Stories first seen on 18 September 2026
Review the tracked stories and available summary evidence for the archived period shown.
Population: stories the tracker first saw on this Coordinated Universal Time (UTC) calendar day; publisher date is used only if first-tracked time is unavailable and remains on each item.
123 stories
- AI Security
Anthropic's AI Plays Major Role in Building Next Models
Anthropic reported that Claude participates in 26% of model research and development work as a lead contributor, with involvement across most remaining tasks. The company disclosed these findings from an internal study examining how frontier artificial intelligence (AI) labs employ their own models to construct successor systems.
Why it matters: Security and AI practitioners need visibility into how AI systems are used in AI development pipelines, given the implications for supply chain integrity, model safety evaluation, and insider risk when AI agents have access to proprietary model architectures.
- OT, ICS, and Critical Infrastructure
Cyber Defense Alone Can't Keep Critical Services Running
State chief information officers must prioritize critical infrastructure protection by mapping dependencies and ranking water systems and hospitals by consequence of failure. NASCIO data indicates that cyber defense alone is insufficient; states should test simultaneous failure scenarios and combine cybersecurity measures with engineering safeguards.
Why it matters: State CIOs and infrastructure operators are responsible for ensuring water and hospital systems remain operational; understanding cross-infrastructure dependencies enables prioritized protection that accounts for cascading failure risks.
- Government and Policy
DARPA Seeks AI Tools to Transform Battlefield Medical Care
The Defense Advanced Research Projects Agency (DARPA) is launching two competitions focused on artificial intelligence (AI), robotics, and automation to improve trauma care in combat settings. The initiatives target surgical robotics, clinical decision support, and medical documentation to address gaps when field surgeons are unavailable or critical treatment decisions go unrecorded.
Why it matters: Military medical personnel and defense contractors should monitor these competitions for funding and partnership opportunities to develop AI-enhanced trauma care systems applicable to forward-deployed environments.
- Industry and Vendors
Post-Mythos Security Rally Rewards Broad Cyber Platforms
Six cybersecurity vendors have seen their valuations double since Anthropic introduced Claude Mythos in April, with Okta leading the group at a 131% stock increase. The market gains reflect investor confidence that artificial intelligence (AI) agents will accelerate adoption of identity, data, network, and integrated security solutions.
Why it matters: Security leaders and practitioners should monitor how AI adoption reshapes vendor ecosystems and platform consolidation, as valuation momentum can signal both market demand and potential M&A activity that affects tools and contracts.
- Ransomware and Leak Sites
Ransomware attack on Kansas county will affect some services
Ellis County, Kansas discovered a ransomware attack affecting portions of its information technology systems. The county isolated the affected systems and engaged cybersecurity experts to contain the incident.
Why it matters: County residents and businesses depending on government services face potential disruptions; practitioners should monitor for service outages and credential exposure from compromised systems.
- Industry and Vendors
Vectra AI Launches Ascent to Help Address New Era of AI-Driven Attacks
Vectra artificial intelligence (AI) launched Ascent, a program that expands its partner strategy to address mounting security challenges driven by artificial intelligence (AI) and increasingly complex threat environments. The initiative aims to broaden access to AI expertise, services, and security outcomes across the partner ecosystem.
Why it matters: Security teams and managed service providers need to evaluate whether Vectra AI's expanded partner program can help them respond to AI-driven attacks and reduce detection and response times in their environments.
- Threat Intelligence
Early Scattered Spider member pleads guilty to cybercrime spree
Ahmed Hossam Eldin Elbadawy, a 24-year-old Texas resident and early member of the Scattered Spider cybercriminal group, pleaded guilty one year ago to wire fraud conspiracy and aggravated identity theft for his role in extortion attacks spanning 2021 to 2023. Operating alongside co-conspirators, Elbadawy and his crew used social engineering to compromise credentials, identify high-net-worth employees, and steal virtual currency from at least 29 victims across entertainment, telecom, technology, and cloud sectors, netting thefts including $6.35 million, $571,000, and $1.7 million in separate incidents. Prosecutors are seeking forfeiture of over $17.6 million in cryptocurrency, luxury vehicles, and designer goods connected to his criminal proceeds.
Why it matters: Security teams at large enterprises in entertainment, telecom, technology, and cloud services should review social engineering defenses and virtual currency wallet access controls, as Scattered Spider continues recruiting replacements despite arrests of early leadership.
- Breaches and Incidents
Foreign actors breach Colorado water systems
Foreign actors breached two small Colorado water utilities serving fewer than 200 people each in late August, gaining access to equipment controlling drinking water systems. The attackers modified equipment settings, disabled remote access, and disabled alarms at the facilities. The incidents were disclosed by Governor Jared Polis' office.
Why it matters: Water utility operators and state regulators need immediate visibility into compromise techniques targeting supervisory control and data acquisition systems, as these small systems often lack advanced threat detection and may indicate broader targeting of critical water infrastructure.
- Breaches and Incidents
The U.S. military leaked more than 93,000 tips via insecure P3 Global Intel. Has anyone been notified?
The U.S. military exposed approximately 94,000 unclassified but sensitive tips through P3 Global Intel applications and websites operated by Navigate360, according to DataBreaches.net's investigation. The incident has been compared to BlueLeaks due to the scale of the disclosure. The status of notification efforts and Homeland Security's response remain unclear.
Why it matters: Military personnel and the public who submitted tips to P3 Global Intel face exposure of sensitive information; defense organizations must assess whether notification protocols were followed and determine the scope of potential misuse.
- Vulnerabilities and Patches
Click2Shell: The RCE WordPress 7.1.1 Just Patched
WordPress 7.1.1 patched Click2Shell, a remote code execution (RCE) vulnerability that chains cross-site request forgery (CSRF) with selector injection to achieve arbitrary code execution. The flaw exploited a mismatch in how server-side and client-side code handled theme slugs, allowing an attacker to silently install a malicious theme and trigger its code through the WordPress Customizer. The patch sanitizes the slug input on the frontend using jQuery escapeSelector and restricts the selector scope to legitimate theme elements.
Why it matters: WordPress administrators must update immediately; attackers can trigger RCE by tricking an admin into clicking a malicious link while logged in or by exploiting existing XSS vulnerabilities to auto-execute the attack in the admin's browser.
- Vulnerabilities and PatchesCVE-2026-76460
Cisco Zero-Day Highlights API Endpoint Authentication Issues
Cisco disclosed an authentication bypass vulnerability, CVE-2026-76460, affecting Identity Services Engine (ISE) that received a critical CVSS 10.0 score. The flaw enables unauthorized access via an unauthenticated application programming interface (API) endpoint.
Why it matters: Organizations running Cisco ISE must patch immediately, as this vulnerability is actively exploited and allows unauthenticated attackers to compromise identity and access control systems.
- AI Security
EY Survey Finds Autonomous AI Implementation Outpaces Oversight
A survey of senior artificial intelligence (AI) executives reveals that organizations are deploying AI and autonomous systems faster than they are establishing oversight processes and controls. The gap between implementation velocity and governance maturity creates potential risks in production environments.
Why it matters: Organizations implementing autonomous AI without proportional controls face unmanaged security, compliance, and operational risks; practitioners should assess whether their AI governance framework matches deployment speed.
- Threat Intelligence
Happy Birthday, Shai-Hulud
One year after a self-propagating worm called Shai-Hulud compromised the npm package @ctrl/tinycolor in September 2025, the supply chain threat evolved into a widespread campaign that spawned multiple variants, copied code, and copycat attacks. The worm harvested credentials and GitHub tokens to inject itself into hundreds of packages, with subsequent waves in November 2025 and throughout 2026 expanding the attack surface and payload capabilities. Australian authorities arrested two alleged members of the cybercriminal group TeamPCP in August 2026 for their role in later waves, though the original authors remain unattributed and the now-public code continues to be adopted by other threat actors.
Why it matters: Developers and DevOps teams using npm packages, especially open source maintainers, face ongoing risk from supply chain compromise as the Shai-Hulud worm code remains public and widely forked; organizations need to rotate credentials, audit lockfiles, and monitor for additional waves.
- Ransomware and Leak Sites
Ten of Fifteen: The Ransomware Leaderboard Turned Over Again in August
Ten of the fifteen most active ransomware families in August differed from July's rankings, reflecting rapid turnover among threat actors. The underlying tactics and techniques employed by these groups remain consistent despite the changing roster of prominent names.
Why it matters: Security teams monitoring ransomware threats need to track both emerging group names and persistent attack methods; focusing solely on known families creates blind spots for newly active threats using established tradecraft.
- Vulnerabilities and Patches
Wordfence Argus Discovers Critical Vulnerability in libheif, the Library That Opens iPhone Photos on Your Server
Wordfence Argus discovered a critical heap buffer overflow in libheif versions 1.18.0 through 1.23.2, a widely used library that decodes HEIC and HEIF images on servers. The vulnerability allows a crafted image to write attacker-controlled data beyond a memory buffer, potentially enabling arbitrary file read or remote code execution, and was fixed in version 1.23.3 released on September 1, 2026. WordPress sites are exposed when they process untrusted HEIC uploads through the PHP Imagick extension and ImageMagick integration with an affected libheif build, though vulnerability depends on specific build configuration rather than version number alone.
Why it matters: Server operators and hosting providers must urgently patch libheif to prevent remote code execution through untrusted image uploads on WordPress sites and other applications (photo services, media servers, document pipelines) that process HEIC or HEIF files; the critical CVSS 9.8 vulnerability can be tailored to specific environments quickly, making high-value targets and standardized hosting fleets particularly at risk.
- Identity and Access
MFA Won't Save You From OAuth Consent Abuse
Multifactor authentication (MFA) alone cannot defend against OAuth consent abuse, which exploits the authorization framework to grant applications excessive permissions. Organizations need to combine MFA with OAuth governance, least-privilege scope assignment, consent monitoring, and rapid revocation capabilities.
Why it matters: Development and security teams relying solely on MFA for OAuth security face risk from compromised or malicious applications gaining broad access to user data and systems; implementing OAuth governance and scope controls provides essential defense today.
- Vulnerabilities and Patches
Public Exploits Released for Four Linux Kernel Flaws That Enable Local Root
A security researcher has published working exploit code for four Linux kernel vulnerabilities that permit local users to escalate privileges to root access. Kernel maintainers have patched all four flaws in recent weeks, but systems running unpatched kernels remain exposed now that exploits are public.
Why it matters: Linux administrators managing systems on older kernel versions face immediate risk of local privilege escalation; prioritize kernel updates to patch these four flaws.
- Vulnerabilities and Patches
Researchers use AI to find widespread software decoder flaw
Researchers used large language models from OpenAI and Anthropic to discover a widespread flaw called HEIF Heist in popular software decoders libheif and libde265 that could allow remote code execution and data theft across major platforms. The vulnerability exploits memory corruption errors triggered by specially crafted image files, potentially compromising systems at OpenAI, Meta, GitHub, Amazon Web Services, and other major services. Patches are available but deployments lacking the latest upstream updates remain at risk.
Why it matters: Organizations running unpatched versions of libheif or libde265 in image processing or decoding services face direct exposure to remote code execution and credential theft; security teams should immediately verify patch status and prioritize updates for any systems parsing HEIF, HEIC, or AVIF image files.
- Threat Intelligence
Don’t Call Us, We’ll Call Your APIs | TraderTraitor Backdoors Resurface on Victim With No Crypto Ties
SentinelOne identified a second victim infected with the macOS backdoors FLATROOF and ROOFDECK used in the April 2026 TraderTraitor attack on LayerZero. Unlike the cryptocurrency-focused initial victim, this target was an IT services organization in India whose DevOps engineer was compromised through a fake job interview lure containing a weaponized Terraform lock file pointing to attacker-controlled provider registries. The backdoors remained dormant for days after infection, then became active in March, beaconing to command and control servers until June 1, 2026, while the attackers deployed additional stages and ultimately removed the original implants.
Why it matters: DevOps and infrastructure engineers at any organization are now active targets for North Korean state-sponsored actors using fake hiring interviews and weaponized Terraform projects; practitioners must monitor development workstations with cloud credentials and source control access, validate Terraform provider registries against official domains, and educate developers on the supply-chain risks of executing unfamiliar coding projects.
- Vulnerabilities and Patches
New WordPress Click2Shell Flaw Forces Theme Installs, Can Chain to Code Execution
WordPress released patches for vulnerabilities in its core software, including a flaw that allows a crafted link opened by a logged-in administrator to automatically install themes from the official WordPress.org directory without explicit user action. Security researchers at pwn.ai reported the attack chain, termed Click2Shell, which can potentially chain to remote code execution.
Why it matters: WordPress administrators are at risk of unwanted theme installation that may lead to code execution if they click a malicious link while logged in; patching is urgent for all WordPress site owners and hosting providers.
- Vulnerabilities and PatchesCVE-2026-91867
CVE-2026-91867: Apache Neethi: Remote policy fetch lacks a total timeout, allowing a slow server to hang the request indefinitely
Apache Neethi before version 3.2.4 contains a denial of service vulnerability in remote policy fetch handling. The affected versions enforce only per-read timeouts, not a total timeout for the entire transfer, allowing a slow server to keep the connection open indefinitely and exhaust the calling thread. Version 3.2.4 addresses this issue.
Why it matters: Teams deploying Apache Neethi before 3.2.4 should upgrade immediately to prevent denial of service attacks via slow remote policy servers that can hang application threads.
- Vulnerabilities and PatchesCVE-2026-91866
CVE-2026-91866: Apache Neethi: Crafted policies cause unbounded work during intersection leading to denial of service
CVE-2026-91866 affects Apache Neethi versions before 3.2.4, where specially crafted WS-Policy documents trigger exponential computational work during policy intersection, causing denial of service. The vulnerability has a CVSS score of 7.5. Upgrade to version 3.2.4 to remediate.
Why it matters: Organizations running Apache Neethi in production for WS-Policy processing should prioritize upgrading to 3.2.4 to prevent attackers from exhausting CPU resources via crafted policy documents.
- Threat Intelligence
FBI: Fake cop and government impersonation scams cost victims $1.6B
Law enforcement and government impersonation scams generated losses exceeding $1.6 billion between January 2025 and July 2026, with nearly 61,000 complaints filed to the FBI's Internet Crime Complaint Center. Scammers employ various tactics ranging from generic threats of arrest or unpaid jury duty to targeted approaches based on victims' professions or immigrant status, with some perpetrators using video calls and costume uniforms to increase credibility. The most profitable variant targets international students and foreign nationals with threats of passport cancellation or extradition, accounting for nearly 10 percent of total losses despite representing less than 3 percent of complaints.
Why it matters: Organizations and individuals, especially professionals in regulated fields and international students in the US, face targeted financial extortion using law enforcement impersonation; practitioners should ensure staff understand these tactics and know that legitimate law enforcement will not demand payment by phone.
- Vulnerabilities and PatchesCVE-2026-91864
CVE-2026-91864: Apache Neethi: Crafted WS-Policy documents bypass element/attribute limits causing memory exhaustion
Apache Neethi versions before 3.2.4 are vulnerable to denial of service when processing specially crafted WS-Policy documents that bypass element and attribute size limits. Affected applications copy unlimited policy assertion content into memory without enforcing heap restrictions, enabling attackers to exhaust available memory. Version 3.2.4 addresses the flaw.
Why it matters: Organizations running Apache Neethi before 3.2.4 in web services environments face denial of service risk from unauthenticated attackers sending malformed WS-Policy documents; upgrade immediately to 3.2.4.
- Vulnerabilities and PatchesCVE-2026-91863
CVE-2026-91863: Apache Neethi: Uncontrolled recursion while parsing crafted WS-Policy documents allows denial of service
Apache Neethi before version 3.2.4 contains a denial of service (DoS) vulnerability in its WS-Policy document parser. Specially crafted documents with deeply nested policy elements can bypass the nesting-depth limit and exhaust thread stack resources, causing the parser to crash. The vendor recommends upgrading to version 3.2.4 to resolve the issue.
Why it matters: Organizations running Apache Neethi in production for web services policy handling must upgrade to 3.2.4 to prevent attackers from triggering DoS conditions through malicious WS-Policy documents.
- Vulnerabilities and PatchesCVE-2026-93019
CVE-2026-93019: Imager versions before 1.036 for Perl exit the process reading a TGA with a colour map length of 32768 or more in tga_palette_read
CVE-2026-93019 affects Imager, a Perl image processing library, in versions before 1.036. The vulnerability causes the process to exit when reading a TGA file with a color map length of 32768 or more in the tga_palette_read function. Users of affected Imager versions should upgrade to 1.036 or later.
Why it matters: Perl developers and applications using Imager to process untrusted TGA image files face a denial of service risk; update to version 1.036 or later.
- Vulnerabilities and PatchesCVE-2026-93018
CVE-2026-93018: Imager versions before 1.036 for Perl disclose uninitialised heap memory reading a paletted image with pixel indexes past its colour map in i_gpix_p and i_glin_p
Imager, a Perl image processing library, contains a vulnerability in versions before 1.036 that discloses uninitialized heap memory when reading paletted images with pixel indexes beyond the color map bounds. The flaw affects the i_gpix_p and i_glin_p functions.
Why it matters: Perl developers using Imager to process untrusted paletted image files face potential information disclosure; upgrade to version 1.036 or later.
- Threat Intelligence
An Undercover Google Analyst Infiltrated a Notorious Supply Chain Hacking Gang
Google's threat intelligence group reports having placed an undercover analyst inside TeamPCP, a hacking group responsible for a major software supply chain attack campaign that compromised thousands of companies. The infiltration provided intelligence on the gang's operations and methods during their spree.
Why it matters: Security teams relying on supply chain defenses need to understand that advanced threat groups may have been actively compromised by intelligence agencies, shifting the threat landscape for future campaigns and validation of third-party risk controls.
Grouped: similar headlines.
- Government and Policy
International security agencies warn about North Korean hackers exploiting job seekers to steal crypto, data
U.S. and allied security agencies have identified WaterPlum, a North Korean hacking group subordinate to the Munitions Industry Department, as targeting job seekers worldwide by posing as recruiters for artificial intelligence, cryptocurrency, and other firms. The group has infected over 30,000 devices across more than 100 countries and stolen approximately $11 million in cryptocurrency from 7,000 crypto wallets. Agencies in Japan, Australia, Germany, and the U.S. have begun dismantling infrastructure, including a laptop farm in Japan, and are coordinating further enforcement efforts.
Why it matters: Software developers, IT professionals, and cryptocurrency users need to scrutinize unsolicited job offers and avoid downloading files from unknown recruiters, as this campaign directly targets technical staff with financial and credential theft as the objective.
Grouped: the same names (NORTH KOREA, NORTH KOREAN, NORTH KOREAN IT).
- OT, ICS, and Critical Infrastructure
FBI, Coast Guard boarded hacked oil tankers heading towards US coast
The Federal Bureau of Investigation and Coast Guard boarded oil tankers that were en route to the US coast after their networks were compromised. In at least one case, the compromise affected the vessel's navigation and propulsion systems, prompting a federal investigation into the incident.
Why it matters: Maritime operators and critical infrastructure owners face immediate risk: compromised vessel systems can disrupt supply chains and pose collision or environmental hazards, and this incident signals adversaries are targeting tanker fleets headed to US ports.
- Threat Intelligence
Transparent Tribe Deploys New Rust Backdoor Using Private GitHub Repositories for C2
Transparent Tribe, a Pakistan-aligned threat group, deployed previously undocumented Rust-based backdoors named RUSTYSHADE and RUSTYMOVE against government and defense targets in India and Afghanistan. The campaign, tracked as Operation, leverages private GitHub repositories for command and control communications.
Why it matters: Government and defense practitioners in India and Afghanistan face direct targeting; organizations should hunt for these new Rust-based tools and monitor GitHub activity for unauthorized C2 communications.
- Threat Intelligence
Fake LastPass Authenticator GitHub repos push new Rapuncel infostealer
Threat actors created fake GitHub repositories that mimic legitimate software vendors to distribute a previously unknown information stealer called Rapuncel. The repositories use search engine optimization techniques to increase visibility and deceive users into downloading malware.
Why it matters: Developers downloading what they believe are official tools from GitHub risk installing infostealer malware that can exfiltrate credentials, tokens, and sensitive data from their systems and development environments.
- Vulnerabilities and Patches
CISA ends weekly vulnerability roundups as part of shift to prioritization approach
The Cybersecurity and Infrastructure Security Agency (CISA) is discontinuing its weekly vulnerability roundup publications and moving toward a prioritization-focused approach. The shift reflects efforts to help organizations navigate the increasing volume of vulnerability disclosures driven by automated discovery tools and artificial intelligence (AI).
Why it matters: Security teams relying on CISA's weekly digests must identify alternative sources for vulnerability tracking, while practitioners should monitor CISA's new prioritization methodology to ensure they focus remediation efforts on the most critical exposures.
- Regulatory and Compliance
Democratic Senators Reintroduce the Health Infrastructure Security and Accountability Act
Two Democratic Senators reintroduced the Health Infrastructure Security and Accountability Act on September 17, 2026, to mandate minimum cybersecurity standards for healthcare entities and provide funding for rural and underserved hospitals. The bill responds to rising healthcare breaches, with year-to-date figures (through August 2026) showing 426 hacking-related breaches affecting 73 million Americans, a 70% increase in affected individuals since the bill's initial introduction in 2024. The legislation establishes mandatory security standards enforced by the Department of Health and Human Services, requires annual audits and risk assessments, increases HIPAA penalties, and allocates 1.3 billion dollars in federal funding for cybersecurity improvements.
Why it matters: Healthcare providers and business associates face new mandatory compliance obligations if the bill passes, including independent audits, annual attestations from leadership, and substantially higher HIPAA penalties for non-compliance, making this a priority for governance and budget planning.
- Threat Intelligence
Nations take action on North Korean IT workers after UN report
Vietnam, Laos, Pakistan, and Argentina have taken action in response to allegations involving North Korean information technology workers documented in a prior UN report. The countries moved to address the issue as of July following an October study highlighting the matter.
Why it matters: Organizations and governments relying on supply chain vetting and labor verification should monitor enforcement actions against North Korean IT worker networks, which pose operational security and sanctions compliance risks.
- Ransomware and Leak Sites
In Other News: Ransomware Developer Sentenced, Plugin4Shell AI Attack, Critical SAP Flaw
A ransomware developer received a sentence, a WordPress plugin vulnerability was exploited in attacks, and a critical SAP flaw emerged. Mandiant released an artificial intelligence (AI) risk report for 2026, while PhantomRaven malware was discovered in use by a bug bounty hunter.
Why it matters: Organizations running WordPress plugins need to patch the disclosed vulnerability immediately; SAP customers should assess their exposure to the critical flaw; teams should review Mandiant's AI risk findings for emerging threat vectors; incident responders should investigate PhantomRaven activity for potential insider threats.
- Vulnerabilities and Patches
[Control systems] Advantech security advisory (AV26-937)
Advantech released security advisory AV26-937 on September 18, 2026, identifying vulnerabilities in the EKI-1242EIMS and EKI-1242IEIMS industrial gateway products at version 2.00.01 and earlier. The Cyber Centre recommends users review the advisory and apply updates as they become available.
Why it matters: Organizations deploying Advantech EKI-1242 series gateways in industrial control environments must prioritize patching to prevent exploitation of these exposed devices in critical infrastructure.
- Cloud and SaaS Security
Secure enterprise sharing with access reviews for Microsoft 365
Microsoft 365 users often grant file access that persists beyond its intended purpose, creating visibility gaps around who retains access to sensitive data. Tenfold Software outlines how centralized access governance and owner-driven reviews help organizations identify and revoke unnecessary permissions.
Why it matters: Microsoft 365 administrators and information governance teams need to audit and clean up stale access rights to reduce the attack surface from overprivileged accounts.
- Cloud and SaaS Security
Microsoft Teams will let admins block custom file extensions
Microsoft Teams is adding a feature that allows administrators to customize file extension blocking policies based on their organization's security needs. This enhancement gives IT teams granular control over which file types are restricted within Teams messaging and file sharing.
Why it matters: Administrators responsible for Teams security need this control to align file restriction policies with company risk tolerance and prevent distribution of potentially dangerous files across the platform.
- AI Security
Researchers used Claude to hack OpenAI
Researchers used Anthropic's Claude tool to gain unauthorized access to an OpenAI employee's ChatGPT account and read private software details. The breach was conducted as part of a paid security testing program designed to identify vulnerabilities before malicious actors could exploit them. The incident underscores security gaps at OpenAI and raises broader concerns about safety practices at leading artificial intelligence (AI) companies.
Why it matters: OpenAI and other AI providers need to assess whether their current access controls and monitoring can detect sophisticated attacks using rival AI tools; security teams should review which third-party AI services have permissions in their environments.
Grouped: similar headlines.
- Ransomware and Leak Sites
New Settra Ransomware Variant Deployed in Attacks on Retail and Manufacturing
Huntress researchers identified a new ransomware variant called Settra and documented post-compromise techniques deployed in recent attacks targeting retail and manufacturing sectors. The analysis provides insight into how the threat actor operationalizes its access after initial compromise.
Why it matters: Retail and manufacturing organizations need to assess exposure to Settra and review post-compromise detection capabilities, given active deployment and documented attack patterns.
Grouped: similar headlines.
- Threat Intelligence
Hacking group ‘NightEagle’ targeting China’s high-tech sector expands operations to Russia
Kaspersky documented multiple incidents over the past year involving a hacking group called NightEagle, which had previously targeted China's high-tech sector and has now expanded operations to Russian businesses.
Why it matters: Organizations in Russia's technology and business sectors face active targeting by NightEagle; practitioners should review incident logs for associated indicators of compromise and strengthen monitoring on systems with intellectual property or strategic value.
- OT, ICS, and Critical Infrastructure
Health-ISAC sets nine-domain MedTech cybersecurity baseline to guide medical device procurement, deployment
Health-ISAC published guidance on nine cybersecurity capability domains for evaluating and procuring medical devices, including authentication, authorization, encryption, logging, patching, secure remote access, privacy controls, cloud connectivity, and network boundary protection. The framework acknowledges that medical devices face unique constraints due to extended lifecycles, regulatory requirements, and patient-safety considerations, and recommends using documented compensating controls and layered protections when devices cannot natively support all security capabilities. The guidance is intended as a decision-support tool rather than a certification standard, emphasizing shared responsibility between healthcare delivery organizations and manufacturers throughout the product lifecycle.
Why it matters: Healthcare delivery organizations and medical device manufacturers need practical procurement and deployment criteria to align security expectations while managing device constraints and patient-safety risks.
- Vulnerabilities and PatchesCVE-2025-40572CVE-2025-40573
Nozomi identifies 12 vulnerabilities in Siemens SCALANCE LPE9403, enabling root access and OT network attacks
Nozomi Networks Labs identified 12 vulnerabilities in Siemens SCALANCE LPE9403 Local Processing Engine devices, with nine affecting firmware below V4.0 HF0 and three affecting systems with SINEMA Remote Connect Edge Client through V2.1. Multiple vulnerabilities can be chained to achieve root-level access, enabling attackers to manipulate telemetry, suppress alarms, exfiltrate operational data, or disrupt critical services across OT networks. Siemens has released patches, and asset owners must apply updates immediately and implement network segmentation to reduce exposure.
Why it matters: Organizations operating Siemens SCALANCE LPE9403 devices in industrial environments face immediate risk of data manipulation, service disruption, and lateral movement across multiple OT networks if devices are not updated to firmware V4.0 HF0 or later, or if SINEMA Remote Connect clients are not updated to version V2.1 or later.
- Breaches and Incidents
Meta’s Copyright System Is Being Weaponized Against Albanian Protesters
Instagram suspended a large number of accounts amid three months of anti-government demonstrations in Albania known as the Flamingo Revolution, raising concerns that the platform's copyright enforcement mechanisms are being misused to target activists. Observers worry the account closures reflect coordinated abuse of Meta's complaint systems rather than legitimate intellectual property violations. The timing and scale of suspensions suggest potential coordination to silence protest activity.
Why it matters: Albanian protesters, journalists covering demonstrations, and platform users relying on Instagram for organizing face account loss and communication disruption; practitioners should monitor whether authoritarian actors are exploiting copyright systems to suppress dissent globally.
- Research and Reports
SANS launches new Dynamic Incident Response framework to help security teams adapt as attacks evolve
SANS Institute released Dynamic Incident Response: A Framework for Security Teams, a 720-page rebuild of its incident response methodology authored by SANS Fellow Joshua Wright. The framework, called DAIR (Dynamic Approach to Incident Response), replaces a linear model with an adaptive, iterative process designed to handle shifting evidence and evolving attack conditions during active incidents. The book is free in digital formats and teaches teams how to adjust priorities and strategies as circumstances change, rather than following fixed stages.
Why it matters: Incident response teams need a framework that reflects how real attacks unfold with changing scope and priorities; this resource provides practical guidance for adapting containment and recovery strategies when initial assumptions prove wrong.
- Vulnerabilities and PatchesCVE-2026-85889
Microsoft Patches CVSS 10.0 Azure AI Foundry Flaw Enabling Unauthorized Privilege Escalation
Microsoft released patches for CVE-2026-85889, a maximum-severity flaw in Azure artificial intelligence (AI) Foundry that stems from missing authentication on a critical function. An unauthenticated attacker could exploit the vulnerability over a network to escalate privileges.
Why it matters: Organizations using Azure AI Foundry should apply the available patches to prevent unauthorized privilege escalation by remote attackers.
- Breaches and Incidents
Raon data leak: Insider leak of 1,894 cases went undetected for 4 years
An insider leaked 1,894 cases of internal data from South Korea's Raon heavy ion accelerator facility in 2022, remaining undetected for approximately four years until discovery in 2026. The Raon project, a state-funded facility built at a cost of approximately 1.5 trillion Korean won, had sensitive information exposed through this insider threat.
Why it matters: Research institutions and government agencies relying on insider access controls face exposure of sensitive scientific and operational data; this case demonstrates how insider threats can persist undetected for years without robust monitoring and audit controls.
- Vulnerabilities and PatchesCVE-2026-91843
AI-Built Exploit and Sign-In Flaw Opened Path to Internal OpenAI Code
Hacktron researchers discovered vulnerabilities combining an artificial intelligence (AI)-built exploit and a sign-in flaw that granted access to OpenAI employee accounts. The researchers disclosed the issues through OpenAI's bug bounty program. The flaws created a pathway to internal OpenAI code.
Why it matters: OpenAI customers and employees face account compromise risk if similar vulnerabilities exist in production systems; security teams should review account access controls and AI-assisted attack methodologies.
Grouped: similar headlines.
- Vulnerabilities and Patches
Microsoft fixes bug behind ‘Defender Antivirus is turned off’ alerts
Microsoft resolved a known issue that generated false alerts indicating Defender Antivirus was disabled following recent updates. The bug affected users who received misleading notifications despite the antivirus remaining active.
Why it matters: Organizations using Microsoft Defender need to verify their antivirus status is not based on these erroneous alerts and confirm actual protection status to avoid operational confusion.
- Vulnerabilities and PatchesCVE-2025-39964CVE-2026-53266
CISA Adds Two Known Exploited Vulnerabilities to Catalog
CISA added CVE-2025-39964 and CVE-2026-53266, two Linux kernel vulnerabilities subject to active exploitation, to its Known Exploited Vulnerabilities Catalog. Binding Operational Directive 26-04 requires federal agencies to prioritize patching of KEV Catalog vulnerabilities on publicly exposed assets, while CISA encourages all organizations to adopt the same risk-based approach.
Why it matters: Federal agencies must treat these two Linux kernel vulnerabilities as high priority for patching on exposed systems; all other organizations should adopt similar prioritization for KEV Catalog entries to reduce attack surface against active exploitation.
- Breaches and Incidents
23 Million User Records Compromised in Gyazo Data Breach
Helpfeel, maker of the Gyazo screenshot tool, disclosed a data breach affecting 23 million user records. An attacker exploited a vulnerability in the image upload server to access the data.
Why it matters: Gyazo users should check if their accounts were affected and reset credentials; security teams tracking third-party service risks need to evaluate whether Gyazo is used in their organization and assess exposure.
Grouped: similar headlines.
- Research and Reports
How to Build a Segmentation Program That Actually Reduces Risk
This article provides guidance on designing and implementing network segmentation strategies to lower organizational risk exposure. It addresses best practices for segmentation program development and deployment.
Why it matters: Security teams responsible for network architecture need actionable segmentation frameworks to limit lateral movement and contain breaches; this resource helps practitioners design programs aligned with risk reduction goals.
- AI Security
Are AIs Still Struggling with CAPTCHAs?
Anthropic's security documentation reveals that Claude struggles to solve CAPTCHAs, including a test where the model repeatedly questioned its choices when identifying non-matching shapes and eventually let the challenge expire. The model exhibited difficulty recognizing new windows and navigating the interface, at one point theorizing the test was broken. Separate reports suggest competing models may have better success with similar challenges, though verification remains unclear.
Why it matters: Security teams relying on CAPTCHAs as a defense against automated artificial intelligence (AI) agents should assess whether this protection remains effective against current and emerging language models, as inconsistent performance across vendors complicates threat modeling.
- Cloud and SaaS Security
An Abandoned CDN Domain Was Re-Registered. Thousands of Sites Still Call It.
A content delivery network (CDN) domain expired and was re-registered by a new owner in July 2025, while thousands of websites and code repositories retained hard-coded references to the old CDN hostnames. The new registrant now controls the infrastructure that these sites attempt to load assets from, creating a potential attack surface.
Why it matters: Website operators and developers relying on the expired CDN domain face supply chain compromise risk if the new owner serves malicious content; practitioners should audit for hard-coded external dependencies and enforce subdomain monitoring.
- Vulnerabilities and Patches
Plugin4Shell Lets Repository Owners Swap Pinned Plugin Code Across Four AI Coding Agents
A vulnerability in four artificial intelligence (AI) coding agents allows repository owners to replace pinned plugin code with malicious versions, bypassing version locks that agents enforce. Anthropic patched the flaw in Claude Code 2.1.179, and OpenAI addressed it in Codex 0.146.0, while GitHub Copilot's status remained unclear.
Why it matters: Developers using these AI agents face supply chain risk if their plugin repositories are compromised; organizations should update to patched versions and review plugin sourcing practices.
- Vulnerabilities and Patches
Microsoft Patches 18 Vulnerabilities in AI, Cloud Products
Microsoft released patches for 18 vulnerabilities affecting Azure and artificial intelligence (AI)-branded products, with most flaws enabling privilege escalation. The fixes address security issues across its cloud and AI services.
Why it matters: Organizations using Azure and Microsoft AI services should review and apply these patches to prevent privilege escalation attacks that could allow unauthorized access to sensitive resources.
- Threat Intelligence
WeaselBiscuit Stealer Spreads via 13 npm Packages to Harvest Chrome Extension Storage
Researchers identified 13 malicious npm packages distributing a JavaScript stealer named WeaselBiscuit that targets Chrome extension storage. The malware shares functional similarities with BeaverTail and other tools linked to the DPRK's Contagious Interview campaign.
Why it matters: Developers who installed these npm packages may have credentials and sensitive data harvested from browser extensions; practitioners should audit dependencies and scan for WeaselBiscuit indicators of compromise (IOCs) in their environments.
- AI Security
Bots with good manners are better at fooling people on social media
A Surfshark study of 1,722 participants found that people identified only 40% of artificial intelligence (AI)-generated bot comments in a social media setting. The AI bots most effective at evading detection were those using polite, measured language rather than aggressive or provocative tones.
Why it matters: Security and communications professionals should understand that AI-generated content modeled on respectful behavior poses greater risks for social manipulation and platform trust than overtly hostile bots.
- Cloud and SaaS Security
A Vault with a Heap-View: The Uncomfortable Space Between AgentCore Harness and Identity
A security analysis reveals that default configurations in AWS AgentCore Harness are vulnerable to prompt injection attacks that can exfiltrate credentials. The research identifies the misconfiguration and provides mitigation steps for securing agents.
Why it matters: Organizations running AWS AgentCore Harness with default settings face immediate credential theft risk; practitioners should review configurations and apply recommended protections now.
- Vulnerabilities and Patches
CISA Upgrades Vulnerability Reporting Platform with More Automation
The US Cybersecurity and Infrastructure Security Agency (CISA) is transitioning to a new vulnerability coordination platform named VINCE-NT. The upgrade aims to enhance automation capabilities for vulnerability reporting and management.
Why it matters: Security researchers, vendors, and other coordinators who report vulnerabilities to CISA should understand the migration timeline and any new processes or requirements for VINCE-NT to ensure continued effective vulnerability disclosure.
- Vulnerabilities and Patches
Patchstack Now Protects Your AI-Built Apps
Patchstack has extended its vulnerability mitigation platform to protect artificial intelligence (AI)-built applications on platforms like Lovable, Replit, and Base44, moving beyond its original WordPress focus. The service uses runtime mitigation to identify and neutralize vulnerabilities without code changes or rebuilds, and introduces Live Hardening to filter sensitive data leaks. Early access is available free today, with standard pricing to follow.
Why it matters: Developers building with AI tools and low-code platforms face escalating security risks from hundreds of unpatched dependencies, and most lack the technical skill or resources to manage them manually; this runtime mitigation approach lets them operate safely while maintaining control over their patch schedule.
- Breaches and Incidents
Brevo Supply Chain Attack Injects Malware Into 100,000 Websites
Attackers compromised an application programming interface (API) key and deployed a malicious Cloudflare worker to inject scripts across approximately 100,000 websites through Brevo, a marketing platform. The supply chain attack leveraged trusted infrastructure to distribute malware at scale.
Why it matters: Website operators and Brevo customers face potential credential theft, malware distribution, and defacement; practitioners should audit API key permissions, review Cloudflare worker deployments, and check their sites for injected scripts immediately.
- Vulnerabilities and Patches
New Check Point flaw lets hackers execute code with root privileges
Check Point Software released security updates to address a critical vulnerability that allows attackers to execute code with root privileges on management systems. The flaw affects Check Point management infrastructure and required patching to contain the exposure.
Why it matters: Organizations running Check Point management systems must apply these updates immediately, as root-level code execution on management infrastructure can lead to full compromise of protected networks.
- Threat Intelligence
Claimed Bug Bounty Hunter Likely Used LLM to Build PhantomRaven npm Stealer
A financially motivated threat actor distributed PhantomRaven, a JavaScript information stealer, through the npm package registry. Security analysis indicates high confidence the malware was written using a large language model, based on verbose comments, placeholder code, and statistical token patterns.
Why it matters: Developers using npm must audit their dependencies immediately for PhantomRaven, as this stealer can harvest sensitive data from applications and build environments.
- AI Security
Arcjet brings security controls and audit trails to AI agents
Arcjet launched agent runtime security, a product designed to secure artificial intelligence (AI) agents in production workflows by providing observability, enforcement, and audit capabilities. The platform enables engineering teams to discover which AI agents are active, control their actions, and maintain compliance records of agent activities.
Why it matters: Engineering and security teams building AI agents need this control layer to govern agent behavior in production, manage audit trails for compliance, and prevent unauthorized agent actions.
- Vulnerabilities and PatchesCVE-2026-58138
Critical Orkes Conductor Vulnerability Exploited in Attacks
CVE-2026-58138 is an unauthenticated remote code execution (RCE) vulnerability in Orkes Conductor that can be exploited through inline workflow definitions. The flaw has been observed in active attacks against deployments.
Why it matters: Organizations running Orkes Conductor are exposed to unauthenticated RCE without authentication; patching is critical to prevent immediate compromise.
Grouped: similar headlines and the same Common Vulnerabilities and Exposures (CVE) record (CVE-2026-58138).
- Cloud and SaaS Security
Android apps can now check security patches down to individual device components
Google released stable versions of AndroidX Security State libraries that let developers check the security patch status of individual device components rather than the entire device. Phone manufacturers and developers building over-the-air (OTA) update systems can now use a standardized approach to determine which security updates are available and ready for installation on specific devices.
Why it matters: Android app developers and OTA system builders need these libraries to provide users with precise security patch information and improve update targeting across diverse device hardware configurations.
- Ransomware and Leak Sites
Manufacturing Accounts for 22% of all Ransomware Victims
Black Kite reports that manufacturing accounts for 22% of all ransomware victims, making it the most targeted sector. The industry experienced a significant increase in ransomware incidents during the first half of 2026.
Why it matters: Manufacturing operations and supply chain partners face elevated ransomware risk and should review incident response plans, network segmentation, and backup strategies to prepare for potential attacks.
Grouped: similar headlines.
- Vulnerabilities and Patches
Microsoft fixes broken copy and paste for Excel 2016 users
Microsoft addressed a copy-and-paste malfunction affecting some Excel 2016 users following the September 2026 security update KB5002914. The issue has been resolved in a follow-up fix.
Why it matters: Excel 2016 users who installed KB5002914 need to apply the corrective update to restore normal clipboard functionality and avoid workflow disruptions.
- Industry and Vendors
MIND Secures $72 Million for AI-Powered DLP
MIND, a data loss prevention platform vendor, raised $72 million in funding to accelerate product development and expand into enterprise markets. The company focuses on artificial intelligence (AI)-powered capabilities for its data loss prevention offerings.
Why it matters: Enterprise security teams evaluating data loss prevention tools should monitor MIND's enhanced capabilities and market expansion, as the funding signals competitive investment in AI-driven protection.
- Vulnerabilities and PatchesCVE-2026-93485
WordPress 7.1.1 Maintenance and Security Release
WordPress 7.1.1 arrived on September 17, 2026, with 11 security fixes and 17 bug fixes addressing an unauthenticated stored cross-site scripting (XSS) vulnerability in the wpautop() function that processes comments on nearly all WordPress content. CVE-2026-93485, with a CVSS 3.1 score of 7.1, stems from a regular expression flaw that fails to account for quoted attribute values, allowing an attacker to inject malicious code through an ordinary comment that bypasses sanitization and executes when displayed to any site visitor. The vulnerability requires comment moderation to pass the payload to the live site, and the release also includes ten additional fixes for access control, information disclosure, and other issues requiring authenticated access.
Why it matters: WordPress site administrators and hosting providers must prioritize updating to 7.1.1 immediately, particularly sites accepting public comments, as unauthenticated attackers can inject executable code visible to all visitors once a comment is approved.
- Vulnerabilities and Patches
HTTP QUERY Method: The Grey Zone Between GET And POST.
In June 2026 the IETF published RFC 10008, which defines a new HTTP method called QUERY that combines properties of GET and POST by allowing a request body while remaining safe and idempotent. The method sits in a gap between existing HTTP verbs, and current security controls like web application firewalls (WAFs), caches, and middleware were built before QUERY existed and handle it inconsistently, creating potential bypass and poisoning risks. Nginx proxies but never caches QUERY requests, Django rejects them outright, while FastAPI and Caddy pass them through untouched, leaving organizations exposed until rules and configurations are explicitly updated.
Why it matters: Web infrastructure defenders must audit WAF signatures, application programming interface (API) gateway allowlists, CSRF middleware, and cache rules to ensure they inspect QUERY request bodies for injection attacks and prevent cache poisoning, as payloads blocked in POST may bypass detection when sent via QUERY.
- Vulnerabilities and Patches
Abandoned IoT apps keep sending sensitive data to broken servers
Researchers analyzed 61,500 abandoned Android IoT companion apps and found that nearly 75 percent contained software dependencies linked to documented vulnerabilities. Millions of users continue running these unmaintained applications for smart home devices like plugs, cameras, and thermostats that no longer receive updates.
Why it matters: Organizations and consumers running legacy IoT apps face active exploitation risk from known vulnerabilities in unpatched dependencies; security teams should audit and deprecate abandoned companion apps in their environments.
- Threat Intelligence
Hardcoded MCP credentials found in public GitHub files
Hush Security analyzed approximately 82,000 Model Context Protocol (MCP) configuration files on GitHub and identified hardcoded credentials in roughly 12% of them, including application programming interface (API) keys, access tokens, and other secrets for connected services. These exposed credentials were discoverable in publicly accessible files, creating potential pathways for unauthorized access to downstream systems.
Why it matters: Developers and organizations using MCP-integrated artificial intelligence (AI) coding tools face credential exposure risk today; security teams should audit their GitHub repositories and connected service accounts for hardcoded secrets and revoke any discovered credentials.
- Identity and Access
98% of fraudulent hires have company credentials by the time they’re caught
A HYPR report reveals that 98% of fraudulent hires gain legitimate company credentials during the 90-day gap between hiring and onboarding completion. Adversaries exploit this window by passing remote interviews and obtaining authentic credentials directly from IT departments, bypassing the need for network breaches. The research indicates that human judgment alone cannot prevent these attacks and calls for dedicated verification technology.
Why it matters: HR, IT, and identity teams need to close the onboarding credential issuance window: fraudulent employees with legitimate access pose a direct insider threat to your network and data.
- Vulnerabilities and PatchesCVE-2026-75157
CVE-2026-75157: Apache Airflow: Asset queued-events DELETE endpoints gated on Dag READ instead of Dag EDIT (asset-triggered scheduling suppression)
CVE-2026-75157 is a low-severity permission flaw in Apache Airflow before version 3.3.2, where asset queued-events DELETE endpoints incorrectly check for Dag READ permission instead of Dag EDIT permission. This allows any authenticated user with read access to a directed acyclic graph (DAG) to delete that DAG's queued asset events, silently disabling asset-triggered scheduling.
Why it matters: Apache Airflow operators should upgrade to version 3.3.2 or later to prevent authenticated users from suppressing asset-triggered scheduling through a read-permission bypass.
- Regulatory and Compliance
Most WordPress pros still lack a breach recovery plan
Melapress surveyed 319 WordPress professionals who build and operate WordPress sites and found that most have experienced at least one security incident. However, fewer than three in ten respondents have a documented breach recovery plan that defines response roles, backup locations, and notification procedures.
Why it matters: WordPress agencies, developers, and site owners need to establish breach recovery plans now: lack of pre-incident coordination delays response, extends dwell time, and increases damage when incidents occur.
- Industry and Vendors
New infosec products of the week: September 18, 2026
This article covers new security product releases from multiple vendors in the week of September 18, 2026, including offerings from Akuity, Bitsight, Cohesity, Dataminr, Nozomi Networks, and Tuskira. Dataminr announced Advanced for Corporate Security, which uses agentic artificial intelligence (AI) to help security teams predict and verify threats. The article provides a weekly roundup of product announcements without detailed technical specifications or comparative analysis.
Why it matters: Security practitioners evaluating new tools for threat prediction, asset monitoring, and operational resilience should review these vendor announcements to assess whether agentic AI capabilities and other new features meet their organization's needs.
- Threat Intelligence
North Korean hackers infect thousands of devices across 100 countries as part of ‘WaterPlum’ campaign
The FBI, Defense Department, and international law enforcement agencies from Japan, Australia, and Germany issued an advisory about WaterPlum, a group of North Korean cyber actors. The group conducts targeted campaigns against job applicants by impersonating artificial intelligence (AI) and blockchain companies to steal cryptocurrency. The campaign has infected thousands of devices across more than 100 countries.
Why it matters: Organizations and individuals recruiting for AI or blockchain roles face credential theft and device compromise from a state-sponsored threat; practitioners should brief hiring teams and monitor for malicious job postings and employment-related phishing linked to WaterPlum.
- AI Security
USA’s Venezuela takeover comes with bonus exposure to Chinese AI surveillance tech
The Australian Strategic Policy Institute reports that Venezuela is adopting Chinese artificial intelligence (AI) systems to enhance its existing surveillance infrastructure, including technology from iFlytek, a company the US sanctioned for roles in Uyghur repression. The think tank warns that Venezuela may become the Western Hemisphere's most advanced adopter of China's latest AI surveillance tools and calls on the US State Department to dismantle the apparatus. The situation presents a tension: the US has taken control of Venezuela while that country deepens reliance on Chinese surveillance technology to maintain political control.
Why it matters: Government decision-makers and intelligence officials need to assess whether dismantling Chinese surveillance systems in Venezuela is feasible and aligned with broader China containment strategy, while civil liberties advocates should monitor whether the US will act on ASPI's recommendations to prevent AI-enabled authoritarianism in the region.
- Cloud and SaaS Security
Mind Raises $72M to Rebuild DLP Around AI Agents
Mind secured $72 million in funding to develop a data loss prevention (DLP) platform that integrates endpoint controls with artificial intelligence (AI) agents. The AI agents analyze data lineage, identify sensitive data movement, and help guide employees through policy violations.
Why it matters: Security teams managing data exfiltration risk need AI-assisted DLP to reduce alert fatigue and distinguish legitimate data flows from policy violations, especially as organizations scale endpoint monitoring.
- OT, ICS, and Critical Infrastructure
States Expand Cyber Support Beyond Their Own Networks
States are expanding cybersecurity support to local utilities and critical infrastructure operators outside their direct control to address protection gaps. The effort requires more than financial grants and software tools, demanding sustained monitoring, operational technology (OT) expertise, and consistent vendor oversight across fragmented local operators.
Why it matters: Utility operators and critical infrastructure managers need to understand state-level support programs available to them and engage with state authorities to ensure adequate monitoring and vendor accountability for their systems.
- Industry and Vendors
CrowdStrike SafeMind: When the Best Offense Builds the Best Defense
CrowdStrike announced SafeMind, a security offering designed to strengthen defensive capabilities through offensive security insights. The product leverages threat intelligence to help organizations identify and remediate vulnerabilities before adversaries can exploit them.
Why it matters: Security teams evaluating endpoint protection and threat prevention tools should understand how CrowdStrike's new offering fits their detection and response workflow.
- Threat Intelligence
Attacker infrastructure, but vibe-coded: tracking the evolution of credential harvesting platforms
Two credential harvesting platforms, Loot and UltraVault, are examined for their operational characteristics and methods. The analysis reveals how attackers use Amazon Bedrock to validate stolen secrets, demonstrating a technical evolution in how harvested credentials are processed and verified.
Why it matters: Security teams managing credential exposure monitoring and cloud service abuse detection need to understand these platforms and the use of legitimate services like Amazon Bedrock in post-compromise validation workflows.
- Vulnerabilities and Patches
Vulnerability in Kaspersky Secure Mail Gateway (September 18, 2026)
A vulnerability in Kaspersky Secure Mail Gateway permits remote code execution by an attacker. The flaw was discovered on September 18, 2026.
Why it matters: Organizations running Kaspersky Secure Mail Gateway face remote code execution risk and should prioritize identifying affected instances and applying patches or mitigations.
- Cloud and SaaS Security
One SOC, 100 projects: running centralized alert triage on Elastic Security Serverless
Elastic Security Serverless now supports cross-project search (CPS) to enable centralized security operations across up to 100 isolated projects while maintaining separate data ownership and environments. Detection rules run once on a central origin project and generate alerts into a single queue, allowing analysts to triage and investigate across linked projects without duplication or tenant isolation breaches. The company validated this model under realistic SOC load, publishing guidance on query patterns, sizing, and operational boundaries such as response actions remaining tied to individual projects.
Why it matters: Security operations teams managing multiple environments, regions, or customers can now consolidate alert triage and detection into one queue while preserving data isolation and compliance boundaries, reducing the operational overhead of managing dozens of rule copies and scattered alert contexts.
- Vulnerabilities and Patches
AI coding agents' 0-click RCE flaw could hand attackers keys to the kingdom
A zero-click remote code execution vulnerability called Plugin4Shell affects major artificial intelligence (AI) coding agents including Anthropic's Claude Code, OpenAI's Codex, Google's Gemini CLI, and Microsoft Copilot by exploiting how they verify plugin commits from trusted marketplaces. The flaw allows attackers to swap legitimate code with malicious payloads while maintaining the appearance of valid security pinning, potentially granting full system access to anyone whose agent downloads a compromised plugin. Anthropic and OpenAI have patched their tools, but Google deprecated Gemini CLI without patching, and Microsoft has not yet released a fix for Copilot despite six months of disclosure.
Why it matters: Organizations deploying artificial intelligence (AI) coding agents across developer teams face immediate remote code execution risk if they run unpatched versions; approximately 90 percent of Fortune 500 companies use GitHub Copilot, which remains vulnerable on non-GitHub marketplace platforms like Bitbucket where the attack succeeds.
Grouped: the same names (CLAUDE CODE, GEMINI CLI).
- Vulnerabilities and PatchesCVE-2026-73639
CVE-2026-73639: Imager::File::PNG versions from 1.003 before 1.004 for Perl write past the end of the row buffer reading a PNG with a tRNS transparency chunk in read_direct8
CVE-2026-73639 affects Imager::File::PNG for Perl versions 1.003 through 1.003, allowing a buffer overflow when processing PNG files with transparency (tRNS) chunks during 8-bit direct color reads. An attacker can trigger the vulnerability by supplying a malicious PNG file to an application using the affected library.
Why it matters: Perl developers using Imager::File::PNG to process PNG images face potential remote code execution if they accept untrusted image files; upgrade to version 1.004 or later.
- Vulnerabilities and PatchesCVE-2026-73638
CVE-2026-73638: Imager versions from 0.45_02 before 1.035 for Perl read outside the EXIF block via unchecked start offsets in tiff_load_ifd
CVE-2026-73638 affects Imager, a Perl image processing module, in versions 0.45_02 through 1.034. The vulnerability allows reading outside the EXIF block due to unchecked start offsets in the tiff_load_ifd function when processing TIFF files.
Why it matters: Developers and systems using Imager to process untrusted TIFF images face potential information disclosure; upgrade to version 1.035 or later.
- Threat Intelligence
Inside the Modern SOC: Defending the Cross-Environment Pivot
This article discusses cross-environment attacks and how security operations centers (SOC) can investigate complete attack paths across multiple systems. Unit 42 Managed Extended Detection and Response and Security Information and Event Management (XSIAM) is presented as a tool to help SOC teams defend against these attacks.
Why it matters: SOC teams responsible for detecting and responding to threats across cloud, on-premises, and hybrid environments need visibility into complete attack chains to identify and stop lateral movement before compromise spreads.
- Threat Intelligence
ThreatsDay: Self-Rewriting Agents, 800+ Flaws Patched, Insider SIM Swaps and 22 More New Stories
A weekly threat roundup covers emerging attack patterns including self-rewriting artificial intelligence (AI) agents, over 800 patched vulnerabilities, insider-based SIM swap attacks, and related security developments. The article emphasizes that attackers exploit multiple attack surfaces spanning AI tools, misconfigured services, legacy vulnerabilities, weak authentication, and software-as-a-service platforms. Security defenders face persistent threats from both novel techniques and reuse of established attack methods.
Why it matters: Security teams must monitor AI-based threats, patch management backlogs, insider threats, and subscription software risks as attackers continue to find and exploit weaknesses across the full infrastructure stack.
- Vulnerabilities and PatchesCVE-2026-77179
Critical Docker Sandboxes Flaw Lets Malicious Guest Code Read and Modify macOS Host Files
A critical vulnerability in Docker Sandboxes on macOS allows malicious code within a virtual machine to escape the project directory and read or modify arbitrary files on the host system with the privileges of the account running the virtual machine. Docker disclosed CVE-2026-77179 on September 15, and the flaw carries a CVSS score of 9.4.
Why it matters: macOS users running Docker Sandboxes with untrusted code or containers face immediate risk of host file compromise; patching or upgrading Docker Sandboxes should be prioritized.
- Threat Intelligence
Iran-Linked Handala Hack Tied to HEAVYGRAM Telegram Backdoor That Can Steal Passwords
Security researchers linked the Iran-affiliated persona Handala Hack to HEAVYGRAM, a Telegram-based backdoor capable of remote command execution, credential theft, and data exfiltration. The malware operates alongside CRUDEEXCLUDE, a Delphi-based utility, to enable surveillance and system reconnaissance.
Why it matters: Organizations and individuals targeted by Iranian threat actors face credential and data theft risks; practitioners should monitor for HEAVYGRAM indicators of compromise (IOCs) and suspicious Telegram-based command and control activity.
- Threat Intelligence
Beware the SparroWock: The backdoor that bites, the commands that catch
ESET researchers have documented SparroWocky, a new backdoor deployed by the FamousSparrow advanced persistent threat (APT) group. The backdoor represents an evolution in the group's offensive capabilities and infrastructure.
Why it matters: Organizations targeted by FamousSparrow APT need to understand SparroWocky's behavior, command set, and indicators of compromise (IOCs) to detect and respond to intrusions using this backdoor.
Grouped: similar headlines.
- Ransomware and Leak Sites
Ukrainian National Sentenced to Four Years in Prison for Wire Fraud Conspiracy in Connection with Conti Ransomware
Oleksii Oleksiyovych Lytvynenko, a Ukrainian national, received a four-year prison sentence for conspiracy to commit wire fraud related to Conti ransomware operations. The conviction reflects ongoing enforcement against individuals involved in ransomware-as-a-service schemes that have targeted organizations globally.
Why it matters: Organizations and incident responders need to track law enforcement actions against Conti-affiliated actors to understand the group's operational status and identify potentially compromised infrastructure or victim data in circulation.
- Threat Intelligence
Phantom in the vault: Obsidian abused to deliver PhantomPulse RAT
Elastic Security Labs identified a social engineering campaign that leverages Obsidian's community plugin ecosystem to distribute PhantomPulse remote access trojan (RAT). The attack targets financial and cryptocurrency sector professionals through LinkedIn and Telegram with elaborate pretexts designed to trick them into installing malicious plugins.
Why it matters: Finance and cryptocurrency professionals are at immediate risk of account compromise and data theft if they install plugins from untrusted sources; practitioners should educate users on plugin vetting and restrict Obsidian plugin installation policies.
- Threat Intelligence
Hooked on Linux: Rootkit Detection Engineering
This article examines techniques for detecting Linux rootkits, with emphasis on moving beyond static detection methods toward behavioral analysis. The piece is presented as the second installment in a series on rootkit detection engineering.
Why it matters: Linux administrators and security teams need practical approaches to identify rootkits that evade file-based signatures, making behavioral detection methods critical for protecting systems already in production environments.
- Threat Intelligence
How we caught the Axios supply chain attack
A researcher named Joe Desimone describes discovering the Axios supply chain attack and developing a proof of concept tool to identify it. The account details how the attack was caught using a tool built in a single afternoon.
Why it matters: Developers and security teams using Axios should understand how this supply chain compromise was detected and apply lessons to their own dependency verification processes.
- Threat Intelligence
Fake Installers to Monero: A Multi-Tool Mining Operation
Elastic Security Labs analyzed a sustained malware campaign distributing remote access trojans (RATs), cryptominers, and cost-per-action fraud tools via counterfeit installer files. The operation evolved across multiple campaigns and generated revenue through Monero payouts.
Why it matters: Practitioners deploying endpoint detection and response (EDR) solutions and monitoring installer vectors should review Elastic's indicators of compromise (IOCs) to detect or block this active infrastructure, as the campaign continues to target users through social engineering.
- Threat Intelligence
Elastic Security Labs uncovers BRUSHWORM and BRUSHLOGGER
Elastic Security Labs identified two custom malware tools deployed against a South Asian financial institution: BRUSHWORM, a modular backdoor capable of spreading via USB, and BRUSHLOGGER, a DLL-side-loaded keylogger. The discovery reveals a targeted intrusion leveraging both network and removable media propagation channels.
Why it matters: Financial institutions and their security teams should assess exposure to USB-based malware delivery and side-loading attacks, and review endpoint detection rules for these custom tools and techniques.
- Threat Intelligence
Illuminating VoidLink: Technical analysis of the VoidLink rootkit framework
Elastic Security Labs published technical analysis of VoidLink, a Linux malware framework that leverages Loadable Kernel Modules and eBPF (extended Berkeley Packet Filter) to establish persistence on compromised systems. The framework combines kernel-level and in-kernel virtual machine techniques to evade detection.
Why it matters: Linux infrastructure operators and threat hunters need to understand VoidLink's kernel-based evasion tactics to detect and defend against this sophisticated persistence mechanism in enterprise environments.
- Threat Intelligence
From Invitation to Infection: How SILENTCONNECT Delivers ScreenConnect
SILENTCONNECT is a multi-stage loader that uses VBScript, PowerShell, and process environment block (PEB) masquerading techniques to covertly install ScreenConnect remote monitoring and management (RMM) software. The malware operates in multiple stages to evade detection and establish unauthorized remote access on compromised systems.
Why it matters: Organizations and incident responders need to monitor for SILENTCONNECT delivery chains and ScreenConnect installations outside approved deployments, as this loader enables attackers to gain persistent remote code execution and control over victim systems.
- Research and Reports
Hooked on Linux: Rootkit Taxonomy, Hooking Techniques and Tradecraft
A security research series examines how Linux rootkits have evolved across different architectural layers, from userland shared object manipulation through kernel modules to contemporary approaches leveraging extended Berkeley Packet Filter (eBPF) and io_uring. The analysis documents the progression of hooking techniques that attackers have employed to maintain persistence and stealth on compromised systems.
Why it matters: Practitioners defending Linux infrastructure need to understand the full spectrum of rootkit attack vectors, from legacy to emerging methods, to design detection and prevention strategies across kernel and userland boundaries.
- Research and Reports
Beyond Behaviors: AI-Augmented Detection Engineering with ES|QL COMPLETION
Elastic introduced an ES|QL COMPLETION command that integrates large language model reasoning into detection rules, allowing detection engineers to perform alert triage within the platform without external tools. The feature enables intelligent processing of security signals through artificial intelligence (AI) capabilities embedded directly in rule logic.
Why it matters: Detection engineers and security operations centers (SOCs) can reduce manual alert triage overhead and improve response speed by automating intelligent filtering within their existing detection workflows.
- Threat Intelligence
MIMICRAT: ClickFix Campaign Delivers Custom RAT via Compromised Legitimate Websites
Elastic Security Labs identified a ClickFix campaign that leverages compromised legitimate websites to deliver MIMICRAT, a custom remote access trojan (RAT) written in C. The five-stage infection chain culminates in a payload capable of command and control (C2) manipulation, stealing authentication tokens, and establishing SOCKS5 tunnels for lateral movement.
Why it matters: Organizations with web-facing assets should monitor for signs of site compromise and suspicious redirects that could distribute MIMICRAT; endpoint teams need to watch for the RAT's token theft and tunneling behavior to detect post-exploitation activity.
- Vulnerabilities and Patches
The Immutable Illusion: Pwning Your Kernel with Cloud Files
Researchers have identified a class of vulnerabilities that allows threat actors to bypass security restrictions and compromise kernel trust chains. The attack leverages cloud file systems to achieve kernel-level code execution through immutable file mechanisms that are not properly validated.
Why it matters: Organizations running cloud-based systems and kernels are exposed to privilege escalation and arbitrary code execution: administrators should audit kernel security mechanisms and file system validation logic.
- Threat Intelligence
BADIIS to the Bone: New Insights to a Global SEO Poisoning Campaign
Elastic Security Labs identified an intrusion at a Southeast Asian multinational organization in November 2025 involving BADIIS malware deployment on a Windows web server. The incident revealed post-compromise techniques and tooling consistent with findings reported elsewhere in industry research.
Why it matters: Organizations running Windows web servers face exposure to BADIIS malware and SEO poisoning tactics; practitioners should review server access logs and web application activity for similar post-compromise indicators.
- Threat Intelligence
NANOREMOTE, cousin of FINALDRAFT
Security researchers identified a fully-featured backdoor named NANOREMOTE that shares characteristics with previously documented malware and resembles the FINALDRAFT implant. The discovery expands the known toolkit of an existing threat actor family.
Why it matters: Organizations and defenders tracking FINALDRAFT-related threats should monitor for NANOREMOTE indicators to identify potential compromises of systems already targeted by this threat actor.
- Threat Intelligence
RONINGLOADER: DragonBreath’s New Path to PPL Abuse
Elastic Security Labs identified RONINGLOADER, a multi-stage loader tied to DragonBreath that deploys an updated gh0st remote access trojan (RAT) variant. The malware uses signed drivers, thread-pool injection, and Privileged Process Light (PPL) abuse to disable Windows Defender and evade Chinese endpoint detection and response (EDR) tools.
Why it matters: Organizations in or targeted by Chinese threat actors need to assess whether their detection tooling and driver-signing processes can spot this technique, and whether PPL-aware defenses are in place.
- Threat Intelligence
TOLLBOOTH: What's yours, IIS mine
REF3927 exploits publicly disclosed ASP.NET machine keys to compromise Internet Information Services (IIS) servers and deploy TOLLBOOTH search engine optimization (SEO) cloaking modules worldwide. The threat actor leverages known key material to gain unauthorized access and inject malicious SEO redirects at scale.
Why it matters: IIS administrators running ASP.NET applications are exposed if they have not rotated machine keys or patched systems; TOLLBOOTH deployment compromises search visibility and site integrity, requiring immediate key rotation and server inspection.
- Research and Reports
NightMARE on 0xelm Street, a guided tour
Elastic Security Labs developed nightMARE, a Python library designed to help malware researchers scale analysis and develop configuration extractors. The tool enables extraction of intelligence indicators from malware samples.
Why it matters: Malware analysts and security researchers can adopt this library to accelerate configuration extraction and indicator carving workflows.
- Threat Intelligence
WARMCOOKIE One Year Later: New Features and Fresh Insights
Elastic Security Labs has released an updated analysis of the WARMCOOKIE backdoor approximately one year after its initial discovery. The report presents new technical features and additional insights into the malware's capabilities and behavior.
Why it matters: Security teams using Elastic tools and defenders tracking backdoor threats need the latest WARMCOOKIE technical details to improve detection and response for this persistent threat.
- Research and Reports
FlipSwitch: a Novel Syscall Hooking Technique
FlipSwitch presents a novel system call hooking technique designed to circumvent Linux kernel security defenses. The research demonstrates a new attack vector in the evolving contest between threat actors and security mechanisms.
Why it matters: Security teams defending Linux systems need to understand this kernel-level attack technique and evaluate whether their detection and prevention controls can identify or block syscall hooking attempts.
- AI Security
MCP Tools: Attack Vectors and Defense Recommendations for Autonomous Agents
Research identifies attack vectors against autonomous agents that use Model Context Protocol (MCP) tools, including tool poisoning, orchestration injection, and rug-pull redefinitions. The study outlines practical defense strategies to mitigate these exposures.
Why it matters: Organizations deploying autonomous agents with MCP integrations need to understand and implement these defenses to prevent attackers from compromising tool behavior and agent decision-making.
- Research and Reports
Investigating a Mysteriously Malformed Authenticode Signature
A technical investigation examines a Windows Authenticode validation failure, tracing vague error codes to undocumented kernel routines. The analysis provides insights into how malformed digital signatures can bypass or trigger unexpected validation behaviors in the Windows authentication system.
Why it matters: Security practitioners and developers need to understand Authenticode validation edge cases to properly verify signed binaries and detect potential signature spoofing or exploitation attempts.
- Threat Intelligence
MaaS Appeal: An Infostealer Rises From The Ashes
NOVABLIGHT is a Node.js-based infostealer offered as a malware-as-a-service (MaaS) platform. It targets user credentials and cryptocurrency wallets through automated theft and distribution mechanisms.
Why it matters: Organizations and individuals managing credentials or holding cryptocurrency assets face credential theft and financial loss if systems are infected with NOVABLIGHT; defenders should monitor for indicators of this infostealer and ensure endpoint detection and response (EDR) coverage.
- Threat Intelligence
Taking SHELLTER: a commercial evasion framework abused in-the-wild
Elastic Security Labs identified infostealers leveraging a pirated copy of SHELLTER, a commercial evasion framework, to deliver post-exploitation payloads in active attacks. The unauthorized use of this off-the-shelf tool suggests threat actors are obtaining and repurposing legitimate software to evade detection during payload deployment.
Why it matters: Organizations running endpoint detection and response (EDR) and antivirus tools need to monitor for SHELLTER signatures and behavior patterns, as criminals now use this framework to bypass standard protections during infostealer campaigns.
- Threat Intelligence
A Wretch Client: From ClickFix deception to information stealer deployment
Elastic Security Labs observed increased ClickFix campaigns that leverage a tool called GHOSTPULSE to distribute Remote Access Trojans (RATs) and information-stealing malware. ClickFix exploits user trust through deceptive technical support mechanics to establish initial access for payload deployment.
Why it matters: Practitioners managing endpoint security need to monitor for ClickFix traffic patterns and GHOSTPULSE indicators, as these campaigns directly target end-users with socially engineered fake error messages that lead to RAT and stealer infections.
- Threat Intelligence
Call Stacks: No More Free Passes For Malware
Call stacks provide critical telemetry data for detecting malware on Windows endpoints by revealing the execution chain of function calls. Elastic considers call stacks vital for security monitoring despite existing architectural constraints. The analysis examines why this telemetry source merits inclusion in endpoint detection strategies.
Why it matters: Security teams and endpoint protection practitioners need to understand call stack value for improving malware detection rates and implementing effective endpoint defense strategies.
- Threat Intelligence
Chasing Eddies: New Rust-based InfoStealer used in CAPTCHA campaigns
Elastic Security Labs documented EDDIESTEALER, a lightweight infostealer written in Rust that is being deployed through CAPTCHA-based campaigns. The malware is distributed as a commodity tool in emerging attack chains.
Why it matters: Practitioners should monitor for EDDIESTEALER distribution campaigns and understand the CAPTCHA-based delivery mechanism as a vector for credential theft and system compromise.
- Threat Intelligence
De-obfuscating ALCATRAZ
A technical analysis examines the obfuscation methods employed by ALCATRAZ, a tool used to obscure malicious code. The research documents the specific techniques that attackers leverage to evade detection and analysis.
Why it matters: Security analysts and threat hunters need to understand ALCATRAZ's obfuscation capabilities to identify and respond to malware samples using this tool in active campaigns.
- Threat Intelligence
Misbehaving Modalities: Detecting Tools, Not Techniques
This article examines execution modality as a detection approach that focuses on identifying the tools and methods used to run code rather than the behavioral patterns of attacks. The piece explains how detections centered on modality can work alongside traditional behavior-based detection strategies to improve threat identification.
Why it matters: Security teams building detection rules can improve coverage by implementing modality-focused detections that catch tool-specific indicators, complementing existing behavior-based detection systems.