By year

Vulnerabilities disclosed in 2026

CVEs published in 2026 with SEC.co analysis.

8541 published vulnerabilities · page 73 of 86

• CVE-2026-28385MEDIUM 5.0

  Canonical LXD has a weakness in how it handles image imports that allows authenticated users to scan internal networks and reach services they shouldn't be able to access. When someone with image-creation permissions requests LXD to import a container image from a URL, the daemon doesn't properly check where that URL points. This means an attacker could probe internal IP ranges, connect to localhost services, or reach cloud metadata endpoints that should be off-limits. The vulnerability requires authentication and specific permissions, which limits exposure, but it's a meaningful network reconnaissance and lateral-movement risk in shared or cloud-hosted LXD environments.

• CVE-2026-34167MEDIUM 5.0

  Coolify, an open-source platform for managing servers and applications, contains an authorization flaw in its ActivityMonitor component. An authenticated user can view activity logs and command output from other teams by guessing sequential activity IDs. This could expose sensitive data like SSH credentials, database passwords, and infrastructure configuration details. The vulnerability affects all versions before 4.0.0-beta.471 and is fixed in that release.

• CVE-2026-35188MEDIUM 5.0

  A flaw in OpenSSL's handling of TLS OCSP stapling—an optimization that allows servers to provide certificate validity proof directly—can cause a double-free memory error in connecting clients. When a malicious server sends a specially crafted OCSP response, it triggers corruption of the client's heap memory. While OCSP stapling is disabled by default, organizations that have explicitly enabled it face exposure. The vulnerability reliably causes denial of service; remote code execution is theoretically possible but difficult to reliably achieve in practice.

• CVE-2026-40992MEDIUM 5.0

  Spring Boot's automatic email configuration leaves mail server connections vulnerable to man-in-the-middle attacks by not verifying the server's identity by default. An attacker positioned on the network path between an application and its mail server could intercept and manipulate email traffic. However, applications that explicitly enable hostname verification through JavaMail properties are protected from this risk.

• CVE-2026-41977MEDIUM 5.0

  CVE-2026-41977 is a denial-of-service vulnerability in a log service that can be triggered by local attackers with minimal privilege requirements. The flaw stems from an integer overflow condition (CWE-190) and may be exploited through user interaction to disrupt system availability and cause minor data integrity issues. While the attack requires local access and user action, the scope crosses security boundaries, making it relevant to systems where unprivileged local accounts or guest access exists.

• CVE-2026-42862MEDIUM 5.0

  Flowise, a popular drag-and-drop interface for building custom AI language model workflows, contains a security flaw that allows authenticated users to move tools between workspaces without proper authorization. When updating a tool, the application fails to validate who should have permission to change ownership fields like workspaceId. An attacker with legitimate access to one workspace can reassign tools to a different workspace, potentially exposing or stealing AI workflows, data pipelines, or proprietary configurations belonging to another team or customer. This breaks the isolation that multi-workspace Flowise deployments rely on to keep organizations separate.

• CVE-2026-43979MEDIUM 5.0

  Local Deep Research versions before 1.6.0 contain a vulnerability where user-supplied search queries and metadata are inserted directly into HTML without proper escaping before being converted to PDF. An authenticated user can inject HTML tags that trick the server into making unauthorized web requests (SSRF), bypassing existing security controls. The vulnerability requires valid credentials but poses moderate risk due to potential confidentiality impact.

• CVE-2026-44173MEDIUM 5.0

  MariaDB server versions within specific ranges contain a privilege-escalation flaw that allows authenticated users to write files to the server's filesystem without possessing the FILE privilege. The vulnerability exists when SELECT statements direct output to files (using INTO OUTFILE or INTO DUMPFILE) and the FROM clause references only subqueries, bypassing the privilege check. An attacker with database login credentials but no explicit FILE permission can exploit this to write arbitrary content to disk, potentially compromising system integrity or enabling further attacks.

• CVE-2026-44936MEDIUM 5.0

  SUSE Rancher Fleet, a Kubernetes fleet management tool, contains a credential leakage vulnerability in how it handles Helm repository authentication. When administrators configure fleet bundles with Helm repositories but don't explicitly set URL validation rules (helmRepoURLRegex), the system will forward stored authentication credentials to any repository URL specified in a fleet.yaml file. An attacker with push access to a monitored Git repository can exploit this by adding a malicious Helm repository URL to a fleet configuration, causing the system to transmit BasicAuth credentials to that attacker-controlled endpoint. This allows credential theft without modifying legitimate configurations.

• CVE-2026-45407MEDIUM 5.0

  Dokku, a containerized platform-as-a-service tool, inadvertently exposes git credentials to local users when setting up authentication. The vulnerability exists in versions before 0.38.2, where a shell command used to initialize credential storage files doesn't set secure file permissions, leaving git authentication tokens readable by anyone with local system access. This is a local privilege escalation and credential theft risk specific to multi-user Dokku deployments.

• CVE-2026-45502MEDIUM 5.0

  A server-side request forgery (SSRF) vulnerability in Microsoft Exchange Server allows an authenticated attacker to make the server issue requests on their behalf, potentially exposing sensitive network information. An attacker with valid credentials can craft requests that cause the Exchange server to connect to internal resources and leak data back to them. This requires prior authentication but poses a meaningful information disclosure risk within trusted network boundaries.

• CVE-2026-46526MEDIUM 5.0

  Local Deep Research versions before 1.6.10 contain a Server-Side Request Forgery (SSRF) vulnerability caused by inconsistent URL validation logic. The application attempts to block malicious URLs using one parsing method but actually sends requests using a different method, creating a gap that attackers can exploit. An authenticated user can craft a specially formatted URL that passes the security checks but reaches an unintended internal or restricted server when the request is actually sent.

• CVE-2026-46561MEDIUM 5.0

  pyLoad, a popular open-source download manager, contains a vulnerability that allows authenticated users to bypass security controls designed to prevent access to internal networks. Specifically, an attacker can craft a specially designed URL that initially appears safe but redirects to a private IP address after the security check passes. This lets the attacker potentially access or interact with systems on the internal network that should be off-limits. The issue affects versions prior to 0.5.0b3.dev100 and has been resolved in that release.

• CVE-2026-48096MEDIUM 5.0

  OpenFGA, an authorization engine used by developers to control permissions, has a caching flaw that could cause it to reuse incorrect permission decisions. When iterator caching is enabled, two different permission requests can accidentally use the same cached result, potentially allowing or denying access incorrectly. The issue affects versions before 1.16.0 and has been patched.

• CVE-2026-48770MEDIUM 5.0

  Notepad++ versions before 8.9.6.1 contain a local denial-of-service vulnerability where a malicious process running on the same Windows session can crash the application by sending a specially crafted message. The vulnerability stems from improper validation of inter-process communication data, allowing an attacker to trigger a memory access violation without requiring special privileges beyond standard user-level access.

• CVE-2026-48956MEDIUM 5.0

  Joomla! contains an access control weakness that allows authenticated users to view a list of installed modules through the frontend interface. While an attacker needs valid login credentials to exploit this, the improper permission check exposes internal system information that should remain hidden from regular users. This information disclosure could help an attacker plan more targeted attacks against the application.

• CVE-2026-49138MEDIUM 5.0

  Nanobot versions prior to 0.2.1 contain a server-side request forgery (SSRF) vulnerability in its web_fetch tool. An attacker with login credentials can trick the application into making requests to internal or private network systems by providing a URL that redirects to a loopback address (like 127.0.0.1) or private IP range. The vulnerability exploits how the underlying httpx library automatically follows HTTP redirects—the application validates the initial URL but not the final destination after redirects are followed, allowing attackers to reach internal services that should not be accessible from the internet.

• CVE-2026-49433MEDIUM 5.0

  DeepAI's email change endpoint lacks CSRF (Cross-Site Request Forgery) protection, allowing attackers to hijack user accounts. An attacker who tricks a logged-in user into visiting a malicious link can silently change that user's email address, potentially locking the legitimate owner out and enabling account takeover. The vulnerability was patched on May 20, 2026.

• CVE-2026-49958MEDIUM 5.0

  Hermes WebUI versions prior to 0.51.303 contain a race condition vulnerability in its file deletion functionality. An attacker with local access and user privileges can exploit a timing gap between when the application validates a file path and when it actually deletes it. By swapping a directory component with a symbolic link at just the right moment, an attacker can trick the application into deleting files outside the intended workspace—potentially removing important system or application files. This requires precise timing and user interaction but could lead to significant data loss or system disruption.

• CVE-2026-54055MEDIUM 5.0

  Kitty, a popular GPU-accelerated terminal emulator, contains a local privilege escalation flaw in its file transmission protocol affecting versions before 0.47.2. A low-privileged user running a child process in the terminal can write files to arbitrary locations on the system by exploiting a race condition in how Kitty validates and creates files. An attacker would need local access and user interaction, but the window for exploitation is small and doesn't require special privileges to trigger.

• CVE-2026-54786MEDIUM 5.0

  Wasmtime, a WebAssembly runtime, contains a resource leak in its implementation of the WASIp1 file descriptor renumbering function (fd_renumber). When a WebAssembly guest program renumbers a file descriptor, Wasmtime updates its internal descriptor table but fails to clean up the corresponding entry in the host operating system's file descriptor table. This leaves orphaned file descriptors and other resources that accumulate in the host until the entire Wasmtime Store is destroyed. An attacker with the ability to execute WebAssembly code and access file descriptors can deliberately call fd_renumber in a loop to exhaust the host's file descriptor pool and memory, causing a denial of service. The vulnerability requires authenticated access (file descriptor capability) and affects only runtimes that expose both fd_renumber and file access to guest modules.

• CVE-2026-55655MEDIUM 5.0

  OpenSSH on Linux has a vulnerability that allows a local attacker to intercept and potentially manipulate X11 graphical display traffic forwarded over SSH. By pre-creating a socket file in a predictable location, an unprivileged user on the same machine can hijack the connection before the SSH client establishes it, gaining access to sensitive window contents, keyboard input, and other display data. The attack requires the victim to have X11 forwarding enabled and an active SSH session, but does not require elevated privileges.

• CVE-2026-56399MEDIUM 5.0

  Open WebUI versions prior to 0.6.27 contain a flaw that allows authenticated users to access internal services on the same network as the application. An attacker with valid credentials can craft specially formed web requests that bypass the application's safeguards against server-side request forgery (SSRF), potentially reaching backend systems and sensitive configurations that should be off-limits. This is particularly concerning because it chains authentication with network access to internal resources.

• CVE-2026-56777MEDIUM 5.0

  n8n, a popular workflow automation platform, contains a security flaw in its Python Code node that allows authenticated users to bypass safety checks and access underlying system modules. This vulnerability only affects self-hosted instances where Python task execution is enabled and environment variable access is permitted. An attacker with workflow creation permissions could potentially read sensitive environment variables that the task runner process can access, such as API keys or database credentials.

• CVE-2026-57282MEDIUM 5.0

  Jenkins Git client Plugin versions 6.6.0 and earlier contain a command injection flaw in how they construct SSH wrapper scripts. When a Jenkins build runs in a workspace with a specially crafted directory name, an attacker with the ability to influence that workspace path can inject arbitrary shell commands that execute on the Jenkins agent. This requires authenticated access to create or influence build job workspace naming, limiting immediate risk but creating a significant privilege-escalation vector in multi-tenant Jenkins environments.

• CVE-2026-57439MEDIUM 5.0

  CyberChef, a popular web application for data encryption, encoding, and analysis, contains a vulnerability in its Series Chart operation that allows an attacker to inject malicious code. By crafting a specially formatted CSV file with a `__proto__` key, an attacker can exploit prototype pollution—a JavaScript technique that modifies object prototypes—to insert harmful JavaScript that gets executed when the data is rendered in HTML. This requires user interaction (opening/uploading a file) but can compromise the security of the application and potentially affect downstream operations like UDP packet parsing.

• CVE-2026-58057MEDIUM 5.0

  Flowise, a visual AI workflow platform, contains a vulnerability in how it validates environment variables for custom integrations on Windows systems. Attackers with legitimate access to configure workflow nodes can bypass security restrictions by using a lowercase variant of a blocked environment variable, allowing them to execute arbitrary code on the Flowise server. This requires authentication and administrator-level access to the workflow configuration interface.

• CVE-2026-59100MEDIUM 5.0

  LobeChat versions up to 2.2.9 suffer from a broken authorization flaw that allows authenticated users to view, modify, and delete chat agent data belonging to other users. By guessing or knowing other users' group identifiers, an attacker with valid login credentials can manipulate agent configurations in groups they don't own, effectively hijacking or disabling shared chat automation workflows without detection or consent.

• CVE-2026-59152MEDIUM 5.0

  The LangSmith Client SDKs contain a vulnerability in their TracingMiddleware component that allows an attacker to read arbitrary files from a server and secretly upload them to LangSmith. An attacker with read access to a LangSmith workspace—such as a low-privilege team member, contractor, or compromised account—can exploit this to access files on any server running the vulnerable middleware, even without direct server access. This breaks the intended security boundary by granting file-read capabilities beyond what workspace permissions should allow. The vulnerability affects all LangSmith SDK versions before 0.8.18.

• CVE-2026-59253MEDIUM 5.0

  n8n, a workflow automation platform, contains a flaw in how it checks permissions when users create or modify workflows. An authenticated user can craft specially-formatted requests to place workflows into folders belonging to other projects—even if they shouldn't have access to those projects. This is a logical integrity issue rather than a confidentiality breach; no data is exposed, but folder organization and project boundaries can be violated.

• CVE-2026-6891MEDIUM 5.0

  A vulnerability in My Image Garden for macOS version 3.6.8 and earlier allows a logged-in user to manipulate the installer through specially crafted symbolic links, potentially gaining permission changes to files they shouldn't normally access. This is a local privilege escalation risk that requires both system access and user interaction during installation.

• CVE-2026-6892MEDIUM 5.0

  A flaw in Canon's CUPS printer driver installers for macOS allows a local attacker who has login access to a machine to manipulate symbolic links during the installation process. By crafting a malicious symbolic link, an attacker can trick the installer into changing file permissions on directories they shouldn't normally be able to modify. This is a local-only attack that requires an attacker to already have user-level access to the system.

• CVE-2026-9903MEDIUM 5.0

  Google Chrome versions prior to 148.0.7778.216 contain a vulnerability in Site Isolation, a security feature designed to prevent malicious websites from accessing data from other sites you visit. An attacker who has already compromised Chrome's renderer process—the part that interprets web content—can craft a specially designed MHTML file (a web archive format) that bypasses this protection. This requires the attacker to have gained initial access to the renderer process and the user to open the malicious file, but if successful, it could allow unauthorized access to sensitive information across site boundaries.

• CVE-2026-9942MEDIUM 5.0

  CVE-2026-9942 is a memory safety issue in ANGLE, the graphics abstraction layer used by Google Chrome. When a remote attacker has already compromised Chrome's renderer process, they can exploit this uninitialized memory condition to break out of Chrome's site isolation sandbox using a specially crafted HTML page. Site isolation is Chrome's primary defense against cross-site data theft; bypassing it allows an attacker to read data from other websites the user is visiting. This requires the renderer process to be already compromised, meaning it is a post-compromise escalation rather than an entry point.

• CVE-2026-9979MEDIUM 5.0

  CVE-2026-9979 is a site isolation bypass vulnerability in Google Chrome that allows an attacker to escape the security boundary between different websites if they have already compromised Chrome's rendering engine. An attacker would need to trick a user into visiting a malicious HTML page while the renderer process is already under their control. Site isolation is Chrome's core defense mechanism that prevents one website's scripts from accessing another website's data; this vulnerability undermines that protection in a limited but serious scenario.

• CVE-2026-9980MEDIUM 5.0

  Google Chrome versions before 148.0.7778.216 contain a flaw in how it validates input when printing documents. An attacker who has already compromised Chrome's rendering engine can exploit this to bypass Site Isolation, a security boundary that separates data between websites. This requires both a prior compromise of the renderer process and user interaction, making it a secondary attack in a chain rather than a standalone entry point.

• CVE-2025-64719MEDIUM 4.9

  Gogs, a self-hosted Git service, contains a flaw that allows users with file creation permissions to crash the repository or wiki interface. When a user creates a file, the system attempts to retrieve commit history for the listing pages. If that retrieval fails, the entire page fails with an HTTP 500 error instead of handling the error gracefully. This makes the file listing unusable until the issue is resolved. The vulnerability affects Gogs versions before 0.14.3.

• CVE-2026-0285MEDIUM 4.9

  A vulnerability in Palo Alto Networks PAN-OS allows authenticated administrators with access to the management interface to make unauthorized requests from the firewall itself to internal services. This server-side request forgery (SSRF) flaw could enable an admin to pivot toward backend systems or services that should only be reachable from within the firewall's network. The risk is materially reduced if you follow Palo Alto Networks' recommended practice of restricting management interface access to trusted internal IP addresses only.

• CVE-2026-10039MEDIUM 4.9

  The Frontend Admin plugin for WordPress contains a SQL injection vulnerability that allows authenticated administrators to extract sensitive data from the website's database. The flaw exists in how the plugin processes the 'order' parameter—it fails to properly escape user input before inserting it into database queries. An attacker with administrator privileges can craft a malicious request containing both 'order' and 'orderby' parameters to inject additional SQL commands and retrieve unauthorized information. This vulnerability affects all versions up to and including 3.28.28.

• CVE-2026-10074MEDIUM 4.9

  DreamMaker, a product developed by Interinfo, contains a vulnerability that allows authenticated administrators or privileged users with local access to read arbitrary files from the system. An attacker with elevated privileges can exploit a path traversal flaw to access sensitive system files they shouldn't normally be able to retrieve, potentially exposing configuration data, credentials, or other protected information.

• CVE-2026-10645MEDIUM 4.9

  Zephyr's ext2 filesystem driver contains memory safety flaws when processing directory entries on ext2-formatted storage devices. An attacker who supplies a crafted ext2 image—via USB drive, SD card, or mounted file—can trigger out-of-bounds reads that leak kernel heap memory, cause the system to hang indefinitely, or corrupt memory. Any filesystem operation (opening files, listing directories, deleting files, renaming) on the malicious volume triggers the defect. The vulnerability affects all Zephyr versions from 3.5.0 through 4.4.0.

• CVE-2026-10736MEDIUM 4.9

  The Tutor LMS plugin for WordPress, a popular learning management system, contains a SQL injection flaw in versions up to 3.9.11. An authenticated administrator can inject malicious SQL commands through the 'data' parameter to extract sensitive database information. This is not an unauthenticated attack—it requires admin-level access—which significantly limits real-world exposure in most WordPress installations.

• CVE-2026-10741MEDIUM 4.9

  Sonatype Nexus Repository Manager versions before 3.93.0 have a flaw that allows repository administrators with delegated authority to access upstream proxy credentials that should be protected. When a Nexus instance is configured to proxy external repositories, it stores credentials needed to authenticate with those upstream servers. An administrator with limited permissions—one who manages only specific repositories—can exploit this vulnerability to retrieve credentials that should remain hidden even from them. This is a credential disclosure issue with medium severity that requires administrator-level access to exploit.

• CVE-2026-11360MEDIUM 4.9

  A SQL injection vulnerability exists in the Advanced Order Export For WooCommerce plugin affecting all versions through 4.0.10. Authenticated users with shop manager privileges or higher can inject malicious SQL commands through the 'sort_direction' parameter to extract sensitive data from the WordPress database. The vulnerability requires valid authentication and specific WordPress capabilities, limiting the attack surface to trusted internal users or compromised administrator accounts.

• CVE-2026-11776MEDIUM 4.9

  The Form Maker by 10Web WordPress plugin is susceptible to SQL injection through the 'groupids' parameter. An authenticated administrator can craft malicious input to execute unauthorized database queries and extract sensitive information. The vulnerability affects all versions up to and including 1.15.43 and requires administrator-level credentials to exploit.

• CVE-2026-11777MEDIUM 4.9

  The Form Maker by 10Web WordPress plugin contains a SQL injection vulnerability in its 'name' parameter that allows authenticated administrators to execute arbitrary SQL queries. An attacker with admin access could extract sensitive database information by injecting malicious SQL code into form submissions. The vulnerability affects all versions up to and including 1.15.43.

• CVE-2026-11789MEDIUM 4.9

  389 Directory Server contains a flaw in its SMD5 password storage plugin that causes the LDAP server to crash when processing certain malformed password hashes. An authenticated attacker with high privileges can trigger this crash by submitting a specially crafted password hash shorter than 16 bytes, which causes the plugin to miscalculate memory boundaries and read beyond allocated buffers. The crash results in denial of service but does not leak sensitive data or compromise authentication logic itself.

• CVE-2026-11790MEDIUM 4.9

  389 Directory Server contains a weakness in how it handles password verification when using PBKDF2-SHA256 encryption. An attacker with administrative access can tamper with a user's stored password hash to include an extremely high iteration count. When that user (or an automated system) attempts to authenticate, the server wastes excessive CPU resources recalculating the password hash, effectively freezing responsiveness for legitimate users. This is a denial-of-service attack that requires the attacker to already have administrative privileges to modify password data.

• CVE-2026-11793MEDIUM 4.9

  A stack buffer overflow vulnerability has been identified in 389 Directory Server's password handling code. When parsing specially crafted credentials, the checkPrefix() function in pw.c fails to validate the length of an algorithm identifier before copying it into a fixed 256-byte buffer. An attacker with Directory Manager credentials can exploit this to crash the LDAP server. While the vulnerability could theoretically allow code execution, compiler protections like FORTIFY_SOURCE limit the practical impact to denial of service in most deployments.

• CVE-2026-11827MEDIUM 4.9

  GitLab Enterprise Edition contains an authorization flaw that allows authenticated maintainers to access other users' stored credentials without proper permission checks. The issue affects multiple release lines and requires an attacker to already have maintainer-level access to the affected project or group. While the technical barrier is moderate, the impact—unauthorized disclosure of sensitive credentials—makes this a notable security concern for organizations running vulnerable GitLab instances.

• CVE-2026-11844MEDIUM 4.9

  CVE-2026-11844 is a path traversal vulnerability in IEI Integration Corp's iVEC-IEI Virtualization Edge Computer that allows an authenticated attacker with elevated privileges to read arbitrary files outside the device's intended directory structure. An attacker with high-level access to the device could potentially access sensitive configuration files, credentials, or other protected data by manipulating file path requests. This is not a vulnerability an unauthenticated external attacker can exploit—it requires someone with administrative or privileged credentials already on the system.

• CVE-2026-11986MEDIUM 4.9

  Keycloak's admin-ui-ext component has a permission-checking flaw that allows limited administrators to strip high-privilege roles from other users or administrators. An attacker with delegated admin rights can exploit bulk role-removal endpoints to bypass intended access controls, potentially locking out legitimate administrators or escalating their own influence within the system.

• CVE-2026-12089MEDIUM 4.9

  The LWS Optimize WordPress plugin contains a path traversal vulnerability that allows authenticated users with Editor-level permissions or higher to read files from the server's filesystem. The plugin's CSS optimization feature trusts stylesheet URLs embedded in HTML without properly validating that resolved file paths stay within WordPress's directory boundaries. An attacker with sufficient WordPress account privileges could exploit this to access sensitive configuration files, database credentials, or other non-public data stored on the server.

• CVE-2026-12920MEDIUM 4.9

  The Cookie Banner for GDPR / CCPA – WPLP Cookie Consent WordPress plugin contains a SQL injection vulnerability in versions 4.3.5 and earlier. An authenticated administrator can craft malicious input through the 's' parameter to inject arbitrary SQL commands and extract sensitive data from the WordPress database. While the vulnerability requires admin-level access to exploit, it poses a meaningful risk in multi-user WordPress environments or where admin credentials have been compromised.

• CVE-2026-12936MEDIUM 4.9

  The Recurio – Ultimate Subscription for WooCommerce plugin contains a SQL injection vulnerability in its 'data' parameter that allows authenticated shop managers and administrators to extract sensitive database information. The vulnerability exists because user input is not properly sanitized before being used in database queries. An attacker with legitimate shop manager credentials or higher can craft malicious requests to execute unintended SQL commands and access confidential data like customer records, payment information, or other sensitive details stored in the WordPress database.

• CVE-2026-13357MEDIUM 4.9

  The Houzez Property Feed plugin for WordPress contains a SQL injection vulnerability in versions up to 2.5.46. An attacker with administrator-level access can manipulate the 'orderby' parameter to execute arbitrary SQL queries and steal sensitive data from the website's database. The vulnerability exists because user input is only lightly sanitized before being inserted into an SQL statement, bypassing WordPress's built-in protection mechanisms.

• CVE-2026-13434MEDIUM 4.9

  KubeVirt's network annotation generator fails to validate user-supplied network names before writing them into pod configuration. A tenant with edit permissions can craft a malicious network name containing JSON that tricks Multus (the Kubernetes networking plugin) into connecting workloads to networks in other namespaces, bypassing network isolation. This risk is amplified when the ExternalNetResourceInjection feature gate is enabled, which skips additional safety checks. The attacker gains access to segregated network segments and can spoof IP and MAC addresses.

• CVE-2026-14342MEDIUM 4.9

  The Mail Mint WordPress plugin, used for email marketing and WooCommerce integration, contains a SQL injection vulnerability affecting all versions up to 1.24.2. An authenticated administrator can manipulate the 'contact_ids' parameter to inject malicious SQL commands and extract sensitive data from the WordPress database. While access requires administrator-level privileges, the vulnerability allows unauthorized data exfiltration once those credentials are compromised or misused.

• CVE-2026-14362MEDIUM 4.9

  HashiCorp's memberlist library, a Go package used for cluster membership and failure detection, contains a flaw in how it processes incoming gossip protocol messages that can cause a process to run out of memory and crash. An attacker with network access to the gossip port can trigger this condition remotely. The vulnerability is resolved in memberlist version 0.6.0 and later.

• CVE-2026-14475MEDIUM 4.9

  The WPLP Cookie Consent plugin for WordPress contains a SQL injection vulnerability affecting all versions up to 4.3.6. An attacker with WordPress administrator access can manipulate the 'scan_id' parameter to inject malicious SQL commands and extract sensitive data from the website's database. While this requires high-level credentials to exploit, the impact on data confidentiality is significant.

• CVE-2026-36778MEDIUM 4.9

  A stack overflow vulnerability has been identified in Tenda O3 Wireless Router version 1.0.0.5(4180). The flaw exists in how the router processes the username parameter when handling certain web requests. An attacker with elevated privileges on the network can send a specially crafted HTTP request that causes the router to crash, resulting in a denial of service. The router would need to be rebooted to restore functionality. This is a medium-severity issue that requires administrative-level access to exploit, limiting its immediate threat in most environments.

• CVE-2026-41280MEDIUM 4.9

  Apache DolphinScheduler contains an authorization flaw that allows authenticated users with system login privileges to delete task definitions in projects they shouldn't have access to. The vulnerability arises from insufficient permission checks when handling task definition deletion requests. An attacker with valid system credentials could exploit this to disrupt workflow orchestration by removing critical task definitions from other projects, potentially causing operational disruption. The issue is limited to versions before 3.4.2 and requires existing system access to exploit.

• CVE-2026-41412MEDIUM 4.9

  alf.io is an open-source ticketing platform used by conferences and events to manage reservations. The vulnerability lies in how alf.io sandboxes custom extensions (plugins) that users can write to extend functionality. The sandbox was designed to safely run untrusted extension code, but it failed to protect file access. Specifically, extensions have access to an HTTP client tool that includes a method for uploading files. This method does not validate or restrict which files can be read—a malicious extension can read any file that the alf.io application has permission to access on the server, then send that file's contents to an attacker's server. An attacker would need to either write a malicious extension or trick an administrator into installing one, but once active, the extension can quietly exfiltrate sensitive data like configuration files, database credentials, or user records.

• CVE-2026-42147MEDIUM 4.9

  Coolify, an open-source platform for managing servers and cloud infrastructure, contains a vulnerability that allows authenticated administrators with storage permissions to trick the application into making requests to sensitive internal systems. When configuring S3 storage endpoints, Coolify validates only the URL format without blocking access to private networks or cloud metadata services. An attacker with legitimate admin credentials could exploit this to probe internal systems, potentially discovering sensitive configuration data or service credentials exposed through metadata endpoints. The vulnerability is fixed in version 4.0.0-beta.474.

• CVE-2026-43752MEDIUM 4.9

  FileMaker Server contains a vulnerability in the Open Source LLM setup feature accessible from the Admin Console. An authenticated administrator can upload a specially crafted file that leads to arbitrary code execution on the underlying host system. The vulnerability requires admin-level access and has been patched in FileMaker Server 26.0.1.

• CVE-2026-44917MEDIUM 4.9

  A vulnerability in OpenStack Ironic before version 35.0.2 allows authenticated project administrators or managers to read sensitive files directly from the Ironic conductor server through a specially crafted PXE template. This is a credential-required attack where an insider with project admin or manager privileges can exploit the template processing mechanism to access files they shouldn't be able to retrieve, potentially exposing configuration secrets, credentials, or other sensitive data stored on the conductor.

• CVE-2026-45559MEDIUM 4.9

  Roxy-WI, a web management interface for HAProxy, Nginx, Apache, and Keepalived servers, contains an LDAP injection vulnerability in its user authentication module. An administrator can manipulate the username parameter to break out of the intended LDAP search filter and enumerate or extract sensitive user attributes—such as email addresses—that should remain hidden. The vulnerability exists because user input is directly concatenated into the LDAP query without proper sanitization or escaping. While exploitation requires administrative privileges, the attack is straightforward once access is obtained, making it a credential harvesting risk in multi-tenant or federated identity environments.

• CVE-2026-45684MEDIUM 4.9

  OpenTelemetry eBPF Instrumentation versions 0.7.0 through 0.8.x contain a buffer handling flaw in the log enricher component. When log injection is enabled, attackers can craft a multi-segment write operation that tricks the instrumentation into reading beyond the intended buffer boundary, potentially overwriting memory. This could lead to information disclosure, data corruption, or application instability on systems using the affected versions.

• CVE-2026-45731MEDIUM 4.9

  WWBN AVideo, an open-source video hosting platform, contains a file-read vulnerability in its database migration feature. An authenticated administrator can manipulate the migration process to read sensitive text files from the server's filesystem. This requires existing admin access, so it poses a targeted insider risk rather than a mass-exploitation threat, but it can expose configuration files, credentials, or other data to a compromised or malicious admin account.

• CVE-2026-46464MEDIUM 4.9

  Dell PowerProtect Data Domain versions 7.7.1.0 through 8.7, along with specific LTS releases (8.6.1.0–8.6.1.10, 8.3.1.0–8.3.1.30, and 7.13.1.0–7.13.1.70), contain a link-following vulnerability that could allow a high-privileged remote attacker to access sensitive information. The flaw stems from improper validation of symbolic links or file path traversal before accessing files on the system. While the attack requires elevated privileges and does not permit system modification or unavailability, the potential for information disclosure in a data protection appliance warrants attention.

• CVE-2026-48947MEDIUM 4.9

  CVE-2026-48947 is a medium-severity access control flaw in Joomla that allows privileged users to overwrite media files even when they lack explicit editing permissions. The vulnerability stems from an improper check during file operations, enabling privilege escalation within the application's permission model. While the issue requires an authenticated attacker with elevated privileges, it can lead to unauthorized modification of critical media assets.

• CVE-2026-49198MEDIUM 4.9

  CVE-2026-49198 is a medium-severity access control flaw in Acer Predator Connect W6X MQTT brokers that allows high-privileged users to subscribe to wildcard topics, inadvertently gaining visibility into all MQTT traffic flowing through the system. While the vulnerability requires authenticated access with elevated permissions, once exploited it enables an insider or compromised admin account to eavesdrop on sensitive IoT communication without additional authorization constraints. This is a confidentiality risk with no impact to system availability or integrity.

• CVE-2026-50219MEDIUM 4.9

  libexpat, a widely-used XML parsing library, contains a use-after-free vulnerability in versions before 2.8.2. The flaw occurs when certain XML parsing functions (XML_GetBuffer, XML_Parse, XML_ParseBuffer, XML_ParserFree, or XML_ParserReset) are called from within event handlers without proper depth tracking. This can lead to memory safety violations and potentially allow attackers to crash applications or, in some scenarios, execute arbitrary code. The vulnerability requires local access and specific conditions to trigger, making it a moderate-risk issue rather than a widespread internet-facing threat.

• CVE-2026-50224MEDIUM 4.9

  The Acer Connect M6E 5G router's web administration panel is configured to listen on all public IPv6 addresses on port 8080, without built-in firewall protections. This means the internal API endpoints used to manage the device can be reached directly over the internet by anyone who knows the device exists and its IPv6 address, potentially allowing unauthorized access to sensitive configuration and status information.

• CVE-2026-50565MEDIUM 4.9

  Fission, an open-source serverless framework for Kubernetes, had a configuration flaw in versions before 1.24.0 where builder pods automatically mounted sensitive service account credentials into user-supplied container images. This meant anyone deploying a function through Fission could potentially access the credentials needed to interact with your Kubernetes cluster, such as listing resources or reading secrets. The issue stems from Kubernetes' default behavior of auto-mounting service account tokens unless explicitly disabled—Fission wasn't disabling this protection for builder pods. The flaw has been patched in version 1.24.0.

• CVE-2026-55079MEDIUM 4.9

  Coder, a platform for provisioning remote development environments via Terraform, contains a memory exhaustion vulnerability in its provisioner daemon. When clients upload files, the daemon fails to validate the claimed file size before allocating memory, allowing an authenticated attacker to trigger out-of-memory conditions by claiming extremely large file sizes. The vulnerability affects versions 2.24.0 through 2.34.1 across multiple release branches, and has been patched in versions 2.29.7, 2.32.7, 2.33.8, and 2.34.2.

• CVE-2026-56080MEDIUM 4.9

  Capgo versions before 12.128.2 have a bug in their password policy enforcement system. When a Super Admin enables the Enforce Password Policy feature and changes their password to meet the requirements, the system incorrectly continues to treat the account as non-compliant. This causes the backend to repeatedly force password reset prompts, effectively locking the Super Admin out of their organization—even though their credentials are valid. The result is a denial of service affecting administrative access.

• CVE-2026-56131MEDIUM 4.9

  libexpat, a widely-used XML parsing library, contains a use-after-free vulnerability in versions before 2.8.2. The flaw occurs when the XML_ResumeParser function is called from within a handler and a policy violation is detected. In such cases, the library fails to properly track the call stack depth, leaving freed memory accessible and potentially exploitable. This is analogous to an earlier vulnerability (CVE-2026-50219) in the same codebase.

• CVE-2026-56149MEDIUM 4.9

  A vulnerability in Elasticsearch allows users with administrative or elevated privileges to crash an Elasticsearch node by submitting a malicious machine learning request. The attack causes the affected node to consume excessive memory until it becomes unavailable, effectively denying service to legitimate users. This is a resource exhaustion issue that requires elevated permissions to exploit, limiting its immediate blast radius but still posing operational risk in multi-tenant or shared Elasticsearch environments.

• CVE-2026-56228MEDIUM 4.9

  Capgo versions before 12.128.2 contain a flaw that allows organization administrators to set an impossibly high minimum password length requirement—potentially billions of characters—without any safeguards. Once enabled, this broken policy locks out all users in the organization from changing their passwords or logging in, effectively shutting down access to the application for everyone. This is a self-inflicted denial of service that requires admin-level access to trigger but can paralyze an entire organization.

• CVE-2026-56412MEDIUM 4.9

  libexpat, a widely-used XML parsing library, contains a use-after-free vulnerability in its CDATA section handling. The issue stems from incomplete depth tracking when handlers are invoked during XML processing, allowing memory already freed by one operation to be accessed by another. This affects versions before 2.8.2. While the immediate trigger requires specific XML policy violations and local system access, the underlying condition can lead to crashes or potentially arbitrary code execution if an attacker controls the malicious XML input.

• CVE-2026-59819MEDIUM 4.9

  LiteLLM, a proxy server used to standardize API calls to large language models, contains a vulnerability in its connection testing endpoint that allows authorized administrators to read sensitive files from the server's filesystem. An attacker with legitimate access to the `/health/test_connection` API could craft requests that reference local files via OIDC configuration parameters, effectively bypassing normal file access restrictions. This affects versions before 1.83.10-stable and is particularly concerning in shared or multi-tenant environments where proxy administration privileges may be distributed.

• CVE-2026-59854MEDIUM 4.9

  SiYuan is a popular open-source personal knowledge management platform. Versions prior to 3.7.1 contain a file-access vulnerability in the global file copy API endpoint that allows authenticated users with administrator privileges or valid API tokens to extract sensitive credential files from the host system. An attacker can exploit a gap in the file-path validation logic to copy commonly-used credential stores—such as Git credentials, SSH keys, database passwords, Kubernetes configuration, and Docker authentication tokens—from the user's home directory into the SiYuan workspace, then download them via the file API. The vulnerability requires valid authentication, limiting exposure to insider threats or compromised service accounts.

• CVE-2026-6448MEDIUM 4.9

  The Quiz and Survey Master plugin for WordPress contains a SQL injection flaw in how it processes the 'order' parameter. An admin-level attacker can craft malicious requests to extract sensitive data from the WordPress database. The vulnerability is time-based and blind, meaning attackers infer results through response delays rather than direct output. If the plugin's secret key becomes public, lower-privileged users could exploit it without admin credentials.

• CVE-2026-7547MEDIUM 4.9

  The Woosa – Marktplaats for WooCommerce plugin contains a path traversal vulnerability that allows authenticated site administrators to read sensitive files from the server, such as the WordPress configuration file. An attacker with admin credentials can manipulate a file parameter to escape the plugin's intended log directory and access arbitrary files, potentially exposing database credentials and other secrets stored in wp-config.php.

• CVE-2026-7565MEDIUM 4.9

  The LearnPress – Backup & Migration Tool plugin for WordPress contains a flaw that allows site administrators to read files from anywhere on the server by manipulating a parameter called 'import-user-file'. While this requires admin-level access to exploit, the impact is serious: attackers with those credentials could retrieve sensitive configuration files, database credentials, or other confidential data stored on the server.

• CVE-2026-8978MEDIUM 4.9

  The OptinCraft WordPress plugin contains a SQL injection vulnerability in its 'order_by' parameter that allows authenticated administrators to extract sensitive database information. The flaw exists because user input is not properly escaped before being used in database queries. While this requires admin-level access to exploit, it represents a significant insider threat risk, especially in multi-user WordPress environments where administrative accounts may be compromised or operated by untrusted parties.

• CVE-2026-9083MEDIUM 4.9

  Keycloak realm administrators with "manage-realm" permissions can craft specially designed filesystem paths when setting up key provider components, allowing them to determine which files exist and can be read by the Keycloak service. This reconnaissance capability enables attackers to map the system's file structure and identify sensitive files before attempting further compromise.

• CVE-2026-9197MEDIUM 4.9

  Smart Slider 3, a popular WordPress plugin, contains a directory traversal vulnerability affecting all versions up to 3.5.1.36. An authenticated WordPress administrator can exploit the replaceHTMLImage function to read sensitive files from the server, including configuration files, database credentials, and other confidential data. The attack requires administrator-level privileges and direct knowledge of the server's file paths, but poses a meaningful risk in environments where admin access is delegated or compromised.

• CVE-2026-9576MEDIUM 4.9

  The Fluent Booking WordPress plugin contains an authorization flaw that allows Calendar Managers to download attendee personal information from booking groups they don't have permission to access. An authenticated user with the Calendar Manager role can exploit this to retrieve names, emails, phone numbers, addresses, and payment details of attendees from groups outside their scope. This vulnerability affects versions prior to 2.1.2 and requires administrator intervention to patch.

• CVE-2026-9801MEDIUM 4.9

  Keycloak has a vulnerability that allows a high-privileged attacker—such as a realm administrator or someone who has compromised an upstream LDAP server—to crash the Keycloak service by sending a specially crafted LDAP password policy response. When triggered during authentication, this causes the Java process to run out of memory and shut down, knocking the service offline for all users on that node. The attack requires either legitimate administrative access to configure a malicious LDAP directory or prior compromise of an existing LDAP backend.

• CVE-2024-1248MEDIUM 4.8

  A vulnerability in federated authentication systems allows an attacker to hijack and reassign user roles when a federated identity provider has silent just-in-time account provisioning enabled. If a federated user shares a username with an existing local account, the provisioning process can overwrite the local user's roles with whatever the federated system assigns. An attacker needs to know a legitimate local username and have access to a federated identity provider configured with silent JIT provisioning to execute this attack. The impact is limited to role changes (typically to minimal access levels) rather than full account takeover.

• CVE-2026-0266MEDIUM 4.8

  A stored cross-site scripting (XSS) vulnerability exists in Palo Alto Networks PAN-OS that allows an authenticated administrator to inject malicious JavaScript into the web interface. The payload persists in the system and executes when other users access the affected interface, potentially compromising their sessions or stealing sensitive data. The vulnerability requires valid administrator credentials to exploit, which significantly limits the attack surface but remains a genuine concern for insider threats or compromised admin accounts.

• CVE-2026-10057MEDIUM 4.8

  ITS Intelligent SCADA System contains a stored cross-site scripting (XSS) vulnerability that allows authenticated users with elevated privileges to inject malicious JavaScript code into the application. Once injected, this code persists in the system and executes automatically whenever other users load affected pages in their browsers. This is distinct from reflected XSS because the payload remains embedded in the application, posing a sustained risk to all users who access the compromised content.

• CVE-2026-10058MEDIUM 4.8

  ITS Intelligent SCADA System contains a stored cross-site scripting (XSS) flaw that lets high-privilege attackers inject malicious JavaScript into the system. When other users load affected pages, that injected code runs in their browsers automatically. This is a persistence threat—the malicious script stays in the system until removed, affecting anyone who accesses the compromised page.

• CVE-2026-10634MEDIUM 4.8

  A race condition in Zephyr's TCP stack allows an attacker on the local network to crash the system or potentially read sensitive memory. The vulnerability exists in how the TCP layer iterates through active connections while a background thread can simultaneously free those connections, causing the iterator to access memory that has already been released. While the attacker needs local network access and user privileges to trigger the issue reliably, the outcome is denial of service or information disclosure on affected embedded and IoT devices running Zephyr.

• CVE-2026-10639MEDIUM 4.8

  Zephyr's IPv4 networking stack contains a use-after-free vulnerability in how it handles ping (ICMP echo) responses. When the device sends back a reply to an incoming ping, it processes the packet through the transmission path, which may immediately free the packet's memory. The code then attempts to read data from that freed memory to update network statistics, creating a window for reading corrupted or recycled data. An attacker can trigger this by sending repeated pings to a Zephyr device, potentially causing statistics corruption or a crash. The vulnerability requires specific configuration options to be enabled and the timing must align with the kernel's memory recycling, making exploitation probabilistic but feasible over a network without authentication.

• CVE-2026-10652MEDIUM 4.8

  A flaw in the Zephyr real-time operating system's DNS resolver allows an attacker to read unintended memory contents from a device. When a malicious or spoofed DNS server sends a specially crafted response with a TXT or SRV record claiming to be larger than it actually is, the resolver reads beyond the DNS response packet into adjacent memory—potentially exposing stale data from prior DNS queries or uninitialized memory pools. An attacker on the network, intercepting DNS traffic, or (if multicast DNS is enabled) any device on the local network can trigger this. The leaked data is typically small (under 64 bytes) and read-only, but in some cases could cause a crash. Zephyr versions 4.3.0 and 4.4.0 are affected.

• CVE-2026-12491MEDIUM 4.8

  vLLM, a widely-used open-source library for running large language models, has a flaw in how it processes images. When the library converts images to a standard format (RGB), it does not properly preserve image metadata like EXIF orientation information or transparency data (PNG tRNS chunks). This mishandling can cause transparent pixels to be rendered incorrectly or remapped unexpectedly, distorting the original image content that the model sees. The result is that the language model may misinterpret what's in the image, compromising data integrity.

• CVE-2026-12549MEDIUM 4.8

  A regression in HTTP Range request handling allows clients to craft requests that bypass security checks, resulting in malformed responses and potential service disruption. When a Range request specifies a suffix length larger than the content being served, the code fails to properly validate the resulting negative offset, leading to incorrect HTTP 206 Partial Content responses and excessive logging that can consume resources.