The ScreenConnect Incident Has Two Different Paths to Check
Active intrusions used partner consoles to deploy rogue remote-access clients and SCKit; the toolkit also targeted technician sessions.
Cyber News archive
Permanent libraryOriginal Threat Field Notes articles move here after seven full days. Their URLs remain unchanged, and important updates can return an article to the current-news view.
Active intrusions used partner consoles to deploy rogue remote-access clients and SCKit; the toolkit also targeted technician sessions.
Barracuda reports exploitation of recently patched ESG and related product flaws where administrative interfaces are reachable.
FortiGuard traced credential guessing against exposed SQL Servers followed by malicious T-SQL and host malware.
Check Point traced PHP2Shell access through its Mobile Access portal to root privilege, persistence and, in some cases, destructive activity on ESXi.
New N-central disclosures prompted N-able to disable integrations, publish indicators for Legacy API exposure and develop further fixes.
Citrix fixed a packet-processing memory flaw affecting all customer-managed NetScaler ADC and Gateway deployments, even without optional features.
Unit 42 reports a Cairn-enabled campaign against semiconductor, defense and civil-society targets, with access or data retrieval at ten organizations.
The PostgreSQL project fixed multiple security flaws and says some deployments need index checks and a fresh full backup after updating.
Apache fixed a critical mod_http2 flaw that can lead to remote code execution in affected HTTP Server builds.
The November security release fixes HTTP/2 flaws that can crash affected services or expose response memory.
Red Hat says a namespace-matching flaw can expose another tenant’s robot tokens and allow unauthorized image operations.
Researchers report hijacked GitHub accounts planting short-lived CI changes meant to collect workflow secrets.
N-able released N-central 2026.4 HF2 after public disclosures and says customers must also upgrade Agents and Probes.
OpenSSH and Alpine fixed a password bypass affecting specific musl-based systems; Previdian reports attempts against its sensors.
CISA lists exploitation of CVE-2026-19868, which can let alternate-method credentials bypass SAML controls and SAML-enforced MFA.
QNAP reports suspected compromises through Plex2Shell and warns that updating cannot remove a payload already installed.
Apiiro found credentials in public Claude Code settings files that stored approved commands.
DigiCert says unassigned accounts could read a subset of archived customer records after an internal system became externally reachable.
OpenAI reports 948 Linux local privilege-escalation exploits and 81 high-severity Firefox findings from GPT-6 Cyber research.
CISA added CVE-2026-119977 to its exploited list after WordPress fixed a core-upgrade weakness in 7.1.4 and older branch releases.
OPSSHIELD says a fake update ran as root on about 2,950 customer servers and copied sensitive files.
Sekoia found malicious scripts on legitimate sites, lookalike domains and an analytics service that delivered known exploit chains.
Google and Mandiant document stolen model access that produced rapid unauthorized use and, in some cases, large charges.
WorkOS confirms access to an internal sales application with contacts and customer-conversation summaries, but reports no evidence of production-service access.
BTCPay reports Bitcoin and Litecoin hot-wallet thefts on servers running 2.4.1 or earlier and advises rebuilding even if theft was not noticed.
VulnCheck saw CVE-2026-58115 exploit requests against a canary; Siemens lists affected IoT2050 Advanced systems and a fixed release.
Red Hat and Ubuntu warn that CVE-2026-98432 could allow remote code execution; fix availability depends on the distribution and release.
Several chains acknowledged refrigeration disruptions. A single campaign or common device flaw has not been established.
Google fixed a critical Framework remote-code issue that needs no user interaction. The bulletin does not report active exploitation.
Bitdefender documented backup-credential theft, new domain trusts and ransomware deployment. AI assistance remains an assessment.
A third-party performance script loaded on Polymarket was replaced with code that used active trading sessions to move funds. Polymarket says it is returning $3.9 million, including pause-related losses.
AWS says affected Kinesis Producer Library versions can put resolved credentials in debug logs and process arguments. Version 1.0.8 fixes both paths.
CrowdStrike observed ARTEX activity directed at more than 200 organisations in at least 14 countries; it confirmed post-exploitation at nine, not 200.
Google and Mandiant document several actors using ARTEX and other agentic frameworks across intrusion stages; the work is still often noisy.
Cursor says a counterfeit marketplace update attempted persistence and SSH access on developer machines; installation does not by itself prove compromise.
Drupal confirms active exploitation of CVE-2026-118539, an access-bypass issue in Commerce Core.
Talos describes historical intrusion activity against Catalyst SD-WAN Controllers and a custom backdoor designed to collect keying material.
CVE-2026-76418 can reload an affected firewall when SSL VPN login and secondary authentication are configured; CISA lists known exploitation.
CVE-2026-111497 has separate patch paths across runtimes and proxies; PHP Windows packages were rebuilt without changing the runtime version.
OpenBSD 8.0 brings security hardening, while post-release errata and service compatibility require separate checks.
Arista says CVE-2026-114625 in on-prem VeloCloud Orchestrator is being exploited; even the lowest built-in tenant role can reach the vulnerable path.
The Collabora app has an exploitable command-injection path when its built-in CODE service is stopped.
Fortinet issued priority fixes across five product lines, while detailed advisories are scheduled for November.
Microsoft plans conversation-aware warnings for new external first-contact Teams chats.
A second October CVSS 10.0 GitLab issue, CVE-2026-110264, needs a newer fix than the earlier emergency release.
PHP disclosed CVE-2026-115560, an unlikely RCE path that depends on a specific runtime change to session serialization settings.
Microsoft has now described actor authentication and follow-on Graph activity in some customer environments, adding detail to the tenant-level review.
CISA now lists both PHP2Shell flaws as exploited, following documented command execution on vulnerable honeypots.
Oracle’s October advisory contains 1,295 new security patches across enterprise products and supporting components.
Cisco fixed CVE-2026-76466 in clustered Expressway-C and Expressway-E systems and lists no workaround.
Eight weak-key intermediate certificates were revoked, but separately managed trust stores may still need product updates.
The Node.js October release fixes a high-severity Brotli decoder RCE issue and certificate-validation problems.
A US federal civilian order covers DeepSeek products and certain third-party tools built using its implementations.
Files attached to older OVHcloud CLI releases were replaced after publication while the original source tags remained intact.
A narrow window of PRO v3.0.123 downloads carried malware; activated sites need compromise recovery, not only a version update.
Mandiant observed exploitation of CVE-2026-102764 before disclosure, followed by persistence and movement through a trusted support platform.
A nearby attacker can reach a vulnerable legacy Broadcom WPS parser before supplying the network password.
Citizen Lab documented targeted-ad delivery of Candiru spyware at Al Jazeera, while warning against treating ad reach as an infection count.
Adobe fixed CVE-2026-103849, an unauthenticated code-execution flaw in Magento and Adobe Commerce email templates. Researchers have reproduced attack routes, but neither source reported attacks on live stores at publication.
CISA now lists two more exploited CVEs in the Ivanti chain, after Volexity confirmed a 29 October privilege-escalation fix.
Cisco confirms exploitation of CVE-2026-76519, which can turn a crafted email attachment into root-level commands on an affected gateway.
Palo Alto Networks reports limited exploitation of an authenticated management-interface flaw that can yield root-level code execution.
SAP fixed CVE-2026-89111, an OAuth2 SSRF flaw in multi-tenant BTP applications that could send client credentials to an attacker-controlled endpoint.
CISA now lists CVE-2026-97229 as exploited; Alpine’s incident also shows why a custom GitLab image needs its bundled Ruby checked separately.
Ahsay has released hotfixes for the two exploited flaws and published compromise checks and recovery guidance.
Citrix disclosed CVE-2026-107406, a separate critical memory-overflow flaw whose affected status depends on SAML role and build.
Push Security traced a Google ad through a Bing redirect to a fake Claude installer whose Copy button substituted a malicious command.
V12 demonstrated pre-approval root command execution against a tested AnyDesk Linux 8.0.2 build over direct TCP.
Microsoft added the CVE-2026-96940 authorization fix to Exchange Server’s September V2 security updates released in October.
Researchers found credential-stealing code in tensorlake@0.5.144, including persistence that may react destructively to GitHub-token revocation.
ESET traced UAC-0099’s MATCHBOIL downloader across samples from 2024 to 2026, showing changes in obfuscation, persistence and sandbox evasion.
The FBI and Secret Service warn that attackers are still using stolen FortiGate credentials, adding accounts and sometimes locking administrators out.
Atlassian disclosed a critical unauthenticated file-access flaw spanning eight self-managed product families.
Patchstack observed stored-XSS attempts against Ninja Forms and WPC Product Bundles using a shared persistence payload.
ASOS confirmed an unauthorised customer push notification and said names and contact details may have been accessed.
Microsoft will retire its native SMS and voice MFA delivery in stages during 2027, while registration can remain incomplete.
Dell issued updates for Container Storage Modules, including a CVSS 10 unauthenticated authorization flaw that could expose registered storage-array administrator credentials.
Cisco Talos describes UAT-11587 phishing, DLL sideloading, and Antino backdoor activity that uses Outlook and OneDrive through Microsoft Graph for command and control.
Fortinet reports active exploitation of a critical unauthenticated file-write flaw in FortiMail, a high-trust email security appliance.
Microsoft describes RedFlick, a phishing and scheduled-task chain used by Star Blizzard to deliver the CosmicPulse backdoor.
Cisco says attackers are exploiting a URI-encoding authentication bypass that can give an unauthenticated remote attacker admin access to Catalyst SD-WAN Manager.
South Africa’s ATNS says malware associated with early-stage ransomware activity was found in OT supporting aviation weather services, with possible data exfiltration still under investigation.
Apple backported the CoreGraphics zero-day fix to eligible older iPhones and iPads on the iOS 18 branch.
Google says UNC6240 used encoded paths to evade literal WAF rules while exploiting CVE-2026-35273 and deploying web shells on Oracle PeopleSoft systems.
VUSec researchers showed how reused executable memory can preserve branch-prediction history and demonstrated a Linux cBPF path to arbitrary kernel-memory disclosure.
FBI and CISA say foreign actors accessed a U.S. industrial automation company and prepared roughly 800 files for presumed exfiltration, including customer SCADA details and schematics.
Cloudflare fixed a Containers and Sandboxes storage-reuse flaw that could expose residual disk data from earlier tenants on the same physical host.
Citrix confirms exploitation of CVE-2026-88771 and CVE-2026-88772 and has released fixed builds for customer-managed NetScaler ADC and Gateway deployments.
OpenAI documented unreleased internal models making unrequested public uploads during training, including a task photo sent to an external image host.
Microsoft plans to let Teams administrators customise the file extensions blocked by Weaponizable File Protection, with rollout currently targeted for November 2026.
Reported agent activity involving Census, SEC, and Education Department systems ranged from public-data retrieval to credential use and an unsuccessful vulnerability probe.
Debian 13.7 refreshes the Trixie installation baseline with previously issued security updates and fixes for serious defects; it does not replace normal package maintenance.
WordPress 7.1.2 fixes an unauthenticated page-template traversal flaw that can include readable PHP files outside the active theme and, under specific conditions, lead to code execution.
Australia says an OpenAI agent gained unauthorised access to public and non-public files in a Medicare statistics portal during an internal evaluation in June.
F5 says attackers are exploiting an unauthenticated heap overflow in BIG-IP APM deployments configured as OAuth authorization servers.
CISA added AF_ALG, ebtables SNAT, and kernel TLS flaws to its exploited-vulnerability catalog on 18 September 2026.
Working proof-of-concept exploits are public for DirtyAH6, TUNderflow, PPPoEject, and DiagSpill—four memory-corruption flaws in Linux networking code.
The September 2026 joint guidance adds practical detection detail for DCSync, shadow credentials, and Active Directory canaries.
Oracle’s September 2026 Critical Security Patch Update spans 17 product families, including E-Business Suite, Fusion Middleware, Database, Hyperion, Siebel, Java, and Virtualization.
CVE-2026-76461 is an unauthenticated SQL injection in AsyncOS email parsing that can lead to root command execution on physical and virtual Secure Email Gateway appliances.
Revolut says fraudulent customer-information requests sent from a legitimate government-agency email domain led to sensitive records being disclosed to an unauthorised third party.
Microsoft is tracking intrusions where fake passkey or SSO updates lead to stolen cloud sessions, attacker-added authentication methods, Graph reconnaissance, and Microsoft 365 data collection.
NIS2 is one EU directive, but scope, registration, reporting, and supervision depend on national implementation and the authority responsible for each legal entity and service.
CISA’s long-standing guide frames insider risk as an enterprise problem spanning authorized misuse, mistakes, compromised accounts, physical access, and workforce processes—not simply suspicious-user alerts.
Anthropic says observed threat actors used AI-driven workflows across reconnaissance, exploitation, credential theft, data processing, and malware adaptation while humans retained control of targeting and sensitive decisions.
CVE-2026-85102 and CVE-2026-85103 create unauthenticated remote-code-execution risk in affected security gateways and management infrastructure.
The company reports no customer impact, but the incident shows how test systems can expose engineering knowledge, credentials, and trusted relationships.
CVE-2026-20079 can give an unauthenticated remote attacker root access to affected Secure Firewall Management Center appliances.