerability in the Weixin reply component that enables remote code execution. The flaw resides in weixinreply.class.php where insufficient input sanitisation of WeChat API parameters allows path traversal and cache manipulation leading to arbitrary PHP execution on affected servers. Exploitation activity was observed following patch release in April 2026, with automated probing and increased targeting of exposed instances in China and Hong Kong.

Antrea CNCF project was compromised through a GitHub pwn request campaign linked to TeamPCP Trivy supply-chain attack, extending access into Jenkins and AWS environments. Attackers used Jenkins slash-command triggered jobs from malicious pull request code to execute payloads that exfiltrated AWS credentials, SSH keys, and metadata via external endpoints. Activity involved throwaway GitHub accounts, iterative multi-project campaigns, spoofed CI identities, and exploitation of CI infrastructure including controller access and EC2 host 35.164.122.165 exposure.

Orca Security research found attack primitives in AI agent skills marketplace enabling malicious skill distribution and persistence via install count inflation, weak scanning, silent overrides. Marketplace issues include unauthenticated telemetry install counts, non-continuous security scans, and skill name collisions allowing replacement of trusted skills and delayed malicious modifications. Demonstrations showed bait-and-switch, nested injection, and delayed update weaponisation achieving code execution on end-user systems via markdown-based skills executed by coding agents.

An ongoing STAC6405 phishing campaign impersonates the US Social Security Administration and has targeted over 80 organisations since April 2025, deploying vendor-signed SimpleHelp 5.0.1 and ScreenConnect via compromised .com.mx infrastructure. The malware installs a Windows service with Safe Mode persistence and dual RMM channels, running automated surveillance loops including WMI queries, network checks and mouse-polling, and communicating with 84.200.205.233 and 213.136.71.246 plus a ScreenConnect relay domain.

Apache HTTP Server 2.4.66 contains a critical HTTP/2 vulnerability tracked as CVE-2026-23918, a double-free flaw in mod_http2 that can trigger denial-of-service and potential remote code execution. The issue occurs when specific HEADERS and RST_STREAM frames are sent in sequence, causing duplicate cleanup of the same stream pointer and memory corruption during stream destruction. Researchers demonstrated a working RCE proof of concept using mmap allocator behaviour and Apache scoreboard memory, with exploitation affecting default deployments using multi-threaded MPM configurations and fixed in version 2.4.67.

The InstallFix campaign distributes malware through fake Claude AI installer pages promoted via Google Ads, targeting users who search for installation instructions and execute malicious PowerShell commands. The infection chain uses mshta.exe to run a ZIP/HTA polyglot payload, deploys obfuscated scripts, disables SSL validation and AMSI protections, and retrieves additional payloads from attacker-controlled infrastructure. Telemetry shows scheduled task creation for persistence and network communication with command-and-control servers, affecting organisations across multiple regions and sectors including government, education, and electronics.