By severity

High-severity vulnerabilities

CVEs rated High by CVSS, with SEC.co remediation and prioritization guidance.

4140 published vulnerabilities · page 24 of 42

• CVE-2026-54672HIGH 7.8

  electron-updater, a component used to deploy automatic updates in Electron-based applications, contains a privilege escalation vulnerability affecting versions before 26.15.0. When AppImage packages are built using app-builder-lib, the dynamic library loader path (LD_LIBRARY_PATH) is set with an empty component that causes the current working directory to be searched for shared libraries. An attacker with local access can exploit this by placing a malicious library in the directory where an AppImage is executed, leading to arbitrary code execution with the privileges of the user running the application.

• CVE-2026-55693HIGH 7.8

  Vim, the widely-used command-line text editor, contains a vulnerability in how it processes spell-check files. When a user loads a specially crafted spell file (with .spl or .sug extension) to enable spell suggestions, the editor's word-counting function descends through a data structure without properly validating its depth. This causes the function to write data beyond the boundaries of fixed-size memory buffers, corrupting Vim's execution state and crashing the application. The vulnerability requires user interaction—specifically, loading a malicious spell file—but no special privileges are needed, and the crafted files are straightforward to create.

• CVE-2026-55878HIGH 7.8

  Symfony UX, a JavaScript framework for Symfony developers, contains a path traversal vulnerability in its ux:install command. When installing recipe kits, the command copies files based on a list provided by the kit. An attacker who creates a malicious kit or compromises an existing one can exploit insufficient path validation to write files anywhere on the system or read sensitive files outside the intended recipe directory. The vulnerability affects versions 2.32.0 through 2.36.0 and 3.0.0 through 3.1.x, and is resolved in versions 2.36.1 and 3.2.0.

• CVE-2026-55895HIGH 7.8

  Vim's file browser (netrw plugin) contains a code injection flaw that allows arbitrary Vimscript execution when deleting files with specially crafted names. An attacker who controls a filename in a directory you browse—such as through a shared folder or downloaded archive—can embed a pipe character (|) to break out of the intended delete command and run malicious code, including shell commands. The vulnerability requires user interaction (opening a directory and triggering file deletion) but no special privileges. Vim versions before 9.2.0663 are affected.

• CVE-2026-56000HIGH 7.8

  A local privilege escalation vulnerability exists in X.org's X server and Xwayland that allows an attacker with an existing X connection to crash the display server or potentially execute code with elevated privileges. The flaw stems from how the graphics subsystem manages memory when processing GLX (OpenGL X extension) commit operations, leaving the server vulnerable to use-after-free attacks. An attacker needs local system access and an active X session to exploit this, but no special user interaction is required once those conditions are met.

• CVE-2026-56137HIGH 7.8

  RPG MAKER MV and MZ game development tools contain a vulnerability that allows attackers to run unauthorized system commands on a developer's machine if they open a specially crafted save file. This is a local attack that requires user interaction—an attacker would need to trick a developer into loading a malicious save file—but once triggered, it grants complete system-level access to the attacker. The vulnerability affects both major versions of the engine and poses significant risk to game developers and anyone using these tools.

• CVE-2026-56437HIGH 7.8

  Pupsman installers prior to version 3.9.0 are vulnerable to arbitrary code execution if an attacker places a malicious DLL file in the same directory as the installer. When a user runs the installer, the vulnerable search path mechanism loads the attacker's DLL instead of the legitimate one, executing malicious code with full system privileges. This is a local attack that requires user interaction—specifically, running the installer—but poses severe risk in shared environments or scenarios where users download installers to predictable locations.

• CVE-2026-57237HIGH 7.8

  A vulnerability in PDF applications allows an attacker to crash the software by crafting a malicious PDF file containing JavaScript that modifies form field properties. When the application processes these modifications, internal object references become corrupted, causing the program to access invalid memory and crash. This affects Foxit PDF Editor and Reader, as well as the underlying operating systems (Windows and macOS) that host these applications.

• CVE-2026-57238HIGH 7.8

  A vulnerability in Foxit PDF Editor and PDF Reader allows a maliciously crafted PDF to crash the application through JavaScript code that deletes form field objects and then attempts to access them. This type of use-after-free flaw can be triggered when a user simply opens a malicious PDF file, requiring no special configuration or user interaction beyond opening the document. The vulnerability affects both products on Windows systems and carries a CVSS score of 7.8 (HIGH severity).

• CVE-2026-57240HIGH 7.8

  A vulnerability in Foxit PDF Editor and PDF Reader allows an attacker to crash the application by crafting a malicious PDF file that uses JavaScript to delete form fields while the application is processing it. The application continues to reference the deleted fields using stale memory pointers, leading to a crash. This requires user interaction—someone must open the malicious PDF—but no special privileges are needed. While the immediate impact is denial of service through application crash, the underlying memory safety issue could potentially be leveraged for more serious attacks.

• CVE-2026-57242HIGH 7.8

  CVE-2026-57242 is a high-severity vulnerability affecting Foxit PDF Editor, Foxit PDF Reader, and Microsoft Windows. When JavaScript within a PDF modifies form elements, the application fails to properly manage the lifecycle of related objects on the page and does not validate null values. This causes the application to repeatedly attempt to access invalid objects as the page state changes, ultimately crashing the application. An attacker can exploit this by crafting a malicious PDF that, when opened, triggers the crash. This is a local, user-interaction-dependent attack that could be used for denial of service or as part of a larger attack chain.

• CVE-2026-57244HIGH 7.8

  A vulnerability in Foxit PDF Editor and Reader allows an attacker to crash the application by exploiting a flaw in how the software handles form resets triggered by JavaScript. When a form is reset, the application fails to properly verify that internal objects still exist before attempting to use them, leading to a memory access error that crashes the program. An attacker can exploit this by crafting a malicious PDF file that, when opened by a user, triggers this crash condition.

• CVE-2026-57245HIGH 7.8

  A flaw in Foxit PDF Editor and Foxit PDF Reader allows malicious PDF files to crash the application. When these applications process a PDF with specially crafted hyperlink annotation relationships, they fail to properly validate the annotation structure. This causes internal objects to become corrupted, and when the application closes or cleans up resources, an invalid memory write occurs, crashing the program. An attacker can exploit this by crafting a malicious PDF and delivering it to a user.

• CVE-2026-57246HIGH 7.8

  CVE-2026-57246 is a local privilege escalation and data integrity vulnerability affecting Foxit PDF Editor, Foxit PDF Reader, and Microsoft Windows. When these applications encounter malformed PDF objects, they fail to validate input parameters before processing. JavaScript in the PDF triggers a signature verification routine, but the underlying signature plugin skips validation while copying abnormal strings, leading to a crash and potential exploitation. The vulnerability requires local access and user interaction to trigger.

• CVE-2026-57247HIGH 7.8

  A use-after-free vulnerability in Foxit PDF Editor and PDF Reader allows local attackers to crash the application and potentially execute code by manipulating document field processing. When a user opens a specially crafted PDF, the application deletes the current page while still holding references to field objects from that page. Continued use of those stale references triggers an illegal memory read, leading to a crash. The vulnerability requires user interaction—opening a malicious PDF file—but no special privileges.

• CVE-2026-57248HIGH 7.8

  A vulnerability in Foxit PDF Editor and PDF Reader allows malicious JavaScript embedded in PDF files to crash the application by corrupting annotation data structures. When a PDF is opened, specially crafted JavaScript can write invalid annotation attributes that bypass type and argument validation, causing the application to fail during subsequent operations. This is a local attack that requires user interaction—someone must open a malicious PDF—but once triggered, results in denial of service and potentially exposes sensitive data through memory corruption.

• CVE-2026-57249HIGH 7.8

  A vulnerability in Foxit PDF Editor and PDF Reader allows an attacker to crash the application by crafting a malicious PDF file that exploits how the software handles form reset operations. When the PDF is opened, the application resets annotation status and then processes a form reset event through additional actions, which can cause it to access memory that has already been freed. This type of memory corruption can lead to application crashes or potentially allow code execution. The vulnerability requires user interaction—a person must open the malicious PDF file—but no special privileges are needed.

• CVE-2026-57250HIGH 7.8

  A vulnerability in Foxit PDF Editor and PDF Reader allows malicious PDFs to crash the application through a use-after-free condition. When a PDF containing JavaScript resets form fields, the script can re-enter the PDF handler in a way that damages the underlying native object. Because the application fails to validate the object's state before using it, a function call on the corrupted object causes an immediate crash. An attacker can exploit this by crafting a malicious PDF and tricking a user into opening it.

• CVE-2026-57251HIGH 7.8

  CVE-2026-57251 is a memory safety vulnerability in Foxit PDF Editor and Reader that occurs when the application processes PDF files. The issue stems from improper bounds checking during PDF parsing—the code fails to enforce an upper limit on memory access, allowing reads or writes beyond the intended array boundaries. When a malicious or malformed PDF is opened, this out-of-bounds access crashes the application. While the immediate symptom is denial of service, the vulnerability's high CVSS score reflects that memory corruption of this type can potentially be exploited for information disclosure or code execution depending on memory layout and attacker sophistication.

• CVE-2026-57252HIGH 7.8

  A vulnerability in Foxit PDF Editor and Reader allows an attacker to crash the application by manipulating JavaScript within a PDF file to delete pages and remove attachment annotations. This triggers the attachment panel to access invalid memory pointers, resulting in a denial of service. The attack requires user interaction—the victim must open a malicious PDF—but no special privileges are needed, making it a practical attack vector for social engineering campaigns.

• CVE-2026-57254HIGH 7.8

  A flaw in PDF parsing allows a specially crafted PDF file to crash applications that read or edit it. The vulnerability exists because these applications fail to properly validate data types when encountering abnormal annotations within PDF documents. An attacker can exploit this by distributing a malicious PDF that causes the application to crash when opened, resulting in a denial of service. The vulnerability requires user interaction—someone must open the PDF file—but does not require special privileges to trigger.

• CVE-2026-57256HIGH 7.8

  A flaw in PDF applications allows attackers to crash the software by crafting a malicious PDF that exploits how the application handles form fields when JavaScript is involved. When the PDF is opened, the application mismandles internal data structures associated with form objects, leading to a memory access violation that terminates the program. An attacker would need to trick a user into opening the malicious PDF file.

• CVE-2026-57260HIGH 7.8

  CVE-2026-57260 is a memory safety vulnerability in Foxit PDF Editor and Reader that can be triggered when opening a specially crafted PDF file containing a malformed Unity 3D object. The application misinterprets data within the abnormal object as a valid memory pointer and attempts to use it, leading to a crash. An attacker could exploit this by distributing a malicious PDF that causes the application to crash when opened by a user, or potentially achieve more severe outcomes depending on how the pointer dereference is handled at runtime.

• CVE-2026-57455HIGH 7.8

  Vim versions before 9.2.0698 contain a stack buffer overflow in spell-checking functionality. When processing words longer than Vim's internal buffer limit (MAXWLEN) with certain spell file configurations, the editor writes data beyond allocated memory, corrupting the program's call stack and causing a crash. An attacker can trigger this by crafting a malicious spell file or providing oversized words during spell-checking operations, though the immediate impact is denial of service rather than code execution.

• CVE-2026-57456HIGH 7.8

  Vim's Python omni-completion feature (which auto-suggests function and class names while editing) contains a code execution vulnerability. When you trigger completion on a malicious Python file, Vim reconstructs function and class definitions from the buffer and executes them using Python's exec() function. An attacker can craft a docstring that breaks out of Python's triple-quote syntax and runs arbitrary code. This happens automatically during normal editor usage—no manual code execution is required. The vulnerability affects Vim versions before 9.2.0699.

• CVE-2026-57851HIGH 7.8

  MSI Feature Manager's kernel driver contains a dangerous flaw that allows any regular user with local access to gain extremely powerful privileges. An attacker with a standard user account can read and write to system memory, control hardware I/O ports, and manipulate critical kernel protections—effectively giving them near-complete control over the system. This can lead to disabling antivirus software, bypassing Windows security features, and taking over the entire machine.

• CVE-2026-57895HIGH 7.8

  Pupsman versions before 3.9.0 suffer from incorrect default folder permissions that allow a local attacker to drop a malicious executable into the installation directory and execute it with SYSTEM-level privileges. This is a local privilege escalation vulnerability that requires an attacker to already have basic user access to the system.

• CVE-2026-57919HIGH 7.8

  A privilege escalation vulnerability exists in Matrix42 Empirum's backup component (PBackupVSS.exe). The application creates a shared communication channel (named pipe) that allows any user logged into the system to send commands to it. An attacker with a basic user account can exploit this to run commands with the highest system privileges, potentially compromising the entire machine. The attack works by placing a malicious file in a directory that the backup process searches through during normal operation.

• CVE-2026-58459HIGH 7.8

  A vulnerability in gpsd (the open-source GPS daemon) allows attackers to run arbitrary shell commands on systems running gpsprof if an attacker can control the GPS device subtype information. The flaw exists because subtype values from GPS data aren't properly sanitized before being inserted into gnuplot commands. When a user generates a plot using gpsprof and gnuplot, the malicious subtype text (containing backticks or shell metacharacters) gets executed as shell commands by the user running gnuplot. This is a local-attack scenario requiring either direct control of a GPS device or manipulation of GPS log data.

• CVE-2026-59856HIGH 7.8

  Vim, a widely-used open-source text editor, contains a code-execution vulnerability in its PHP autocompletion feature. When a user opens a specially crafted PHP file and triggers the omni-completion function, malicious PHP class or trait names embedded in the file can break out of the search command and execute arbitrary operating-system commands on the user's machine. The vulnerability requires user interaction—opening the file and invoking completion—but does not require special privileges. Vim versions prior to 9.2.0736 are affected.

• CVE-2026-59858HIGH 7.8

  Vim's C language code completion feature contains a command injection vulnerability that can execute arbitrary commands when you open a C source file in a project with a malicious tags file. An attacker crafts a specially formatted tags entry designed to break out of Vim's search pattern syntax and inject executable commands. When you trigger code completion, Vim runs those commands with your user privileges. The vulnerability is fixed in Vim 9.2.0735 and later.

• CVE-2026-61437HIGH 7.8

  PraisonAI, a Python framework for building agent-based workflows, contains a code execution vulnerability in versions before 1.6.78. When processing workflow files that reference Pydantic models as strings, the framework automatically loads a sibling tools.py file from the same directory without proper security checks. An attacker who can place or modify both the workflow configuration and tools.py file can inject malicious Python code that executes with full privileges when the workflow runs. This is a local attack vector requiring file system access, but the impact is severe—complete code execution in the context of the application using PraisonAI.

• CVE-2026-6509HIGH 7.8

  Pardus Update, a software maintenance tool from TUBITAK BILGEM, contains a missing authorization check that allows an authenticated local user to escalate their privileges on affected systems. An attacker with basic user access can bypass intended permission controls to perform administrative actions without proper authorization, potentially gaining full system control.

• CVE-2026-6676HIGH 7.8

  Avira Antivirus contains a memory corruption flaw in its scan engine that triggers when processing specially crafted tar archive files. An attacker who can get a user to scan a malicious tar file can exploit this to either run arbitrary code on the system with the victim's privileges or crash the antivirus engine, disabling active protection. The vulnerability affects Windows, macOS, and Linux installations running engine versions before 8.3.27.12.

• CVE-2026-8357HIGH 7.8

  LibreOffice Calc has a critical flaw in how it processes spreadsheet formulas when opening files. The vulnerability occurs because the application allocates a buffer one element too small to handle extremely long, deeply nested formulas composed of many opening tokens. When such a formula is compiled, it writes past the buffer's boundary, corrupting adjacent memory. This heap buffer overflow can allow attackers to execute arbitrary code simply by convincing a user to open a malicious spreadsheet file.

• CVE-2026-8501HIGH 7.8

  CVE-2026-8501 is a privilege escalation vulnerability in PC Tools Internet Security's kernel driver (PCTCore64.sys). The driver fails to properly restrict access to its device interface, allowing any user-mode process to issue commands that should be reserved for system-level operations. An attacker with local access can send specially crafted requests to the driver to gain elevated privileges and execute sensitive operations on the compromised system.

• CVE-2026-8637HIGH 7.8

  LanSchool Classic contains a vulnerability where the application searches for files or libraries in locations that an authenticated local user can manipulate. An attacker with local access could place malicious files in predictable locations and trick the application into loading them with elevated privileges, leading to complete system compromise. This is a local-only issue requiring prior system access, but the damage potential is severe.

• CVE-2026-8795HIGH 7.8

  Rapid7 Velociraptor contains a vulnerability where specially crafted evidence collections can inject malicious code into YAML configuration files. When a security analyst processes a compromised collection using Velociraptor's remapping feature, arbitrary commands execute on their workstation with full permissions. The attack requires an attacker to control a collection ZIP file (obtained through network compromise or social engineering) and a user to run a specific analysis command. This is a local privilege escalation risk for incident response teams.

• CVE-2026-8863HIGH 7.8

  CVE-2026-8863 is a Secure Boot bypass vulnerability affecting Microsoft-signed UEFI SHIM bootloaders. An attacker with administrative access or the ability to manipulate the boot process can exploit these flawed bootloaders to execute malicious code before the operating system starts, effectively bypassing Secure Boot protections. Remediation requires a specific UEFI DBX (Deny List) update to block the vulnerable bootloaders from loading.

• CVE-2026-9045HIGH 7.8

  Lenovo Accessories and Display Manager for Enterprise for Windows contains a privilege escalation vulnerability that allows a local authenticated user to execute arbitrary code with elevated privileges. An attacker with valid credentials on a system running the affected software could bypass normal access controls and run malicious code with system-level permissions, potentially leading to complete system compromise.

• CVE-2026-9987HIGH 7.8

  A flaw in Google Chrome for Android allows a malicious file to trick the app's installation handler into running unauthorized code on your device. An attacker would need local access (physical proximity or prior compromise) and some user interaction to pull this off. The vulnerability stems from insufficient validation when Chrome processes installation-related files or requests.

• CVE-2023-37524HIGH 7.7

  HCL Traveler for Microsoft Outlook relies on .NET Framework 4.5, which reached end-of-life years ago and no longer receives security patches. This means the application sits on top of a foundation with known, unfixed vulnerabilities that attackers can exploit. When a third-party component falls out of support, you're essentially running with one hand tied behind your back—security gaps accumulate and become actionable by threat actors.

• CVE-2025-60223HIGH 7.7

  WPBot Pro, a WordPress chatbot plugin, contains a vulnerability that allows subscribers (low-privilege users) to delete arbitrary files from a WordPress installation. This affects versions 13.6.5 and earlier. The flaw does not require user interaction and can be exploited remotely by any authenticated subscriber, making it a direct threat to site integrity and availability.

• CVE-2026-10107HIGH 7.7

  MoviePilot v2 has a security flaw in its image proxy feature that lets authenticated users access internal network resources they shouldn't be able to reach. An attacker with valid login credentials can trick the proxy into fetching files or services from private networks—like personal media servers—by crafting requests with specific cookies and domain names. The vulnerability exists because the security check only verifies that a domain name appears to be allowed, but doesn't block access to internal, private, or loopback addresses. This means an attacker could map out internal services or steal data from them.

• CVE-2026-10835HIGH 7.7

  The SALESmanago & Leadoo WordPress plugin contains a SQL injection vulnerability that allows attackers with minimal user privileges—even subscriber-level accounts—to extract sensitive data from the WordPress database. The flaw stems from insufficient input validation in an AJAX endpoint that fails to properly sanitize user-supplied parameters before inserting them into database queries. This affects versions prior to 3.11.3.

• CVE-2026-11297HIGH 7.7

  A vulnerability in Google Chrome's Reader Mode on Android allows local attackers to bypass navigation restrictions by providing a specially crafted file. While the underlying input validation flaw is classified as low severity by Google's Chromium team, the CVSS scoring reflects the potential for high-impact integrity and availability consequences when successfully exploited. This is a local-only attack, meaning an attacker must already have some level of access to the device to deliver the malicious file.

• CVE-2026-14373HIGH 7.7

  HashiCorp Nomad has a flaw in how it enforces security restrictions on the Docker task driver. Specifically, Nomad failed to properly block authenticated users from launching containers that can access the host operating system's namespace and resources. An attacker with the ability to submit jobs to Nomad (a legitimate but restricted privilege) could exploit this to break container isolation, potentially reading sensitive data from the host or other workloads running on the same physical machine. The vulnerability affects both open-source and enterprise versions and has been patched.

• CVE-2026-14468HIGH 7.7

  HashiCorp Terraform Enterprise has a path traversal vulnerability in how it processes modules from version control systems. An authenticated user can craft a malicious module that escapes the intended repository boundaries and includes files from elsewhere on the server—such as configuration files or credentials—then download those files. The vulnerability affects Terraform Enterprise versions prior to v2.0.4 and v1.2.4.

• CVE-2026-21734HIGH 7.7

  A vulnerability in the GPU shader compiler library can be triggered when a malformed web page containing specially crafted GPU shader code is loaded and processed. The compiler crashes due to an out-of-bounds memory write when handling an edge case involving a very small value in the shader code. On systems where the compiler runs with elevated privileges, this crash could potentially be chained with additional exploits to compromise the device. The vulnerability does not grant direct access to sensitive data, but the ability to crash a privileged process with controlled memory corruption creates a meaningful attack surface.

• CVE-2026-32174HIGH 7.7

  A flaw in Azure Bot Service's authentication system allows someone who already has legitimate access to the service to bypass normal permission controls and gain elevated privileges. The vulnerability is network-accessible, meaning an attacker doesn't need local system access, but they do need valid credentials to exploit it. The impact is primarily on the integrity of the system—an attacker could modify configurations, data, or access controls—rather than stealing data or causing downtime.

• CVE-2026-33235HIGH 7.7

  AutoGPT versions before 0.6.52 contain a denial-of-service vulnerability in the Fill Text Template block. An authenticated user can craft computationally expensive expressions that lock up or crash the server, affecting all other users. The vulnerability stems from the platform's failure to enforce CPU and memory limits on template processing, even though it does block dangerous attribute access. This is particularly damaging in shared or self-hosted deployments where one user's action can take down the entire service.

• CVE-2026-33655HIGH 7.7

  New API is an AI gateway and asset management platform. Versions prior to 0.12.0-alpha.1 suffer from inadequate validation of internal IP addresses in notification webhook configurations. An authenticated attacker can bypass hostname filtering to point notification endpoints (Webhook, Bark, or Gotify) at internal servers or cloud metadata services, potentially exfiltrating sensitive configuration and credentials. The root cause is that domain allow/block rules were applied, but resolved IP addresses were never validated against an internal IP blocklist—a setting (ApplyIPFilterForDomain) that was disabled by default. The vulnerability requires authentication and has limited scope within the notification system, but access to metadata services represents elevated risk in cloud environments.

• CVE-2026-34044HIGH 7.7

  Coolify, an open-source platform for managing servers, applications, and databases, contains an authorization flaw that allows authenticated users to view application logs belonging to other teams. An attacker with valid credentials can supply a UUID of a victim's application and retrieve its logs without owning or having permission to access that application. The vulnerability affects versions prior to 4.0.0-beta.466 and has been patched in that release.

• CVE-2026-34192HIGH 7.7

  A use-after-free vulnerability exists in GPU memory management that can be triggered by unprivileged users through improper GPU system calls. When the GPU driver encounters an error during certain operations, it fails to properly clean up allocated memory before freeing it, leaving page table structures accessible even after deallocation. An attacker can exploit this to read or corrupt GPU memory, potentially affecting system integrity and data confidentiality.

• CVE-2026-34592HIGH 7.7

  Coolify, an open-source platform for managing servers, applications, and databases, contains an authorization flaw that allows any logged-in user to view and access infrastructure belonging to other teams. An attacker with valid credentials can retrieve servers and projects owned by different teams simply by guessing or obtaining their IDs, bypassing the team isolation that should protect multi-tenant deployments. This was fixed in version 4.0.0-beta.471.

• CVE-2026-37149HIGH 7.7

  A SQL injection vulnerability has been discovered in GROCERY-STORE-MANAGEMENT-SYSTEM-USING-PHP-AND-MYSQL-PHPMYADMIN v1.0, specifically in the product search functionality. An attacker can craft malicious input in the scost parameter to execute arbitrary SQL commands against the underlying MySQL database, potentially exposing customer data, inventory records, pricing information, and other sensitive business information. The vulnerability requires local access to the affected system and does not require authentication.

• CVE-2026-4035HIGH 7.7

  MLflow, a popular open-source machine learning platform, contains a credential exposure vulnerability affecting versions before 3.11.0. The flaw allows attackers to extract sensitive server-side environment variables—such as AWS credentials—by manipulating how the AI Gateway handles secrets. An attacker with basic authentication access (or no authentication in default setups) can craft requests that trick MLflow into exposing these credentials to attacker-controlled endpoints. This is particularly dangerous because exposed cloud credentials could allow further compromise of artifact repositories and downstream systems.

• CVE-2026-41156HIGH 7.7

  CVE-2026-41156 is a use-after-free vulnerability in GPU driver firmware interactions that allows a non-privileged user to trigger improper GPU system calls, leading to resource mismanagement and potential system instability or data corruption. The flaw occurs when a CPU driver thread deallocates shared memory before the GPU firmware has finished accessing it, creating a race condition that can be exploited locally without requiring elevated privileges.

• CVE-2026-42129HIGH 7.7

  A vulnerability in Grafana's Loki data source plugin allows users with basic read-only (Viewer) permissions to bypass intended access controls and reach administrative endpoints they should not access. By crafting malicious file paths, an authenticated attacker can retrieve sensitive backend configuration details and internal service information that would normally be restricted to administrators. This is a path traversal flaw—a classic technique for escaping intended directory boundaries.

• CVE-2026-42398HIGH 7.7

  CVE-2026-42398 is a Server-Side Request Forgery (SSRF) vulnerability in Kibana that allows authenticated users with connector management privileges to circumvent network egress restrictions. An attacker with the right permissions can configure a malicious Webhook connector that tricks Kibana into making outbound HTTP requests to internal or restricted destinations that should have been blocked by the organization's firewall or allowlist policies. This effectively punches through network security controls by leveraging Kibana's own trusted outbound connection capability.

• CVE-2026-42965HIGH 7.7

  OpenShift Router contains a flaw that allows users with EndpointSlice write permissions to redirect traffic through the router to cloud metadata endpoints. By crafting a Service backed by an FQDN-based EndpointSlice pointing to a cloud metadata service, an attacker can intercept and read sensitive instance credentials and metadata that should never be exposed. This circumvents existing IP address validation protections designed to block such access.

• CVE-2026-44285HIGH 7.7

  FastGPT, an AI Agent building platform, contains a Server-Side Request Forgery (SSRF) vulnerability in versions before 4.15.0-beta1. An authenticated user can bypass the platform's internal network protection and send HTTP requests to services on the internal network that should be inaccessible. The flaw exists in the dataset preview feature when using the externalFile import type. This allows an attacker with valid credentials to potentially access sensitive internal services, databases, or administrative endpoints that are normally restricted from external access.

• CVE-2026-44692HIGH 7.7

  Sharp, a Laravel-based content management framework, contains a path traversal vulnerability in its file download functionality. An authenticated user who has legitimate access to view one record in Sharp can exploit a flaw in the download endpoint to retrieve unrelated files from any configured Laravel Storage disk. The vulnerability requires an attacker to be already logged into Sharp, but once authenticated, they can bypass the authorization checks that should restrict downloads to files associated with their authorized records. This affects confidentiality but not integrity or availability.

• CVE-2026-45296HIGH 7.7

  OpenReplay, a self-hosted session replay platform, contains a multi-tenant authorization bypass that allows attackers with valid API credentials for one tenant to access another tenant's sensitive session data. The vulnerability exists because the API does not verify that an API key and requested project belong to the same tenant—it only confirms the API key is valid and the project exists. Since project IDs are exposed in browser-side code, an attacker can discover victim project identifiers and exploit this flaw to enumerate user sessions and extract sensitive event details across tenant boundaries. This is a critical tenant isolation failure that affects confidentiality.

• CVE-2026-45497HIGH 7.7

  Microsoft Copilot contains a command injection vulnerability that allows an authenticated attacker to execute arbitrary code through a network connection. The vulnerability arises from improper handling of special characters in command inputs, enabling an attacker with valid credentials to bypass intended restrictions and run code on affected systems. This is not a pre-authentication flaw, meaning the attacker must already have authorized access to Copilot.

• CVE-2026-45807HIGH 7.7

  Kestra, an open-source workflow orchestration platform, contains a path traversal vulnerability in versions before 1.0.43 and 1.3.19. An authenticated user can craft a specially encoded request using %2E%2E (URL-encoded dots) to bypass the platform's path validation guard. Once the request reaches the file-reading logic, the encoded dots are decoded back to .. and allow navigation outside intended directories. An attacker with valid credentials can read sensitive files on the host system—including /etc/passwd, environment secrets, and other users' execution data—that the Kestra process has permission to access. The vulnerability requires authentication and affects the confidentiality of data stored on the server.

• CVE-2026-46123HIGH 7.7

  A vulnerability in the Linux kernel's Bluetooth virtio backend driver allows a malicious or buggy virtual device to expose uninitialized kernel memory to unprivileged processes. The driver fails to properly validate the length of data reported by the virtual device, permitting reads beyond the intended 1000-byte receive buffer. An attacker with the ability to control a virtio Bluetooth backend—such as a compromised hypervisor or malicious VM—can leak sensitive kernel heap data or trigger denial of service. This is a local attack requiring some form of device emulation control, but it directly compromises memory isolation guarantees in virtualized environments.

• CVE-2026-46518HIGH 7.7

  OpenEMR versions before 8.0.0.1 contain a stored cross-site scripting (XSS) flaw in the prescription multi-print feature that allows a patient portal user to inject malicious code into a clinician's browser session. By manipulating patient demographic fields through the API, an attacker can execute arbitrary JavaScript when a clinician views prescription reports, potentially compromising the clinician's session and enabling unauthorized access to patient records or system actions.

• CVE-2026-46717HIGH 7.7

  Nezha Monitoring, a self-hosted server and website monitoring tool, contains an access control vulnerability in versions 1.4.0 through 2.0.7 that allows unprivileged users to trigger outbound HTTP requests to arbitrary destinations and retrieve the response data. The vulnerability exists in the notification configuration endpoints, which were inadvertently made accessible to standard member-level users instead of being restricted to administrators only. An attacker with member-level access can exploit this to perform server-side request forgery (SSRF) attacks, potentially accessing internal services, extracting sensitive information from cloud metadata endpoints, or conducting reconnaissance against internal infrastructure. The issue has been resolved in version 2.0.8.

• CVE-2026-46821HIGH 7.7

  CVE-2026-46821 is a high-severity flaw in Oracle Financials Common Modules that allows a low-privileged attacker with network access to view sensitive financial data without authorization. The vulnerability exists in Oracle E-Business Suite versions 12.2.3 through 12.2.15 and can be exploited via standard HTTP requests. What makes this particularly concerning is that while the defect lives in the Common Modules component, successful exploitation can grant attackers access to confidential information across multiple interconnected Oracle applications, expanding the blast radius beyond a single product.

• CVE-2026-46823HIGH 7.7

  CVE-2026-46823 is an authorization flaw in Oracle Public Sector Financials (International), a module of Oracle E-Business Suite. An attacker with a low-privileged account and network access can bypass authorization controls to read sensitive financial data that should be restricted. The vulnerability requires valid credentials but no user interaction, making it a straightforward privilege escalation path. Affected versions range from 12.2.6 through 12.2.15.

• CVE-2026-47170HIGH 7.7

  Garlic-Hub is a self-hosted digital signage management platform that controls networked displays, content libraries, and playback schedules. A vulnerability in versions before 1.1 allows any authenticated user to trick the server into making HTTP requests to internal systems on behalf of the attacker. Those responses are then saved to the publicly accessible media pool, effectively giving an insider a window into your internal network topology and service configuration.

• CVE-2026-47179HIGH 7.7

  Arcane, a container management interface for Docker, has a path-traversal vulnerability that allows authenticated users to read arbitrary files from the server. An attacker can craft a malicious Docker Compose configuration file with include directives pointing to sensitive system files—such as /etc/passwd or Arcane's own database—and then retrieve the contents through the project file API. Since Arcane stores password hashes and API keys in its SQLite database, successful exploitation can lead to account takeover and remote code execution on the host system. The issue affects all versions prior to 1.19.4.

• CVE-2026-47260HIGH 7.7

  Koel, a free open-source music streaming platform, contains a Server-Side Request Forgery (SSRF) vulnerability in its podcast feature. While the application correctly validates podcast feed URLs to prevent abuse, it fails to validate the individual audio file URLs embedded within those feeds. An authenticated attacker can craft a malicious podcast feed containing URLs pointing to internal services, and when users play episodes from that feed, the Koel server will fetch and stream the content from those internal URLs, effectively bypassing network segmentation. This allows an attacker to read sensitive data from services that should only be accessible internally.

• CVE-2026-47684HIGH 7.7

  Sync-in Server, an open-source file synchronization and collaboration platform, contains a vulnerability in how it validates and blocks requests to private IP addresses when downloading files from URLs. The system's security check fails to recognize IPv4-mapped IPv6 addresses—a special format that represents traditional IPv4 addresses within IPv6 protocol—allowing attackers to bypass the protection and reach internal resources they shouldn't access. This issue affects systems running dual-stack networking (supporting both IPv4 and IPv6). The vulnerability has been patched in version 2.3.0.

• CVE-2026-47937HIGH 7.7

  Adobe Acrobat Reader contains a flaw in how it searches for and loads libraries or components from the file system. An attacker with administrator or elevated system privileges could craft a malicious file that, when opened by a user, tricks the application into running attacker-controlled code with the same permissions as the user who opened the file. The vulnerability requires both high system privileges to set up and user action to trigger—someone must explicitly open the crafted file.

• CVE-2026-49465HIGH 7.7

  n8n workflow automation platform contains a path traversal vulnerability affecting versions before 1.123.48, 2.21.8, and 2.22.4. An authenticated user with workflow creation or modification permissions can bypass file sandbox restrictions by supplying a local filesystem path in Git node operations. This allows an attacker to read the contents of any git repository accessible to the n8n process, even if those paths are normally blocked by the N8N_RESTRICT_FILE_ACCESS_TO sandbox configuration. The vulnerability requires authentication and workflow modification privileges but could expose sensitive repository data to unauthorized access.

• CVE-2026-49821HIGH 7.7

  Fission, a Kubernetes serverless framework, contained a namespace validation flaw in its Package controller prior to version 1.24.0. The buildermgr controller failed to verify that the environment namespace specified in a Package resource matched its actual namespace, potentially allowing authenticated users to reference functions or configurations from unintended namespaces. This cross-namespace access could expose sensitive data or functions, though it requires valid Kubernetes authentication to exploit. The issue is resolved in version 1.24.0 and later.

• CVE-2026-49822HIGH 7.7

  A vulnerability in Fission, an open-source serverless framework for Kubernetes, allows low-privilege developers to spy on activity in other namespaces. Specifically, a developer with limited access who can create a KubernetesWatchTrigger (KWT) in their own namespace can exploit this to establish unauthorized monitoring of unrelated namespaces. This violates namespace isolation, a core security boundary in Kubernetes. The issue has been patched in version 1.24.0.

• CVE-2026-49823HIGH 7.7

  Fission, an open-source serverless framework for Kubernetes, contains a namespace isolation bypass in versions prior to 1.24.0. When defining Fission Functions, administrators can reference Secrets, ConfigMaps, and Packages. The admission webhook—a Kubernetes security control—validated namespace boundaries for Secrets and ConfigMaps but failed to validate the namespace referenced in PackageRef. This gap allows an authenticated user in one namespace to access packages from other namespaces, violating expected isolation. Version 1.24.0 patches this validation gap.

• CVE-2026-49957HIGH 7.7

  Hermes WebUI versions before 0.51.296 contain a flaw in how they validate workspace boundaries for remote terminal sessions. An attacker who has legitimate access to the system can trick the application into accepting a system directory like /etc as a valid workspace root, bypassing security checks designed to prevent this. Once bypassed, the attacker can read sensitive files from the system through the workspace file-read features, effectively escaping the intended sandbox that should contain their access to user-specific directories.

• CVE-2026-49984HIGH 7.7

  Kestra, an open-source workflow orchestration platform, contains a path-traversal vulnerability that allows authenticated users to read files anywhere on the server filesystem. The flaw exists because Kestra checks user-supplied file paths for directory traversal attacks using forward slashes, but an attacker can bypass this check by using Windows-style backslashes instead. After the validation passes, the backslashes are converted to forward slashes, enabling the traversal. Any user with the ability to view a workflow execution—the lowest privilege level in Kestra—can exploit this to escape the storage sandbox and read sensitive files across all tenants and namespaces, including databases, secrets, and system credentials.

• CVE-2026-50005HIGH 7.7

  Brickcom IP cameras are shipped with hardcoded default credentials that cannot be easily changed, allowing anyone with network access to view live camera feeds without authentication. An attacker does not need to exploit a software flaw—they simply use the well-known default username and password to log in remotely. This is a configuration and design issue, not a traditional code vulnerability, but the impact is severe: loss of surveillance confidentiality and potential reconnaissance for physical security breaches.

• CVE-2026-50245HIGH 7.7

  Brickcom IP cameras contain a vulnerability that allows anyone with local network access to view live camera snapshots without any credentials. The flaw exists in the /ONVIF endpoint, which normally handles standard camera communication protocols but fails to enforce authentication before serving image data. An attacker on the same network segment could passively gather surveillance footage without triggering alarms or leaving obvious traces.

• CVE-2026-50567HIGH 7.7

  Fission, an open-source serverless framework for Kubernetes, contains a path traversal vulnerability in its archive extraction logic. When Fission fetches and extracts archives (zip files) from URLs specified in Package or Deployment configurations, it fails to validate that extracted files remain within the intended destination directory. An attacker who controls the archive URL can craft a malicious zip file with entries using relative path sequences like `../../` to write files outside the target directory. This allows unauthorized file writes to sensitive locations including other tenants' package directories, mounted secrets, configuration volumes, or even the fetcher process's own binary. The vulnerability exists in versions prior to 1.25.0 and affects the fission-fetcher sidecar that runs in each environment pod.

• CVE-2026-52906HIGH 7.7

  A bug in the Linux kernel's 9P filesystem implementation causes access control flags to accumulate incorrectly when users specify mount options. When administrators mount a 9P share and explicitly request a particular access mode (like 'access=user'), the kernel fails to replace the default access setting. Instead, both the default and requested modes are set simultaneously. This confusion causes the filesystem to reject all legitimate access attempts and fall back to using a default 'nobody' user ID (65534) for file operations. As a result, even the root user cannot perform privileged operations like changing file ownership on the mounted filesystem.

• CVE-2026-53812HIGH 7.7

  OpenClaw versions before 2026.5.18 contain a server-side request forgery (SSRF) vulnerability that allows authenticated users to access private-network resources they should not be able to reach. An attacker with legitimate credentials can manipulate the browser control functionality to bypass security checks that normally prevent navigation to internal networks, then extract sensitive data from those restricted pages. The vulnerability requires existing user authentication but can affect the confidentiality of internal systems.

• CVE-2026-53832HIGH 7.7

  OpenClaw versions prior to 2026.5.18 fail to properly validate identity headers used by trusted proxies. A local attacker who can reach the Gateway port can forge these headers to impersonate an operator, potentially gaining elevated privileges. The vulnerability requires local network access but no authentication, making it a meaningful risk in shared or containerized environments where multiple applications share network paths.

• CVE-2026-53833HIGH 7.7

  OpenClaw before version 2026.4.29 has a security flaw where users with basic access can bypass authorization controls to change QQBot streaming settings that should only be modifiable by administrators. The vulnerability exists because the system doesn't properly enforce access restrictions on the QQBot streaming command, allowing authenticated attackers to reconfigure critical settings even without explicit permission to do so.

• CVE-2026-54017HIGH 7.7

  Open WebUI, a self-hosted AI platform, contains a path traversal vulnerability in its terminal-server reverse proxy. An authenticated user with terminal access can manipulate the request path using encoded traversal sequences (like `../`) to bypass intended boundaries and reach files or endpoints they shouldn't access on the terminal server or connected internal services. The vulnerability affects versions before 0.9.6 and requires valid user authentication to exploit, but no special privileges beyond terminal access.

• CVE-2026-54018HIGH 7.7

  Open WebUI versions before 0.9.6 contain a server-side request forgery (SSRF) vulnerability that allows authenticated users to access internal network resources despite security restrictions. The flaw exists because the application validates the initial URL provided by a user but fails to validate the final destination after HTTP redirects. An attacker can exploit this by providing a legitimate-looking URL that redirects to internal services like localhost or cloud metadata endpoints, effectively bypassing SSRF protections.

• CVE-2026-54033HIGH 7.7

  LibreChat, a self-hosted ChatGPT alternative that integrates with multiple AI providers, contains a server-side request forgery (SSRF) vulnerability in its custom API endpoint configuration feature. Authenticated users can specify any URL as a custom OpenAI-compatible API endpoint, including internal network addresses that should not be accessible from the internet. The application fails to validate these URLs before making HTTP requests, allowing an attacker with valid credentials to probe or interact with internal services, databases, or APIs that would normally be isolated from external access. This issue was fixed in version 0.8.4-rc1.

• CVE-2026-54304HIGH 7.7

  n8n, an open source workflow automation platform, contains a credential exfiltration vulnerability affecting versions prior to 1.123.55, 2.25.7, and 2.26.1. An authenticated user with workflow creation or modification permissions can misconfigure the SecurityScorecard node to send API credentials to attacker-controlled URLs, bypassing the platform's domain allowlist protections. This requires the attacker to have valid n8n access and knowledge of the target's SecurityScorecard credential configuration, but once exploited, directly exposes sensitive API tokens to external parties.

• CVE-2026-54311HIGH 7.7

  n8n versions before 2.25.7 and 2.26.2 contain a cross-user data leak vulnerability in the Merge node's SQL Query mode. An authenticated user can craft a malicious workflow that modifies the shared sandbox environment, poisoning it for all subsequent SQL queries executed by other users on the same instance. This means workflows belonging to different users or projects can inadvertently expose sensitive data to attackers who have created a polluted sandbox. The vulnerability requires authenticated access and only impacts multi-user deployments where multiple users can create workflows with the Merge node.

• CVE-2026-54313HIGH 7.7

  n8n, an open-source workflow automation platform, contains a flaw in how it validates filter parameters passed to MongoDB operations. A user with permission to edit workflows can inject malicious filter values into the MongoDB node's Find And Replace operation, causing the system to match and overwrite unintended documents. The vulnerability exists in versions prior to 2.24.0 and is resolved in version 2.24.0 and later.

• CVE-2026-54322HIGH 7.7

  Daytona, an infrastructure platform for AI code execution, contains an authorization bypass vulnerability in its role management system. An authenticated attacker who owns any organization can modify or delete roles that belong to a completely different organization. The vulnerability stems from the application checking whether the caller owns the organization in the URL path, but then operating on a role based solely on its ID without confirming that role belongs to the target organization. This allows privilege escalation and lateral damage across organization boundaries. The issue affects Daytona versions prior to 0.185.0.

• CVE-2026-54401HIGH 7.7

  A critical privilege escalation flaw in Ubiquiti UniFi OS allows attackers who already have network access and basic user permissions to abuse the system's ability to make outbound requests (Server-Side Request Forgery) to gain higher-level administrative control. The attack requires an attacker to be inside the network or have valid low-privilege credentials, but once exploited, could grant them full system access.

• CVE-2026-54607HIGH 7.7

  FastGPT versions before 4.15.0-beta4 contain a vulnerability in their HTTP-tool OpenAPI schema importer that allows authenticated team members to read sensitive data from internal services and cloud metadata endpoints. When users import OpenAPI schemas, the system validates only the main URL but then passes the full schema to SwaggerParser.bundle, which resolves remote JSON references ($ref pointers) without the same security restrictions. An attacker with team-member access can craft a malicious schema containing references to internal IP addresses, private cloud metadata services, or other restricted endpoints, and the parser will fetch and return that content inline, effectively bypassing the application's internal-address protections.

• CVE-2026-54699HIGH 7.7

  Warp, a terminal-based development environment designed for WSL (Windows Subsystem for Linux), has a command injection vulnerability affecting versions from March 2024 through early May 2026. When Warp fails to open a URL through its primary WSL mechanism (wslview), it falls back to Windows command processing. An attacker can craft malicious URLs that appear in terminal output; when a developer clicks such a link, arbitrary OS commands execute with the user's privileges. The vulnerability requires user interaction but can cross security boundaries, making it a meaningful risk in shared or compromised terminal environments.

• CVE-2026-55189HIGH 7.7

  RustFS, a distributed storage system, has a critical authorization bypass in its FTP interface. When FTP is enabled, users can read files and check file/bucket existence without proper permission checks—even if their account explicitly denies read access. The HTTP S3 interface is not affected. This vulnerability affects RustFS versions 1.0.0-alpha.1 through 1.0.0-beta.8 and is resolved in 1.0.0-beta.9.