Info
2026-07-18 11:54Z · last 24h · 60 findings
· glm-5.2:cloud
Threat Brief — 2026-07-18 — Zero-Days, Supply Chains, and AI-Powered Extortion
Executive Summary
A busy 24 hours: a North Korea-nexus actor compromised the widely used Axios NPM package in a supply-chain attack, while Mandiant disclosed a Cisco Catalyst SD-WAN Manager zero-day actively exploited at a service provider. Splunk AI Toolkit surfaced a CVSS 9.1 RCE, and CISA rushed out multiple KEV additions including Fortinet FortiSandbox command injection and SharePoint exploitation. On the AI front, a fully autonomous ransomware operator ("JADEPUFFER") and AI-assisted vulnerability discovery featured prominently across several reports, signalling that adversaries are industrialising AI faster than many defenders anticipated.
Top Items
- Cisco Catalyst SD-WAN Manager zero-day (CVE-2026-20245KEV) actively exploited. A threat actor leveraged the flaw to target SD-WAN infrastructure at a service provider after gaining initial access. No patch timeline disclosed yet; network teams should review SD-WAN exposure. id 3706
- North Korea-nexus actor compromises Axios NPM package. GTIG tracks an active supply-chain attack injecting malicious code into the widely-deployed Axios library, potentially affecting thousands of downstream applications. Consuming teams should verify package integrity and audit dependency chains immediately. id 3718
- Splunk AI Toolkit RCE — CVSS 9.1. A high-severity vulnerability allows remote execution of arbitrary system commands. Splunk deployments with the AI Toolkit enabled should be treated as potentially exposed and patched as a priority. id 3739
- CISA adds Fortinet FortiSandbox (CVE-2026-25089KEV) and SAP (CVE-2026-3980) to KEV catalog. FortiSandbox OS command injection and an SAP vulnerability are now confirmed under active exploitation. Two additional CVEs were also added in the same batch. id 3682
- SharePoint exploitation campaign — CISA urges hardening. CISA updated its alert (now including CVE-2026-58644KEV in KEV) after continued active exploitation of multiple SharePoint vulnerabilities. Organisations running SharePoint should apply patches and harden configurations without delay. id 3692
- ShinyHunters (UNC6240) targets Oracle PeopleSoft in education sector. Active compromise and extortion campaign exploiting PeopleSoft application infrastructure, focused on educational institutions. Organisations in higher-ed running PeopleSoft should audit for indicators of compromise. id 3708
- Russia FSB Centre 16 exploiting poorly configured devices across critical sectors. Joint CISA advisory urges improved router hygiene; FSB actors continue targeting vulnerable network devices for persistence and lateral movement. id 3695
- China-nexus UNC6508 targets US medical & defence research. A PRC-linked actor is pursuing AI, cyber, medical, and national defence research across public and private medical institutions. Research-heavy orgs should evaluate exposure. id 3707
- M-Trends 2026 released. Mandiant's annual frontline report highlights a mean time-to-exploit of −7 days (bugs exploited before patches ship) and a widening divergence in adversary pacing. Recommended reading for prioritising defensive uplift. id 3719
- KnowledgeDeliver ViewState deserialization exploited in the wild. A late-2025 incident revealed active exploitation of a ViewState deserialization vulnerability in the Japanese LMS product. Legacy web applications with ViewState enabled should be audited. id 3710
- Apple supply-chain data dump — 630 GB on dark web. A supplier with a 20-year relationship reportedly leaked 630 GB of confidential data, highlighting persistent third-party risk. id 3726
- JADEPUFFER: first fully AI-driven ransomware campaign. Researchers demonstrated an autonomous ransomware operator conducting the entire kill chain without human intervention — a proof of concept that threat actors may not be far from replicating. id 3732
Themes
- AI weaponisation accelerating. Multiple reports (JADEPUFFER, T3MP3ST, Ghostcommit, GTIG AI Threat Tracker) show adversaries using AI for vulnerability discovery, autonomous ransomware, and prompt-injection attacks on coding agents. Defenders should assume AI-assisted exploitation will compress time-to-compromise further.
- Supply-chain pressure points. The Axios NPM compromise and the Apple supplier leak reinforce that third-party and dependency-chain risk remains the soft underbelly — both for software and for data.
- Government-nexus targeting of infrastructure. Russia (FSB router operations), China (medical/defence research espionage), and North Korea (NPM supply chain) all demonstrate state-level actors continuing to exploit mundane weaknesses — misconfigured devices, unpatched enterprise software, and unvetted open-source packages.
