LLMKube versions prior to 0.9.0 contain a vulnerability in the ModelReconciler component where the Model.spec.source field accepts arbitrary absolute paths and file:// URIs without path allowlisting. This allows users with create/update permissions on Model and InferenceService custom resources to read arbitrary files accessible to the controller pod or on the node hosting the inference pod, including sensitive files like Kubernetes service-account tokens and kubelet configuration. Model caching, which triggers the vulnerable hostPath volume mount, is enabled by default. The vulnerability has a CVSS 3.1 base score of 7.7 (High).
A low-severity vulnerability exists in the MySQL Cluster product of Oracle MySQL, specifically in the Cluster JS module. The flaw affects multiple supported versions including 8.0.0 through 8.0.50, 8.4.0 through 8.4.13, 9.7.0 through 9.7.4, and version 26.10. This vulnerability allows a low-privileged attacker with access to the infrastructure hosting MySQL Cluster to gain unauthorized read access to some data. The CVSS 3.1 base score is 2.5, indicating limited confidentiality impact without integrity or availability effects. Oracle has released patches for this vulnerability as part of their November 2026 Critical Security Patch Update.
Between September 12 and 18, 2026, a North Korea-linked threat actor conducted spearphishing attacks targeting South Korean seminar attendees. The attacker impersonated a South Korean National Assembly aide and used honorarium payment forms as lures. The attack chain was fileless and multistage, involving malicious LNK files, obfuscated PowerShell scripts, and abuse of public note services (aNotepad) as command and control channels. Payloads included XenoRAT, information collection scripts, document exfiltration tools, and NodeLinkRAT, a JavaScript-based remote access tool running on Node.js. Data exfiltration used rotating channels such as Dropbox, MEGA cloud, and free hosting services. The campaign employed living-off-the-land techniques, polymorphic scripts, and legitimate infrastructure abuse to evade detection while conducting cyberespionage operations targeting South Korea.
Control Web Panel (CWP) before webpanel_version 1.30.6 contains an authorization bypass vulnerability in the validate() function of api/v1/index.php that allows attackers holding a password-recovery token and a predictable legacy signature to obtain valid remote-API authorization without a proper API key. Attackers can exploit this flaw via the POST /cronjobsusers/ endpoint to add arbitrary cron commands and achieve command execution as the affected hosting account, and can also abuse the same token to access global account-list and account-detail routes exposing usernames, contact emails, and password hashes.
Control Web Panel (CWP) before webpanel_version 1.30.6 contains a host header injection vulnerability that allows unauthenticated attackers to manipulate the password-recovery reset URL by injecting the HTTP Host header into the /v1/lostpass/ route's token generation process. Attackers can lure an automated mail-security scanner into visiting the poisoned reset link to exfiltrate the live recovery token, replay it at /login/index.php?acc=newpass to set a new password, and subsequently abuse the File Manager and Crontab modules to achieve command execution as the targeted hosting account.
Multiple threat actors have been using GSocket, an open source remote access tool, to maintain persistent access on compromised Linux servers, especially targeting web applications and hosting environments. The activity surged by over 350% between March and August 2026, with education and research sectors notably affected. GSocket uses a shared public relay network for command and control, bypassing traditional infrastructure that defenders can block. Attackers deploy GSocket with renamed processes in memory-backed directories and establish persistence via cron and systemd. Some attacks involve privilege escalation exploiting CVE-2026-43284. The tool's outbound-only connections and shared relay infrastructure complicate detection and remediation.
GSocket is an open source remote access tool weaponized by threat actors to maintain persistent access on compromised Linux servers, especially web applications and hosting environments. It uses a public relay network for command and control, avoiding direct threat actor infrastructure and evading traditional inbound firewall rules. This allows attackers to maintain access even after patching the initial vulnerability. GSocket employs outbound-only encrypted connections, process renaming, memory-backed staging, and persistence mechanisms that survive reboots, complicating detection and remediation. Microsoft telemetry shows a significant increase in GSocket-related detections, primarily targeting PHP web applications and Apache servers in education and technology sectors. Defenders should focus on detecting GSocket deployment, persistence, privilege escalation, and outbound connections rather than relying solely on patching the exploited application.
Following the arrests of key ShinyHunters members in September 2026, multiple rival factions emerged claiming the ShinyHunters name. These factions publicly denounce each other on Telegram but have been observed directing victims to the same download server hosting stolen data. This situation complicates attribution and victim response, as victims face conflicting claims over the same stolen data. The investigation highlights that Telegram handles and posted signatures are unreliable for attribution, while infrastructure and account IDs provide stronger evidence. The arrests did not end ShinyHunters activity but fragmented it into competing groups.
This report summarizes multiple cyber threats and vulnerabilities discovered in early November 2026, including ransomware attacks, supply-chain compromises, AI-powered exploitation frameworks, and several high-severity software vulnerabilities. Notable incidents include a ransomware attack on Queens Public Library, a destructive attack on Russian web hosting provider Beget, and a supply-chain attack on the Linux security tool cPGuard. AI-assisted threat actors targeted financial and cryptocurrency sectors, while multiple critical vulnerabilities were patched in SonicWall, Android, Barracuda Email Security Gateway, and OpenSSH. Some vulnerabilities have been linked to active exploitation.
A time-of-check time-of-use (TOCTOU) vulnerability exists in olivier-ls PHP-FTS versions up to 1.1.3 affecting the TrigramIndex::open function in Storage.php. The vulnerability requires local access with write permissions to the index directory by a user other than the application user, a scenario common in shared hosting environments. Exploitation could lead to limited confidentiality, integrity, and availability impacts, with potential code execution only if the attacker can create a symlink to a web-served PHP file. Upgrading to version 1.1.4 addresses this issue.
The Static Snapshot WordPress plugin up to version 0.2.1 contains a vulnerability where it fails to sanitize shell metacharacters in user-supplied input before using it in operating system commands. This flaw allows unauthenticated attackers to execute arbitrary commands on the server hosting the plugin.
A URL hosting an attacker payload related to CVE-2017-9841 was observed in the wild by OffSeq Mirage honeypot sensors on 2026-11-05. The URL was seen three times and is associated with the T1105 technique (Ingress Tool Transfer). The attacker infrastructure was identified in the United Arab Emirates (AE). This URL is used for tool download, callback, or exploitation targeting. Defenders are advised to block and monitor this URL and search their proxy, web, and WAF logs for any requests to it.
The banks library versions prior to 2.4.3 contain a remote code execution vulnerability due to unsafe use of importlib.import_module and getattr on attacker-controlled import_path values in Tool JSON objects parsed from template bodies. This allows an attacker who can influence prompt content to execute arbitrary Python code in the hosting process. The vulnerability is distinct from a prior sandboxing fix and remains exploitable in version 2.4.2. A fix is available in version 2.4.3.
A phishing campaign uses over 300 fake domains impersonating mostly US financial institutions to target business banking and treasury logins. The phishing sites serve legitimate-looking login pages to first-time visitors but show fake 'account suspended' pages to returning visitors to evade takedown efforts. The campaign leverages Cloudflare to hide hosting details and uses IP tracking to differentiate visitors. This tactic complicates detection and takedown since security teams see only the fake suspended page on repeat visits. The campaign also targets victims in the UK, Australia, and New Zealand.
Between late September and early October 2026, seven South Korean financial institutions were breached, impacting around 66,000 individuals and 2,200 corporations. The attacks were linked to ARTEX, an AI-driven autonomous penetration testing framework. Investigations uncovered exposed infrastructure hosting AI model-routing components, exploitation tools, and proxy management utilities. The operation leveraged multiple AI models and included various attack artifacts such as reconnaissance tools, exploitation chains, privilege escalation exploits, and scripts targeting cryptocurrency exchanges and mobile devices. The infrastructure spanned multiple countries with over 1,300 IP addresses identified.
Metalama's Visual Studio integration uses a local named pipe for inter-process communication that deserializes messages without restricting CLR types, enabling local code execution. This unsafe deserialization affects all Metalama releases from 2023.0 through 2026.1.18 and is fixed in 2026.1.19. The pipe is accessible only by processes running as the same user, limiting exploitation to local same-user attackers. The vulnerability allows arbitrary code execution within the Visual Studio processes hosting Metalama components, impacting confidentiality, integrity, and availability of those processes. The practical severity is medium due to the same-user restriction and lack of privilege escalation. Remediation requires upgrading to Metalama 2026.1.19 or later, or updating affected Visual Studio extensions accordingly.
The Yoba malware campaign distributes cryptocurrency mining malware by abusing digitally signed AI application executables from OpenAI and Anthropic to evade detection. It uses DLL sideloading with malicious DLLs loaded alongside legitimate signed binaries. The infection chain includes YobaDropper and YobaLoader components that employ anti-analysis techniques such as process verification, execution delays, and regional language exclusions. Command and control communication uses Telegram profile images, domain generation algorithms, DNS over HTTPS via Google, and public hosting services to retrieve encrypted payloads. While primarily observed deploying the XMRig miner, the loader architecture can support additional Windows executables and PowerShell payloads, indicating potential expansion beyond cryptojacking.
An unauthenticated arbitrary file upload vulnerability exists in Joomla Extension Custom Tables versions prior to 3.7.8. Attackers can upload files with attacker-controlled extensions to a web-accessible temporary directory via a public submission form. The uploaded files are served as static content, enabling hosting of malicious content such as phishing or malware pages under the victim's domain.