# Today's Top Stories

November 18, 2026

  1. 1
    0
    The Hacker News general Nov 17
    Hijacked Optimizely Scripts Served Card Skimmer and iOS Exploit Kits to CNN, BBC and Zoom Visitors

    Attackers exploited a vulnerability in Optimizely Web Experimentation to inject malicious JavaScript into scripts loaded by at least 178 organizations' websites, including CNN, BBC, Zoom, Salesforce, GitLab, and Equifax. The malicious code delivered a multi-payload cocktail including a card skimmer, cryptocurrency theft tools, and iOS exploit kits to unsuspecting visitors. This is a significant supply chain attack affecting major global brands and underscores the risk of third-party web experimentation platforms as an attack vector.

  2. 2
    0
    BleepingComputer general Nov 17
    Critical Java4Shell flaw allows RCE attacks on major Java applications

    A critical vulnerability dubbed Java4Shell (CVE-2026-114938) in OpenJDK enables unauthenticated remote code execution on major enterprise applications including Jenkins, Atlassian Jira, Confluence, and Keycloak. The flaw affects Java 17.0.6 and later running on x86-64 processors supporting AVX-512 VBMI instructions, limiting but not eliminating its attack surface. Oracle shipped patches this week; security teams running affected JVM versions on modern Intel/AMD hardware should prioritize patching immediately.

  3. 3
    0
    BleepingComputer general Nov 17
    Uber notifies drivers of data breach affecting nearly 150,000 people

    Uber is notifying nearly 150,000 drivers of a data breach stemming from the compromise of AI infrastructure provider Fireworks AI in early October, where attackers stole internal credentials and used them to access Uber's account. The breach originated from a stolen Fireworks access key, which was also linked to the separate Cursor supply chain attack involving malicious Claude Code extensions. This incident illustrates cascading third-party breach risk when AI infrastructure providers are compromised.

  4. 4
    0
    CyberScoop general Nov 16
    Google says Russian-speaking hackers ran the first fully automated AI cyberattack

    Google's Threat Intelligence Group has documented the first fully automated AI cyberattack, attributed to Russian-speaking threat actors using autonomous multi-agent offensive frameworks including Artex, CAIRN, CyberStrikeAI, and PentAGI to execute multi-stage intrusions at increased velocity. While the tools leave noisier forensic traces than manual attacks, they dramatically increase operational tempo. This marks a qualitative shift in the threat landscape that defenders must account for in detection and response planning.

  5. 5
    0
    The Hacker News general Nov 17
    N-able Patches 17 N-central Flaws as Attackers Exploit Earlier Bugs to Deploy Crypto Miners

    N-able released N-central 2026.4 Hotfix 3 patching 17 zero-day vulnerabilities in its Remote Monitoring and Management platform, with some flaws already actively exploited in the wild to deploy cryptocurrency miners on managed endpoints. The update also kills legacy APIs that were being abused. MSPs and enterprises using N-central should treat this as an emergency patch given active exploitation and the privileged access RMM tools have across managed environments.

  6. 6
    0
    The Hacker News general Nov 17
    ConnectWise Warns ScreenConnect Attacks Continue as SCKit Targets Unpatched Systems

    ConnectWise has completed its investigation into unauthorized access targeting partner ScreenConnect environments and is urging all customers to upgrade the technician client to version 26.6.5 or later. A month-long campaign exploiting CVE-2026-84869 sent over 688,000 requests targeting more than 40 servers to execute guest-pushed files. Active exploitation of unpatched on-premises ScreenConnect instances continues, particularly where vulnerable third-party extensions remain enabled.

  7. 7
    0
    SecurityWeek general Nov 16
    Google Project Zero Sets Course for 30-Day Disclosure Deadline

    Google Project Zero announced it will shorten its vulnerability disclosure deadline from the current 120-day window to 60 days in February 2027 and then to 30 days in January 2028. The move is designed to pressure vendors to patch faster but has divided the security community, with critics arguing shorter windows favor attackers when patches are delayed. This policy change will directly affect how vendors prioritize and resource their patch development cycles.

  8. 8
    0
    The Hacker News general Nov 17
    New GitDoor and GraphDoor Backdoors Turn GitHub and SharePoint Into Covert C2 Channels

    Kaspersky researchers disclosed Operation ShadowCloud, a cyber espionage campaign targeting government sectors in Asia and Africa using two novel backdoors — GitDoor and GraphDoor — that abuse GitHub and Microsoft SharePoint respectively as covert command-and-control channels. Identified in early 2026, the campaign leverages legitimate cloud services to blend malicious traffic with normal enterprise activity, complicating detection via traditional network monitoring. Security teams defending government networks should audit outbound traffic to these platforms for anomalous patterns.

  9. 9
    0
    SecurityWeek general Nov 16
    Barracuda Hacked Amid Surge in Artex-Powered Cyberattacks

    Barracuda confirmed it was breached amid a surge in attacks powered by Artex, an agentic AI penetration testing framework being broadly adopted by Chinese threat actors to accelerate intrusion campaigns. The same Artex framework was used in attacks against advanced manufacturing, defense, and civil rights organizations, with at least 10 organizations confirmed compromised. The incident highlights how AI-driven attack frameworks are enabling nation-state actors to scale and speed up targeted intrusions against high-value sectors.

  10. 10
    0
    BleepingComputer general Nov 17
    Researchers hijack Hugging Face account for popular Kokoro AI model

    Researchers exploited a lack of ownership verification on the Hugging Face platform to hijack the account hosting Kokoro, a widely used open-source text-to-speech AI model integrated into projects from Google, NVIDIA, AMD, Intel, and AWS. The takeover demonstrates a systemic trust gap in the Hugging Face model repository that could enable malicious model substitution at scale. Organizations consuming AI models from Hugging Face should implement cryptographic verification and monitor for unexpected model updates.