2026 · Low

Low-severity vulnerabilities disclosed in 2026

Low-rated CVEs published in 2026, with SEC.co remediation and prioritization guidance.

385 published vulnerabilities · page 3 of 4

• CVE-2026-14789LOW 3.3

  A stack-based buffer overflow vulnerability exists in radare2's Memory64ListStream Parser component, which processes memory dump files. An attacker with local access can manipulate a specially crafted input file to trigger a buffer overflow, potentially causing the application to crash. This is a localized denial-of-service issue affecting radare2 versions up to 6.1.6.

• CVE-2026-14790LOW 3.3

  A null pointer dereference vulnerability exists in GPAC 26.02.0's NHML file writing component. An attacker with local system access can trigger this flaw by manipulating input to the nhmldump_send_frame function, causing the application to crash. The vulnerability has limited scope—it only affects availability on the local machine and requires authenticated access. While an exploit has been published, the overall risk is low due to the access requirements and lack of remote exploitability.

• CVE-2026-14801LOW 3.3

  A divide-by-zero flaw exists in GPAC's TeXML file handler that can crash the application when processing specially crafted text files. The vulnerability requires local access and a user with at least limited system privileges to trigger; it causes a denial of service but does not leak data or allow code execution. The issue affects GPAC version 26.03-DEV-rev342-g80071f700-master and is remediated by applying the identified patch commit.

• CVE-2026-15115LOW 3.3

  A vulnerability in Google Chrome on Android allows a local attacker to bypass the same-origin policy—a core browser security feature that prevents websites from accessing data from other origins—through a specially crafted HTML page. The issue stems from insufficient validation of user-supplied input in the WebAppInstalls component. An attacker with local access to the device would need to trick a user into visiting a malicious webpage to exploit this. The vulnerability was patched in Chrome version 150.0.7871.115.

• CVE-2026-15184LOW 3.3

  GNU LibreDWG versions up to 0.13.4 contain a flaw in how the DWG file handler processes certain file structures. When a specially crafted DWG file is opened, the software can crash due to a null pointer dereference. An attacker with local access to a system could exploit this to cause a denial of service. The issue requires user interaction—specifically opening a malicious file—and affects only the availability of the application, not data confidentiality or integrity. Version 0.14 and later resolve this issue.

• CVE-2026-15185LOW 3.3

  GPAC version 26.03-DEV contains a flaw in how it processes subtitle index files (vobsub format) through its MP4Box tool. An attacker with local access can manipulate specific data in these files to trigger the software to read beyond allocated memory boundaries. This out-of-bounds read could crash the application or potentially leak sensitive data from memory. The vulnerability is of low severity due to its local-only nature and limited impact surface, though the public disclosure means exploit code may be available.

• CVE-2026-15194LOW 3.3

  Open5GS version 2.7.7 contains a use-after-free vulnerability in the AMF (Access and Mobility Management Function) component's context cleanup routine. An authenticated local attacker can trigger this flaw by manipulating the amf_context_final function, potentially leading to information disclosure. While the vulnerability requires local access and low privileges, public exploit code now exists, increasing practical risk despite the low CVSS rating.

• CVE-2026-15274LOW 3.3

  CVE-2026-15274 is a low-severity denial-of-service vulnerability in fbxcel, a Rust library for parsing FBX (Autodesk FBX) files. The flaw exists in the Node Header Handler component of the pull parser and can be triggered by a local attacker to crash or hang the application. The vulnerability affects fbxcel versions up to 0.9.0, and a fix is pending in the development repository but has not yet been released to the public.

• CVE-2026-15276LOW 3.3

  Symphonia, a popular audio processing library, contains a flaw in how it handles metadata that can cause the application to become unresponsive or crash. The issue is local only, meaning an attacker must already have access to the system running the software. A patch is being prepared but has not yet been officially released. The overall risk is low because of these constraints, but you should still monitor for the fix if your organization uses this library.

• CVE-2026-21027LOW 3.3

  CVE-2026-21027 is a low-severity vulnerability in Samsung's ImsSettings application that allows a local attacker with existing device access to trigger logging functions through improperly exported application components. The attack requires the attacker to already have user-level privileges on the device and does not enable data theft or system takeover—the primary risk is integrity impact through manipulation of logging behavior.

• CVE-2026-21034LOW 3.3

  Samsung Auto versions prior to 3.1.2.61 (Android 15) and 3.2.0.38 (Android 16) contain a flaw that improperly exposes application components. A local attacker with user-level access can exploit this exposure to modify audio settings without user consent. The vulnerability is rated LOW severity and does not affect confidentiality or system availability, only the integrity of audio configuration.

• CVE-2026-28586LOW 3.3

  CVE-2026-28586 is a local information disclosure vulnerability in Android's AppOpsService that allows an already-authenticated user to bypass permission checks and read sensitive data they shouldn't have access to. The flaw requires the attacker to already have a local account on the device; there's no way to exploit it remotely. The exposure is classified as low-severity because the data leaked is limited and no system functions are disrupted.

• CVE-2026-34049LOW 3.3

  Coolify, an open-source server and application management tool, contains a command injection vulnerability in its MongoDB backup feature. Versions 4.0.0-beta.451 through 4.0.0-beta.470 fail to properly sanitize collection names before passing them to shell commands, allowing an authenticated attacker with backup configuration privileges to inject arbitrary commands. The vulnerability requires high-level access and uncommon circumstances to exploit, but could allow a malicious administrator or compromised high-privilege account to execute code on the Coolify host.

• CVE-2026-34149LOW 3.3

  Coolify, an open-source self-hosted platform for managing servers and databases, contains a command injection vulnerability in its backup functionality. When authenticated users with database management permissions configure backup jobs, credentials and collection names they enter are inserted directly into shell commands without proper escaping. An attacker with those permissions could craft malicious input to execute arbitrary commands on the servers Coolify manages. This affects versions prior to 4.0.0-beta.471.

• CVE-2026-41434LOW 3.3

  OP-TEE, a security-focused runtime environment for Arm processors, contains a flaw in its PKCS#11 trusted application that can be triggered by a local attacker with basic privileges. When exploited, the vulnerability causes the PKCS#11 component to enter an infinite recursive loop, crashing that trusted application. The issue affects versions 3.10.0 through 4.10.x; version 4.11.0 and later include the fix.

• CVE-2026-41579LOW 3.3

  runc, the container runtime tool used by Docker and other container platforms, contains a vulnerability in how it sets up container filesystems. A malicious container image can exploit this by making /dev a symlink, causing runc to delete or create files in unexpected locations on the host system. Docker users are protected because Docker adds a read-only layer that blocks this attack, but other container tools built on runc may remain vulnerable if they don't have similar protections. The issue affects versions before 1.3.6, 1.4.3, and 1.5.0.

• CVE-2026-42201LOW 3.3

  Coolify, a self-hosted platform for managing servers and applications, contains a command injection vulnerability in how it handles database credentials. When administrators configure database connections (Redis, KeyDB, Dragonfly, ClickHouse, PostgreSQL, MySQL), the credential values they enter are inserted directly into Docker Compose configuration files without any safety checks. An attacker with administrative access could exploit this by injecting shell metacharacters into credential fields to execute arbitrary commands on the host system. The vulnerability affects all versions prior to 4.0.0-beta.474.

• CVE-2026-45277LOW 3.3

  Nextcloud's approval workflow feature contains an information disclosure flaw that allows authenticated users to determine whether arbitrary files are connected to specific approval processes. An attacker with valid credentials can probe the system to learn if particular files have approval workflows attached, potentially revealing organizational file structures and approval dependencies that should remain confidential. The issue affects versions prior to 2.7.2 and does not require user interaction to exploit.

• CVE-2026-45278LOW 3.3

  Nextcloud's user OIDC (OpenID Connect) module contains an open redirect vulnerability that allows attackers to craft malicious login links. When users click these links to authenticate via OIDC, they are redirected to attacker-controlled websites after logging in. This affects Nextcloud versions 6.1.0 through 8.2.1. The vulnerability has a low CVSS score because it requires user interaction and does not directly compromise confidentiality or availability.

• CVE-2026-45324LOW 3.3

  Rizin, a reverse engineering framework used for binary analysis and code inspection, contains a double free vulnerability in its search functionality. This occurs when the same memory location is freed twice, potentially causing application crashes or unexpected behavior. The vulnerability requires physical access to the system and user interaction to trigger, making it a lower-risk issue in most operational environments.

• CVE-2026-45455LOW 3.3

  A flaw in Microsoft Office Excel can allow an attacker to read memory that shouldn't be accessible, potentially exposing sensitive information on a local system. The vulnerability requires user interaction—someone must open a specially crafted Excel file—but once triggered, it could leak data like file contents or system details. This is a low-severity issue with no direct impact on system availability or file integrity.

• CVE-2026-45459LOW 3.3

  Microsoft Excel has a flaw that allows someone with local access to bypass a built-in security protection mechanism. An attacker would need to trick a user into opening a specially crafted Excel file on their machine. The vulnerability exposes some information (such as file contents or formulas) but cannot be used to modify data or crash the application. This is a low-risk issue with limited real-world impact.

• CVE-2026-45466LOW 3.3

  A flaw in Microsoft Office Word allows an attacker to trigger a heap-based buffer overflow by crafting a malicious document. When a user opens the document, sensitive information stored in the application's memory could be read by the attacker. This is a local attack—the attacker cannot exploit it remotely—and it requires user interaction to open the malicious file. The confidentiality risk is limited; no system damage or data modification occurs.

• CVE-2026-45485LOW 3.3

  Microsoft Office contains a flaw that allows an attacker to read data from memory locations outside the intended bounds, potentially exposing sensitive information stored locally on a user's machine. The vulnerability requires local system access and user interaction (such as opening a malicious document), but does not allow the attacker to modify or delete data or crash the application. This is classified as a low-risk issue because it requires presence on the affected system and a user action to trigger exposure.

• CVE-2026-45613LOW 3.3

  Rizin, a reverse engineering framework used by security researchers and analysts, contains a heap buffer overflow vulnerability in its OMF (Object Module Format) file parsing code. An attacker could craft a malicious OMF binary file that, when opened by a user in Rizin, could read small amounts of sensitive data from the program's memory. This requires local access and user interaction—the user must deliberately open a malicious file.

• CVE-2026-47327LOW 3.3

  CVE-2026-47327 is a denial-of-service vulnerability in Ubuntu Linux affecting versions 6.8, 6.17, and 7.0. A NULL pointer dereference in the AppArmor notification handling code allows any unprivileged local user to crash the kernel without authentication or special permissions. The attack requires only local system access and can be triggered with a single action, causing a kernel oops that disrupts availability but does not compromise confidentiality or integrity.

• CVE-2026-47329LOW 3.3

  Ubuntu Linux versions 6.8, 6.17, and 7.0 contain a flaw in SAUCE patches that handle AppArmor security notifications. The vulnerability stems from improper validation of the name field size in these notifications. An unprivileged local user can exploit this by sending crafted AppArmor responses that bypass validation checks, potentially leading to unexpected behavior in the kernel's handling of these security-related messages. This is a local-only issue with low severity impact.

• CVE-2026-47330LOW 3.3

  CVE-2026-47330 is a local privilege escalation and cache poisoning vulnerability affecting Ubuntu Linux systems with AppArmor SAUCE patches. An unprivileged user can trigger uninitialized variable handling in AppArmor's notification code, causing incorrect caching of security policy responses. While the CVSS score is low (3.3), the issue undermines AppArmor's integrity by allowing cache corruption that could affect subsequent policy enforcement decisions.

• CVE-2026-47336LOW 3.3

  Ubuntu Linux 6.8 has a bug in its AppArmor security module that could allow an unprivileged local user to bypass or weaken network socket access controls. The issue stems from an uninitialized variable in the code that mediates AF_INET and AF_INET6 (IPv4 and IPv6) socket access. While the vulnerability requires local access and does not enable data theft or system crashes, it undermines the purpose of AppArmor's fine-grained network policy enforcement, potentially allowing a local user to perform network operations that should have been restricted.

• CVE-2026-47337LOW 3.3

  A NULL pointer dereference flaw in Ubuntu Linux kernel versions 6.8, 6.17, and 7.0 can be triggered by any unprivileged local user to crash the kernel. The vulnerability exists in socket mediation code that handles both IPv4 and IPv6 traffic. While the flaw itself does not enable data theft or system compromise, it can be exploited to cause a denial of service by forcing a kernel panic, disrupting availability for all users on the affected system.

• CVE-2026-47712LOW 3.3

  Dulwich, a Python library for working with Git repositories, has a path-traversal vulnerability in how it generates patch file names. When creating patch files from Git commits, the library previously used the commit's subject line to create the filename without properly sanitizing it. An attacker could craft a malicious commit with special characters (like slashes, backslashes, or double-dots) in the subject line to trick the library into writing patch files outside the intended directory. This could allow unauthorized file creation in unexpected locations on a system using Dulwich to process untrusted commits.

• CVE-2026-48156LOW 3.3

  pypdf, a popular open-source Python library for PDF handling, contains a vulnerability that allows an attacker to craft malicious PDF files that cause the library to consume excessive processing time during parsing. The issue stems from how pypdf processes cross-reference streams—a mechanism PDFs use to index internal objects—when they contain specific structural patterns. An attacker would need to trick a user or application into opening a specially crafted PDF, but once opened, the library can hang or freeze during PDF processing, resulting in a denial-of-service condition on that system.

• CVE-2026-48935LOW 3.3

  Node.js has a permission model that allows administrators to restrict file system access using flags like `--allow-fs-read` to declare read-only paths. A flaw in how this Permission API enforces those restrictions lets a local user with low privileges modify file metadata—such as permissions, timestamps, and ownership—on paths that should be read-only. This bypasses the intended security boundary and could allow an unprivileged process to alter files it should not be able to touch.

• CVE-2026-48936LOW 3.3

  Node.js 26 contains a weakness in its Permission API that allows a local attacker to start a server using a Unix domain socket even when the `--allow-net` permission flag has not been granted. This bypasses the intended access control mechanism, potentially allowing unauthorized network binding on the local system. The vulnerability requires local access and is limited in scope—it does not expose confidentiality or availability, but it does undermine the integrity of the permission model by enabling unintended network operations.

• CVE-2026-49383LOW 3.3

  CVE-2026-49383 is a low-severity vulnerability in JetBrains IntelliJ IDEA's UI Designer form parser that could allow local attackers to read sensitive information from a user's system. The vulnerability requires user interaction—specifically opening a malicious or compromised form file in the IDE—and affects versions prior to 2026.1. The exposure is limited to information disclosure; the vulnerability does not enable code execution or system modification.

• CVE-2026-49460LOW 3.3

  pypdf, a widely-used open-source Python library for PDF processing, contains a vulnerability in versions prior to 6.12.2 that allows an attacker to craft a malicious PDF file causing the library to consume excessive CPU resources and run for an unusually long time. The attack specifically targets PDFs that use compressed streams with a FlateDecode filter and PNG predictors. While this does not result in data theft or corruption, it can degrade application performance or be used in denial-of-service scenarios where PDF processing is part of a critical workflow.

• CVE-2026-49497LOW 3.3

  Ghidra, the NSA's reverse-engineering framework, contains a path traversal flaw in its debug symbol handler. When you open a malicious ELF binary, Ghidra automatically tries to load debugging information referenced in the binary's .gnu_debuglink section. An attacker can craft that section with path traversal sequences (like "../") to trick Ghidra into checking whether arbitrary files exist on your system and leaking their CRC32 checksums. This is a local attack—the attacker needs you to open a malicious file—but it can reveal information about your system's filesystem structure and contents.

• CVE-2026-56085LOW 3.3

  Dell PowerProtect Data Domain contains a vulnerability in how it initializes system resources. An attacker with local access and basic user privileges can read sensitive information from memory that should have been cleared. This is a localized information disclosure issue with limited scope—the attacker cannot modify data or crash the system.

• CVE-2026-56361LOW 3.3

  ImageMagick versions before 7.1.2-19 contain a subtle off-by-one error in how they validate morphology parameters. This flaw allows attackers to read data from a single memory location just outside the intended buffer boundary. An attacker would need to craft a malicious image or morphology parameter and have a local user open it, potentially leaking a small amount of sensitive information from memory.

• CVE-2026-56362LOW 3.3

  ImageMagick versions before 7.1.2-15 contain a heap buffer overflow vulnerability that occurs when the image processing library reads from memory it hasn't properly allocated. The flaw stems from a sequence of operations where metadata is updated before the pixel cache memory is fully initialized, creating a window where out-of-bounds reads can occur. An attacker with administrative privileges can manipulate memory or disk allocation to trigger this condition, potentially exposing sensitive data from adjacent memory regions.

• CVE-2026-56363LOW 3.3

  ImageMagick versions prior to 7.1.2-22 contain a vulnerability in how they process binomial kernel values used in image filtering operations. When an attacker supplies an unusually large binomial kernel value, the application performs an integer overflow that leads to division by zero, crashing the ImageMagick process. This is a local denial-of-service condition—the attacker needs local access and user interaction to trigger the crash, but the impact is straightforward: service disruption rather than data theft or system compromise.

• CVE-2026-56366LOW 3.3

  ImageMagick versions before 7.1.2-18 have a memory leak flaw in how they process certain image metadata. When an attacker sends a specially crafted APP1JPEG image file, ImageMagick fails to properly release memory, causing it to consume increasing amounts of system resources until the application crashes or becomes unresponsive. This is a local denial-of-service issue that requires user interaction to trigger.

• CVE-2026-56370LOW 3.3

  ImageMagick versions prior to 7.1.2-19 contain a flaw in how they process connected-components image data. When a user or script provides specially crafted connected-components definitions—particularly those with invalid array indices—ImageMagick can attempt to access memory locations outside of allocated buffers. This out-of-bounds access can crash the application (denial of service) or, in certain conditions, potentially allow an attacker to execute arbitrary code. The vulnerability requires local access and user interaction to trigger, making it a lower-risk but still important issue for organizations that process untrusted image data or accept image uploads from external sources.

• CVE-2026-56374LOW 3.3

  ImageMagick versions before 7.1.2-19 have a memory safety bug in how they process FTXT image files. When an attacker crafts a specially designed FTXT file with malformed format data, ImageMagick fails to validate boundaries before reading from memory, potentially crashing the application or exposing data in memory. The vulnerability requires user interaction—someone must open or process the malicious file.

• CVE-2026-56377LOW 3.3

  ImageMagick versions before 7.1.2-24 contain a flaw in how they enforce security policies that govern file operations. An attacker with local access can exploit this weakness to create or truncate files that the security policy should have blocked, potentially allowing them to write data outside the intended sandbox boundaries in conversion services. This is a path traversal issue masked by an incomplete policy validation mechanism.

• CVE-2026-57588LOW 3.3

  A SQL injection flaw exists in Nessus that can be triggered when a privileged user imports a specially crafted scan result file. An attacker could design such a file to inject malicious SQL commands into the scan results database, potentially allowing unauthorized access to sensitive scan data. The vulnerability requires local file access and user interaction, limiting its immediate threat scope, though it could be valuable in targeted attacks against security teams.

• CVE-2026-58052LOW 3.3

  7-Zip for Windows has a flaw in how it handles specially crafted RAR5 archive files. When extracting such an archive, 7-Zip fails to properly maintain the 'Mark-of-the-Web' (MotW)—a security marker that Windows uses to warn users about potentially unsafe files downloaded from the internet. An attacker can exploit this by embedding malicious stream records in a RAR5 file that bypass 7-Zip's protection and allow the tool to extract content that appears safe to Windows, defeating SmartScreen warnings and potentially tricking users into running malware.

• CVE-2026-8662LOW 3.3

  A path traversal flaw exists in Rapid7 InsightConnect's Compression Plugin on Linux systems. An authenticated attacker can supply specially crafted filenames to the create_archive function that cause files to be written outside their intended directory. However, because the attacker cannot control the file contents—only the path—the damage is limited to corrupting legitimate files rather than injecting malicious data. Exploitation requires valid credentials, making this primarily a risk for insider threats or compromised service accounts.

• CVE-2026-46815LOW 3.2

  CVE-2026-46815 is a low-severity information disclosure flaw in Oracle VM VirtualBox 7.2.8 affecting the VMSVGA graphics device driver. A high-privilege user already logged into a system running VirtualBox can read a limited subset of VirtualBox data that should not be accessible to them. The vulnerability does not enable attackers to modify data, crash the application, or gain system-level control. Its scope extends beyond VirtualBox itself—successful exploitation could indirectly expose data relevant to other products running on or managed by the affected host.

• CVE-2026-46816LOW 3.2

  CVE-2026-46816 is a low-severity information disclosure vulnerability affecting Oracle VM VirtualBox 7.2.8. A high-privileged user already logged into the host system running VirtualBox can read a limited subset of data accessible to VirtualBox. The attack requires existing local access with administrative-level privileges and occurs through the VMSVGA device component. Because VirtualBox often runs on infrastructure managing multiple virtual machines, successful exploitation could expose sensitive data across guest systems, though the scope of readable data is restricted.

• CVE-2026-46874LOW 3.2

  Oracle VM VirtualBox version 7.2.8 contains a privilege escalation vulnerability affecting the Core component. An attacker with high-level privileges and local access to the system running VirtualBox can read sensitive data that VirtualBox manages or processes. While the vulnerability itself resides in VirtualBox, successful exploitation could expose information relevant to other systems or virtual machines it hosts, expanding the security footprint beyond VirtualBox alone. The vulnerability requires the attacker to already have administrative or high-privilege access to the infrastructure, which significantly limits the practical attack surface in most environments.

• CVE-2026-46977LOW 3.2

  A vulnerability in Oracle VM VirtualBox's VMSVGA device component could allow a high-privileged user already logged into the host system to read a limited subset of sensitive data stored within VirtualBox. Version 7.2.8 is affected. The issue is classified as low-severity because successful exploitation requires administrative-level access to the infrastructure where VirtualBox runs, and the data exposure is partial rather than complete. The scope impacts extend beyond VirtualBox itself, meaning the unauthorized read could potentially affect other systems or data managed through the hypervisor.

• CVE-2026-49356LOW 3.2

  Babel, a widely used JavaScript compiler, contains a vulnerability that allows attackers to read arbitrary source map files from systems running the compiler. If an attacker can control the JavaScript code being compiled and see the compiler's output, they can craft malicious code containing a sourceMappingURL comment that tricks Babel into exposing source map files they know the path to. This is a low-severity issue because it requires multiple preconditions: the attacker must have input control, output visibility, and knowledge of the exact file path of the target source map.

• CVE-2024-42206LOW 3.1

  HCL iReflection contains third-party components that are vulnerable and outdated, creating a potential integrity risk within the web application. An authenticated user with low privileges could potentially exploit this condition, though the attack surface is constrained by difficult environmental conditions and limited authentication requirements.

• CVE-2025-52608LOW 3.1

  HCL iControl contains a cookie security misconfiguration that leaves session identifiers and authentication tokens vulnerable to interception and cross-site request forgery attacks. The vulnerability stems from the absence of the Secure and SameSite cookie attributes, combined with an overly permissive cookie path set to root. While the immediate risk is moderate, this configuration flaw can enable attackers to hijack user sessions or trick authenticated users into performing unintended actions.

• CVE-2025-52611LOW 3.1

  HCL iControl v4.0.0 contains a vulnerability where the application crashes and exposes internal error messages, including stack traces, when certain code paths are triggered. The underlying cause is a programming error where the application attempts to access a property (the 'dashboard key') from an object that hasn't been properly initialized or is missing entirely. While an attacker would need valid login credentials to trigger this issue, the exposure of stack trace information could help them understand the application's internal structure and identify further attack vectors.

• CVE-2025-62340LOW 3.1

  HCL iControl contains a session management weakness where user sessions are not automatically terminated after periods of inactivity. This means that if a user walks away from an authenticated session without logging out, an attacker with physical or network access could potentially interact with the application under that user's privileges. The vulnerability requires an attacker to have some level of access and be able to reach the application, limiting the immediate threat, but it does increase risk in shared or public-facing deployments.

• CVE-2026-10011LOW 3.1

  A flaw in Chrome's Skia graphics library could allow an attacker who has already compromised Chrome's renderer process to extract sensitive data from websites you visit. The attacker would need to serve you a specially crafted web page to perform the attack. While the underlying issue received a High severity rating from Chromium, the overall exploitability is limited because it requires both renderer compromise and user interaction, making it a low-risk vulnerability in practical terms.

• CVE-2026-10565LOW 3.1

  A race condition vulnerability has been discovered in Open5GS versions up to 2.7.6 that affects the NGAP Handover security mode processing function. The flaw allows an authenticated attacker to trigger a timing-dependent race condition that results in a denial-of-service condition. While a public exploit exists, successful exploitation requires specific conditions and careful timing, making real-world attacks difficult to execute reliably.

• CVE-2026-10654LOW 3.1

  A race condition in Zephyr's Bluetooth RFCOMM stack can cause a session to become permanently stuck when a local disconnect and an incoming peer disconnect happen at the same time. When this occurs, the session slot doesn't get released properly, preventing future Bluetooth connections with that peer. While the timing required to trigger this is difficult to achieve naturally, repeated occurrences could exhaust available RFCOMM session capacity and deny Bluetooth service to users. The vulnerability only affects availability; no data is leaked or corrupted.

• CVE-2026-10705LOW 3.1

  Dask, a Python library for parallel computing and distributed data processing, contains a resource exhaustion vulnerability in its HyperLogLog (approximate distinct count) functionality. An authenticated remote attacker can trigger excessive resource consumption through the nunique_approx function, potentially degrading system availability. The flaw requires significant attack complexity and specific preconditions, making real-world exploitation difficult despite being theoretically possible.

• CVE-2026-11240LOW 3.1

  CVE-2026-11240 is a low-severity input validation flaw in Google Chrome's Loader component that allows a remote attacker to bypass the browser's site isolation security feature, but only if they have already compromised the renderer process. Site isolation is Chrome's defense mechanism that runs each website in a separate process to prevent one compromised site from accessing data from another. An attacker would need to deliver a specially crafted HTML page to exploit this, making it a post-compromise risk rather than a direct remote code execution vector. The vulnerability affects Chrome versions prior to 149.0.7827.53.

• CVE-2026-11244LOW 3.1

  CVE-2026-11244 is a low-severity flaw in Google Chrome's WebAuthentication feature that allows inadequate validation of user-supplied input. An attacker with prior access to Chrome's renderer process—the component responsible for displaying web pages—could craft a malicious HTML page to circumvent the browser's same-origin policy, a fundamental security boundary that prevents scripts from one website accessing data from another. This is not a direct remote code execution and requires both renderer process compromise and user interaction to succeed.

• CVE-2026-11247LOW 3.1

  A flaw in Google Chrome's CustomTabs feature on Android allows an attacker to leak data across website boundaries through a specially crafted webpage. The vulnerability requires user interaction and is difficult to exploit, affecting Android devices running Chrome versions before 149.0.7827.53. While the risk is low, it represents a potential privacy leak in a widely used mobile browser component.

• CVE-2026-11251LOW 3.1

  A flaw in Chrome's password manager allows a sophisticated attacker to read stored password information if they can first compromise Chrome's renderer process through a malicious web page. The vulnerability requires multiple conditions to exploit: the attacker must already control the rendering engine, the user must interact with the page, and the attack surface is limited to sensitive credential disclosure. Chrome versions before 149.0.7827.53 are affected. This is not a zero-click issue and does not allow code execution or system-level access.

• CVE-2026-11464LOW 3.1

  JeecgBoot versions up to 3.9.2 contain a vulnerability in the User List Endpoint that allows authenticated users to disclose sensitive information by manipulating a salt parameter. An attacker with valid credentials can exploit this flaw to access restricted data, though doing so requires specific conditions and technical knowledge. A fix is planned for a future release.

• CVE-2026-11465LOW 3.1

  A logic flaw exists in songquanpeng one-api versions up to 0.6.11-preview.7 that affects the redemption code top-up functionality. An authenticated attacker with specific knowledge of the system could bypass or manipulate the business logic governing how redemption codes are processed, potentially allowing unauthorized credit issuance or redemption manipulation. The attack is complex to execute and requires an active user account, making widespread exploitation unlikely in typical deployments.

• CVE-2026-11502LOW 3.1

  JeecgBoot versions up to 3.9.2 contain an open redirect vulnerability in the third-party login flow. When users are directed to log in via OAuth providers like DingTalk or WeChat, an attacker can manipulate the state parameter to redirect victims to a malicious website after authentication. The attack requires social engineering—tricking a user into clicking a specially crafted login link—and only affects deployments that have enabled third-party login functionality.

• CVE-2026-11675LOW 3.1

  Google Chrome contained a memory reading vulnerability in its Skia graphics library that could allow an attacker to steal sensitive data from other websites. The attacker would first need to compromise Chrome's renderer process—the sandboxed component that handles web page rendering—and then trick a user into visiting a specially crafted webpage. If successful, the flaw could leak cross-origin data, meaning information from a different website than the one the user thought they were visiting. This vulnerability affects Chrome versions prior to 149.0.7827.103 across Windows, macOS, and Linux systems.

• CVE-2026-11684LOW 3.1

  A policy enforcement gap in Google Chrome's Network component allowed attackers who had already compromised Chrome's utility process to steal cross-origin data by serving a specially crafted HTML page. This is a post-compromise attack where the attacker has already gained some level of access to the browser process itself, then exploits this weakness to read data that should be isolated between different websites.

• CVE-2026-11686LOW 3.1

  A flaw in Google Chrome's Dawn graphics library on macOS allows an attacker who has already compromised the browser's renderer process to trick the system into leaking data from other websites. The vulnerability requires the attacker to already have control over the renderer and the user to interact with a malicious webpage, making it a limited but real risk in scenarios where renderer escapes are already being exploited.

• CVE-2026-11691LOW 3.1

  Google Chrome contained a flaw in its New Tab Page that could allow attackers who had already compromised Chrome's renderer process to steal data from websites across different origins. The vulnerability required an attacker to have already broken into the renderer—the sandboxed component that runs web content—and then trick a user into visiting a malicious HTML page. While the Chromium security team rated this High severity internally, the calculated CVSS score is Low (3.1) because the attack requires both prior renderer compromise and user interaction.

• CVE-2026-11880LOW 3.1

  Fluent Forms, a popular WordPress plugin for building forms and managing subscriptions, contains an authorization flaw in versions before 6.2.1. The plugin fails to verify that a logged-in user actually owns a subscription before processing cancellation requests. This means a low-privilege account holder could cancel subscriptions belonging to other users, potentially disrupting their services or causing account manipulation. The vulnerability requires an authenticated attacker—someone with a valid WordPress user account—making it a lower-risk issue than unauthenticated attacks.

• CVE-2026-12017LOW 3.1

  Google Chrome versions before 149.0.7827.115 contain a flaw in how browser extensions are implemented that could allow an attacker who has already compromised Chrome's rendering engine to escape site isolation—the security boundary that prevents malicious websites from accessing data belonging to other websites. The attacker would need to trick a user into viewing a specially crafted webpage, but the core vulnerability requires prior control of the renderer process, which significantly limits real-world attack scope.

• CVE-2026-12032LOW 3.1

  Google Chrome on Android versions prior to 149.0.7827.115 contain a flaw in how the browser handles password-related features that could allow an attacker to bypass site isolation protections. Site isolation is a critical Chrome security feature that prevents malicious websites from accessing data from other sites. To exploit this issue, an attacker would first need to compromise Chrome's rendering engine through another vulnerability, then use a specially crafted webpage to break site isolation. While this requires multiple attack prerequisites, the underlying flaw affects password handling and could expose cross-site data to compromised processes.

• CVE-2026-12458LOW 3.1

  Google Chrome versions before 149.0.7827.155 contain a flaw in how the browser handles password-related features that could allow an attacker to extract sensitive data from websites the user has visited. The vulnerability requires an attacker to trick a user into performing specific interactions—such as clicking or gesturing—on a malicious webpage. When exploited, it may leak information that should remain isolated between different websites, but only within the user's current browsing session. The severity is considered low because it demands active user participation and the exposed data scope is limited.

• CVE-2026-12566LOW 3.1

  CVE-2026-12566 is a vulnerability in bbot's docker_pull module where an attacker positioned between bbot and a Docker registry can intercept and modify the authentication redirect instructions sent by the registry. By tampering with the WWW-Authenticate response header, an attacker could trick bbot into sending authentication tokens to a malicious server under their control instead of the legitimate Docker registry, resulting in credential theft.

• CVE-2026-13483LOW 3.1

  A weakness has been identified in arc53 DocsGPT versions up to 0.18.0 that allows an authenticated attacker to tamper with stored credentials without the system detecting the manipulation. The flaw exists in the credential encryption component and requires moderate technical effort to exploit. While a proof-of-concept has been publicly disclosed, successful attacks depend on the attacker already having login access to the application. The development team has submitted a patch, though it has not yet been formally released.

• CVE-2026-13489LOW 3.1

  A synchronization flaw has been discovered in xiaozhi-esp32 versions up to 2.2.6, affecting the MCP Response Handler component. An authenticated attacker with network access could exploit this weakness to cause minor integrity issues, though the attack requires non-trivial technical skill and is not currently a widespread threat. The vulnerability carries a low CVSS score of 3.1 and is not yet tracked as actively exploited in the wild.

• CVE-2026-13493LOW 3.1

  A vulnerability has been identified in AIDC-AI ComfyUI-Copilot versions up to 2.0.28 that allows an authenticated attacker to improperly manipulate resource identifiers through the Workflow Checkpoint Restore Handler. The flaw resides in the conversation API backend and requires specific conditions to exploit, but an exploit has already been publicly disclosed. This is a low-severity issue affecting only authenticated users with legitimate access to the application.

• CVE-2026-13511LOW 3.1

  VoltAgent versions up to 2.1.17 contain an authorization bypass vulnerability in the Memory REST API. An authenticated attacker can manipulate the conversationId parameter to access memory records they should not have permission to view. The vulnerability requires an existing login and is difficult to exploit in practice, but proof-of-concept code has been released publicly. A patch is under review by the maintainers.

• CVE-2026-13939LOW 3.1

  A flaw in Google Chrome's WebShare feature on Android devices could allow an attacker who has already compromised the browser's rendering engine to trick users into thinking they are interacting with legitimate interface elements when they are actually seeing forged content. The vulnerability requires the attacker to have significant pre-existing access to the browser process and relies on user interaction, making it a limited-impact issue in practice.

• CVE-2026-13944LOW 3.1

  Google Chrome on macOS contains a flaw in how it handles data transfers that could allow an attacker to trick users into exposing information across website boundaries. The vulnerability requires the attacker to convince the user to perform specific interactions with a malicious webpage. This is a low-severity issue affecting Chrome versions before 150.0.7871.47 on Mac systems.

• CVE-2026-13945LOW 3.1

  This vulnerability affects Google Chrome on Linux systems and involves insufficient controls over how browser extensions are validated. An attacker could create a malicious extension that, once installed by a user, could perform UI spoofing—essentially creating fake interface elements that trick users into thinking they're interacting with legitimate Chrome features. The vulnerability requires user action (installing the extension) to be exploited, making it a social engineering vector rather than a direct technical flaw. Google has addressed this issue in Chrome version 150.0.7871.47 and later.

• CVE-2026-13948LOW 3.1

  Google Chrome versions prior to 150.0.7871.47 contain a flaw in how it enforces policies on browser extensions. An attacker could convince a user to install a malicious extension, then use that extension to create fake or spoofed user interface elements—making it appear as though the user is interacting with legitimate Chrome features when they are actually engaging with attacker-controlled content. This is a social engineering attack that relies on initial user action to install the extension.

• CVE-2026-13963LOW 3.1

  A vulnerability in Google Chrome's developer tools allows an attacker to trick users into revealing sensitive information from other websites through a specially crafted web page. The attack requires the user to perform specific interactions with Chrome's UI, such as clicking or dragging elements in a particular way. While the risk is considered low due to these interaction requirements and the limited scope of data exposure, organizations should still apply the patch to eliminate the attack surface.

• CVE-2026-13982LOW 3.1

  Google Chrome's password manager UI can be spoofed by attackers who have already compromised the browser's rendering engine. An attacker who gains control of the renderer process—the component responsible for displaying web content—can craft a malicious HTML page that mimics legitimate Chrome password UI elements, potentially deceiving users into revealing credentials or performing unintended actions. This requires prior compromise of the renderer, which is a significant prerequisite but still represents a real escalation risk once initial access is achieved.

• CVE-2026-14617LOW 3.1

  A case-sensitivity handling flaw exists in NousResearch's hermes-agent streaming reasoning component (versions up to 2026.4.30). The vulnerability allows a low-impact information disclosure when filtering streaming data, but requires both authentication and specific conditions to exploit. The vendor has elected not to patch this issue, determining the maintenance burden outweighs the security benefit given the limited attack surface and high complexity required to trigger it.

• CVE-2026-14621LOW 3.1

  A logic flaw in FederatedAI FATE's OSX Broker component can cause session data to be routed to the wrong recipient when certain queue parameters are manipulated. An authenticated attacker can exploit improper validation of the rollSiteSessionId, dstRole, and dstPartyId arguments to leak sensitive data between federated learning sessions. The vulnerability is difficult to exploit in practice and requires authenticated access, keeping real-world risk relatively contained.

• CVE-2026-14630LOW 3.1

  ForceInjection AI-fundermentals versions 2.0 and 3.0 contain a weakness in how the Memory Recall Handler component stores and retrieves conversation history. The system uses cryptographic hashing methods that fall short of current security standards, potentially allowing an authenticated attacker to bypass session ownership checks and access conversation data they should not be able to retrieve. The vulnerability requires valid user credentials and non-trivial technical effort to exploit, limiting its immediate threat to most deployments.

• CVE-2026-14742LOW 3.1

  LangGraph, a component of LangChain AI, contains a weakness in how it hashes cache keys within its Task Result Cache feature. Versions up to 1.2.4 use a weak hashing algorithm when processing cached task results, potentially allowing an authenticated attacker with network access to infer or predict cache contents. The vulnerability has low severity and requires significant technical effort to exploit, but the risk exists in environments where cache confidentiality is important and attackers have valid credentials.

• CVE-2026-14966LOW 3.1

  BBOT's unarchive module contains a symlink validation bypass in zip and 7z archives. The module is designed to reject archives containing symlinks before extraction, but it fails to detect symlinks when they are encoded with a DOS-attribute prefix before the Unix file mode—a format produced by legacy versions of p7zip. An attacker can craft such an archive and have it extracted during a scan (such as through file download operations), allowing a malicious symlink to be written to the extraction directory. The symlink itself is planted but its target is not accessed, limiting the immediate impact. The vulnerability only affects systems using legacy p7zip builds; current mainline 7-Zip is not vulnerable.

• CVE-2026-14967LOW 3.1

  A path traversal vulnerability in BBOT's GitHub workflows reconnaissance module allows an attacker to manipulate a crafted repository URL to write downloaded artifacts outside the intended output directory. The vulnerability exploits insufficient path sanitization—the module fails to resolve `..` sequences in file paths. However, the impact is constrained: writes are limited to two directory levels above the configured output folder, and the final destination is ultimately determined by the operator's own configuration settings, not attacker choice.

• CVE-2026-28378LOW 3.1

  A Grafana vulnerability allows an administrator in one organization to delete public dashboards belonging to a different organization. The issue stems from insufficient access controls on the dashboard deletion endpoint, which does not properly verify organizational boundaries. An authenticated org admin can exploit this by crafting requests with another organization's dashboard identifiers to remove their dashboards without authorization.

• CVE-2026-3176LOW 3.1

  GitLab Enterprise Edition versions 18.6 through 18.11.5, 19.0.0 through 19.0.2, and 19.1.0 contain a flaw that could allow an authenticated user with limited permissions to view project information they should not have access to. The vulnerability requires specific conditions to be triggered and is rated low-severity, but it does bypass intended authorization controls.

• CVE-2026-35193LOW 3.1

  Django's cache middleware has a flaw that can leak private user data. When Django caches web responses, it's supposed to mark cached data as private (via the `Vary` header) if a request included authentication credentials. In Django 5.2 before version 5.2.15 and 6.0 before version 6.0.6, this protection doesn't work correctly. An attacker can make an unauthenticated request to the same URL a logged-in user visited, and Django may serve the cached private response—revealing sensitive information that should have been protected. Older Django versions (5.0.x, 4.1.x, 3.2.x) haven't been formally evaluated but may have the same problem.

• CVE-2026-3553LOW 3.1

  GitLab Community Edition and Enterprise Edition contain an authorization flaw that allows authenticated users to view confidential issue details they should not have access to. The vulnerability affects versions 12.0 through 18.10.7, 18.11.0 through 18.11.4, and 19.0.0 through 19.0.1. An attacker must already have valid GitLab credentials to exploit it, and success depends on specific configuration or state conditions. While the exposure is limited to reading sensitive issue metadata rather than system compromise, it represents a meaningful confidentiality breach for organizations using affected versions.

• CVE-2026-40963LOW 3.1

  Apache Airflow's UI structure_data endpoint was leaking metadata about linked workflows (DAGs) to users who shouldn't see them. An authenticated user with permission to view one workflow could discover the names and dependency relationships of other workflows they weren't authorized to access. This is a read-only information disclosure—no data modification or system disruption occurs—but it can undermine team isolation in multi-tenant Airflow deployments where workflow topology is considered sensitive.

• CVE-2026-42145LOW 3.1

  Coolify, a self-hosted server and application management platform, contains a vulnerability in its database backup upload feature that allows authenticated users to upload files without proper validation. An attacker with valid credentials could upload oversized or malicious files to disrupt service availability. The flaw affects all versions before 4.0.0-beta.474 and requires user authentication, limiting the immediate threat surface but remaining a denial-of-service risk in environments where insider risk or credential compromise is a concern.

• CVE-2026-42172LOW 3.1

  Coolify is a self-hosted platform for managing servers, applications, and databases. Prior to version 4.0.0-beta.474, API tokens used by Coolify (specifically Sanctum tokens, which handle API authentication) did not automatically expire. This meant that if an attacker obtained a valid token—through a leaked credential, compromised backup, or other disclosure—that token would continue working indefinitely. An administrator would need to manually revoke the token in Coolify's settings to stop access. The vulnerability has been patched in 4.0.0-beta.474 and later.