Mitigation bypass in the File Handling component. This vulnerability was fixed in Firefox 150, Firefox ESR 140.10, Thunderbird 150, and Thunderbird 140.10.

A vulnerability was detected in langflow-ai langflow up to 1.8.3. The impacted element is the function get_client_ip/install_mcp_config of the file src/backend/base/langflow/api/v1/mcp_projects.py of the component Model Context Protocol Configuration API. Performing a manipulation of the argument X-Forwarded-For results in injection. The attack may be initiated remotely. The exploit is now public and may be used. The vendor was contacted early about this disclosure but did not respond in any way.

NEMU contains an implementation flaw in its RISC-V Hypervisor CSR handling where henvcfg[7:4] (CBIE/CBCFE/CBZE-related fields) is incorrectly masked/updated based on menvcfg[7:4], so a machine-mode write to menvcfg can implicitly modify the hypervisor's environment configuration. This can lead to incorrect enforcement of virtualization configuration and may cause unexpected traps or denial of service when executing cache-block management instructions in virtualized contexts (V=1).

Protection mechanism failure in Windows Shell allows an unauthorized attacker to bypass a security feature over a network.

Protection mechanism failure in Windows Shell allows an unauthorized attacker to perform spoofing over a network.

October is a Content Management System (CMS) and web platform. Versions prior to 3.7.13 and versions 4.0.0 through 4.1.4 contain a sandbox bypass vulnerability in the optional Twig safe mode feature (CMS_SAFE_MODE). Certain methods on the collect() helper were not properly restricted, allowing authenticated users with template editing permissions to bypass sandbox protections. Exploitation requires authenticated backend access with CMS template editing permissions and only affects installations with CMS_SAFE_MODE enabled (disabled by default). This issue has been fixed in versions 3.7.13 and 4.1.5. To workaround this issue, users can disable CMS_SAFE_MODE if untrusted template editing is not required, and restrict CMS template editing permissions to fully trusted administrators only.

MaxKB is an open-source AI assistant for enterprise. In versions 2.7.1 and below, an authenticated user can bypass sandbox result validation and spoof tool execution results by exploiting Python frame introspection to read the wrapper's UUID from its bytecode constants, then writing a forged result directly to file descriptor 1 (bypassing stdout redirection). By calling sys.exit(0), the attacker terminates the wrapper before it prints the legitimate output, causing the MaxKB service to parse and trust the spoofed response as the genuine tool result. This issue has been fixed in version 2.8.0.

MaxKB is an open-source AI assistant for enterprise. Versions 2.7.1 and below contain a sandbox escape vulnerability in the ToolExecutor component. By leveraging Python's ctypes library to execute raw system calls, an authenticated attacker with workspace privileges can bypass the LD_PRELOAD-based sandbox.so module to achieve arbitrary code execution via direct kernel system calls, enabling full network exfiltration and container compromise. The library intercepts critical standard system functions such as execve, system, connect, and open. It also intercepts mprotect to prevent PROT_EXEC (executable memory) allocations within the sandboxed Python processes, but pkey_mprotect is not blocked. This issue has been fixed in version 2.8.0.

MaxKB is an open-source AI assistant for enterprise. In versions 2.7.1 and below, an incomplete sandbox protection mechanism allows an authenticated user with tool execution privileges to escape the LD_PRELOAD-based sandbox. By env command the attacker can clear the environment variables and drop the sandbox.so hook, leading to unrestricted Remote Code Execution (RCE) and network access. MaxKB restricts untrusted Python code execution via the Tool Debug API by injecting sandbox.so through the LD_PRELOAD environment variable. This intercepts sensitive C library functions (like execve, socket, open) to restrict network and file access. However, a patch allowed the /usr/bin/env utility to be executed by the sandboxed user. When an attacker is permitted to create subprocesses, they can execute the env -i python command. The -i flag instructs env to completely clear all environment variables before running the target program. This effectively drops the LD_PRELOAD environment variable. The newly spawned Python process will therefore execute natively without any sandbox hooks, bypassing all network and file system restrictions. This issue has been fixed in version 2.8.0.

ImageMagick is free and open-source software used for editing and manipulating digital images. Versions below 7.1.2-19 and 6.9.13-44 contain a heap use-after-free vulnerability that can cause a crash when reading and printing values from an invalid XMP profile. This issue has been fixed in versions 6.9.13-44 and 7.1.2-19.

PraisonAI is a multi-agent teams system. Prior to 4.5.128, PraisonAI's AST-based Python sandbox can be bypassed using type.__getattribute__ trampoline, allowing arbitrary code execution when running untrusted agent code. The _execute_code_direct function in praisonaiagents/tools/python_tools.py uses AST filtering to block dangerous Python attributes like __subclasses__, __globals__, and __bases__. However, the filter only checks ast.Attribute nodes, allowing a bypass. The sandbox relies on AST-based filtering of attribute access but fails to account for dynamic attribute resolution via built-in methods such as type.getattribute, resulting in incomplete enforcement of security restrictions. The string '__subclasses__' is an ast.Constant, not an ast.Attribute, so it is never checked against the blocked list. This vulnerability is fixed in 4.5.128.

Policy bypass in ServiceWorkers in Google Chrome prior to 147.0.7727.55 allowed a remote attacker to bypass content security policy via a crafted HTML page. (Chromium security severity: Low)

Policy bypass in IFrameSandbox in Google Chrome prior to 147.0.7727.55 allowed a remote attacker who convinced a user to engage in specific UI gestures to bypass navigation restrictions via a crafted HTML page. (Chromium security severity: Low)

Policy bypass in Downloads in Google Chrome prior to 147.0.7727.55 allowed a remote attacker to bypass of multi-download protections via a crafted HTML page. (Chromium security severity: Low)

Policy bypass in Audio in Google Chrome prior to 147.0.7727.55 allowed a remote attacker who convinced a user to engage in specific UI gestures to bypass sandbox download restrictions via a crafted HTML page. (Chromium security severity: Low)

PraisonAI is a multi-agent teams system. Prior to 1.5.115, execute_code() in praisonaiagents.tools.python_tools defaults to sandbox_mode="sandbox", which runs user code in a subprocess wrapped with a restricted __builtins__ dict and an AST-based blocklist. The AST blocklist embedded inside the subprocess wrapper (blocked_attrs of python_tools.py) contains only 11 attribute names — a strict subset of the 30+ names blocked in the direct-execution path. The four attributes that form a frame-traversal chain out of the sandbox are all absent from the subprocess list (__traceback__, tb_frame, f_back, and f_builtins). Chaining these attributes through a caught exception exposes the real Python builtins dict of the subprocess wrapper frame, from which exec can be retrieved and called under a non-blocked variable name — bypassing every remaining security layer. This vulnerability is fixed in 1.5.115.

Directus is a real-time API and App dashboard for managing SQL database content. Prior to 11.17.0, Directus's Single Sign-On (SSO) login pages lacked a Cross-Origin-Opener-Policy (COOP) HTTP response header. Without this header, a malicious cross-origin window that opens the Directus login page retains the ability to access and manipulate the window object of that page. An attacker can exploit this to intercept and redirect the OAuth authorization flow to an attacker-controlled OAuth client, causing the victim to unknowingly grant access to their authentication provider account (e.g. Google, Discord). This vulnerability is fixed in 11.17.0.

SandboxJS is a JavaScript sandboxing library. Prior to 0.8.36, SandboxJS blocks direct assignment to global objects (for example Math.random = ...), but this protection can be bypassed through an exposed callable constructor path: this.constructor.call(target, attackerObject). Because this.constructor resolves to the internal SandboxGlobal function and Function.prototype.call is allowed, attacker code can write arbitrary properties into host global objects and persist those mutations across sandbox instances in the same process. This vulnerability is fixed in 0.8.36.

A vulnerability was determined in Campcodes Complete POS Management and Inventory System up to 4.0.6. This affects an unknown function of the file app/Http/Controllers/SettingsController.php of the component Environment Variable Handler. Executing a manipulation can lead to injection. It is possible to launch the attack remotely. The exploit has been publicly disclosed and may be utilized.

PraisonAI is a multi-agent teams system. Prior to version 1.5.90, execute_code() in praisonai-agents runs attacker-controlled Python inside a three-layer sandbox that can be fully bypassed by passing a str subclass with an overridden startswith() method to the _safe_getattr wrapper, achieving arbitrary OS command execution on the host. This issue has been patched in version 1.5.90.

Cr*nMaster (cronmaster) is a Cronjob management UI with human readable syntax, live logging and log history for cronjobs. Prior to version 2.2.0, an authentication bypass in middleware allows unauthenticated requests with an invalid session cookie to be treated as authenticated when the middleware’s session-validation fetch fails. This can result in unauthorized access to protected pages and unauthorized execution of privileged Next.js Server Actions. This issue has been patched in version 2.2.0.

Insufficient policy enforcement in WebUSB in Google Chrome prior to 146.0.7680.178 allowed a remote attacker to obtain potentially sensitive information from process memory via a crafted HTML page. (Chromium security severity: High)

Business::OnlinePayment::StoredTransaction versions through 0.01 for Perl uses an insecure secret key. Business::OnlinePayment::StoredTransaction generates a secret key by using a MD5 hash of a single call to the built-in rand function, which is unsuitable for cryptographic use. This key is intended for encrypting credit card transaction data.

A vulnerability has been found in PromtEngineer localGPT up to 4d41c7d1713b16b216d8e062e51a5dd88b20b054. The impacted element is the function _route_using_overviews of the file backend/server.py of the component LLM Prompt Handler. Such manipulation leads to injection. The attack may be performed from remote. The exploit has been disclosed to the public and may be used. This product utilizes a rolling release system for continuous delivery, and as such, version information for affected or updated releases is not disclosed. The vendor was contacted early about this disclosure but did not respond in any way.

vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.18.0, two model implementation files hardcode `trust_remote_code=True` when loading sub-components, bypassing the user's explicit `--trust-remote-code=False` security opt-out. This enables remote code execution via malicious model repositories even when the user has explicitly disabled remote code trust. Version 0.18.0 patches the issue.

PinchTab is a standalone HTTP server that gives AI agents direct control over a Chrome browser. PinchTab `v0.8.3` through `v0.8.5` allow arbitrary JavaScript execution through `POST /wait` and `POST /tabs/{id}/wait` when the request uses `fn` mode, even if `security.allowEvaluate` is disabled. `POST /evaluate` correctly enforces the `security.allowEvaluate` guard, which is disabled by default. However, in the affected releases, `POST /wait` accepted a user-controlled `fn` expression, embedded it directly into executable JavaScript, and evaluated it in the browser context without checking the same policy. This is a security-policy bypass rather than a separate authentication bypass. Exploitation still requires authenticated API access, but a caller with the server token can execute arbitrary JavaScript in a tab context even when the operator explicitly disabled JavaScript evaluation. The current worktree fixes this by applying the same policy boundary to `fn` mode in `/wait` that already exists on `/evaluate`, while preserving the non-code wait modes. As of time of publication, a patched version is not yet available.

OneUptime is an open-source monitoring and observability platform. Prior to version 10.0.35, a low-privileged authenticated user (ProjectMember) can achieve remote command execution on the Probe container/host by abusing Synthetic Monitor Playwright script execution. Synthetic monitor code is executed in VMRunner.runCodeInNodeVM with a live Playwright page object in context. The sandbox relies on a denylist of blocked properties/methods, but it is incomplete. Specifically, _browserType and launchServer are not blocked, so attacker code can traverse `page.context().browser()._browserType.launchServer(...)` and spawn arbitrary processes. Version 10.0.35 contains a patch.

An access issue was addressed with additional sandbox restrictions. This issue is fixed in macOS Sequoia 15.7.5, macOS Sonoma 14.8.5, macOS Tahoe 26.4. An app may be able to connect to a network share without user consent.

This issue was addressed through improved state management. This issue is fixed in Safari 26.4, iOS 18.7.7 and iPadOS 18.7.7, iOS 26.4 and iPadOS 26.4, macOS Tahoe 26.4, tvOS 26.4, visionOS 26.4, watchOS 26.4. Processing maliciously crafted web content may prevent Content Security Policy from being enforced.

A vulnerability was found in Foundation Agents MetaGPT up to 0.8.1. This vulnerability affects unknown code of the file metagpt/actions/di/write_analysis_code.py of the component DataInterpreter. The manipulation results in injection. It is possible to launch the attack remotely. The exploit has been made public and could be used. The vendor was contacted early about this disclosure but did not respond in any way.

A security vulnerability has been detected in vanna-ai vanna up to 2.0.2. Affected is the function exec of the file /src/vanna/legacy. Such manipulation leads to injection. The attack can be executed remotely. The exploit has been disclosed publicly and may be used. The vendor was contacted early about this disclosure but did not respond in any way.

A vulnerability was identified in bagofwords1 bagofwords up to 0.0.297. This impacts the function generate_df of the file backend/app/ai/code_execution/code_execution.py. Such manipulation leads to injection. The attack may be launched remotely. The exploit is publicly available and might be used. Upgrading to version 0.0.298 will fix this issue. The name of the patch is 47b20bcda31264635faff7f6b1c8095abe1861c6. It is recommended to upgrade the affected component.

Harden-Runner is a CI/CD security agent that works like an EDR for GitHub Actions runners. In versions 2.15.1 and below, a DNS over HTTPS (DoH) vulnerability allows attackers to bypass egress-policy: block network restrictions by tunneling exfiltrated data through permitted HTTPS endpoints like dns.google. The attack works by encoding sensitive data (e.g., the runner's hostname) as subdomains in DoH queries, which appear as legitimate HTTPS traffic to Harden-Runner's domain-based filtering but are ultimately forwarded to an attacker-controlled domain. This effectively enables data exfiltration without directly connecting to any blocked destination. Exploitation requires the attacker to already have code execution within the GitHub Actions workflow. The issue was fixed in version 2.16.0.

Harden-Runner is a CI/CD security agent that works like an EDR for GitHub Actions runners. In versions 2.15.1 and below, the Harden-Runner that allows bypass of the egress-policy: block network restriction using DNS queries over TCP. Egress policies are enforced on GitHub runners by filtering outbound connections at the network layer. When egress-policy: block is enabled with a restrictive allowed-endpoints list (e.g., only github.com:443), all non-compliant traffic should be denied. However, DNS queries over TCP, commonly used for large responses or fallback from UDP, are not adequately restricted. Tools like dig can explicitly initiate TCP-based DNS queries (+tcp flag) without being blocked. This vulnerability requires the attacker to already have code execution capabilities within the GitHub Actions workflow. The issue has been fixed in version 2.16.0.

Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. In versions up to and including 1.20.1, a security control bypass exists in onnx.hub.load() due to improper logic in the repository trust verification mechanism. While the function is designed to warn users when loading models from non-official sources, the use of the silent=True parameter completely suppresses all security warnings and confirmation prompts. This vulnerability transforms a standard model-loading function into a vector for Zero-Interaction Supply-Chain Attacks. When chained with file-system vulnerabilities, an attacker can silently exfiltrate sensitive files (SSH keys, cloud credentials) from the victim's machine the moment the model is loaded. As of time of publication, no known patched versions are available.

HCL AION is affected by a vulnerability where untrusted file parsing operations are not executed within a properly isolated sandbox environment. This may expose the application to potential security risks, including unintended behaviour or integrity impact when processing specially crafted files.

A vulnerability allowing an authenticated domain user to bypass restrictions and manipulate arbitrary files on a Backup Repository.

A vulnerability allowing an authenticated domain user to perform remote code execution (RCE) on the Backup Server.

A vulnerability allowing an authenticated user with the Backup Administrator role to perform remote code execution (RCE) in high availability (HA) deployments of Veeam Backup & Replication.

A weakness has been identified in CodeGenieApp serverless-express up to 4.17.1. This affects an unknown part of the file utils/dynamodb.ts of the component Users Endpoint. This manipulation of the argument filter causes injection. The attack may be initiated remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.

A security vulnerability has been detected in whyour qinglong up to 2.20.1. Affected is an unknown function of the file back/loaders/express.ts of the component API Interface. The manipulation of the argument command leads to protection mechanism failure. The attack may be initiated remotely. The exploit has been disclosed publicly and may be used. Upgrading to version 2.20.2 is able to address this issue. The identifier of the patch is 6bec52dca158481258315ba0fc2f11206df7b719. It is advisable to upgrade the affected component. The code maintainer was informed beforehand about the issues. He reacted very fast and highly professional.

In oobconfig, there is a possible bypass of carrier restrictions due to a logic error. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.

Parse Server is an open source backend that can be deployed to any infrastructure that can run Node.js. Prior to 8.6.12 and 9.5.1-alpha.1, the requestKeywordDenylist security control can be bypassed by placing any nested object or array before a prohibited keyword in the request payload. This is caused by a logic bug that stops scanning sibling keys after encountering the first nested value. Any custom requestKeywordDenylist entries configured by the developer are equally by-passable using the same technique. All Parse Server deployments are affected. The requestKeywordDenylist is enabled by default. This vulnerability is fixed in 8.6.12 and 9.5.1-alpha.1. Use a Cloud Code beforeSave trigger to validate incoming data for prohibited keywords across all classes.

A vulnerability was identified in opencc JFlow up to 5badc00db382d7cb82dad231e6a866b18e0addfe. Affected by this vulnerability is the function Calculate of the file src/main/java/bp/wf/httphandler/WF_CCForm.java. Such manipulation leads to injection. The attack may be performed from remote. The exploit is publicly available and might be used. This product utilizes a rolling release system for continuous delivery, and as such, version information for affected or updated releases is not disclosed. The project was informed of the problem early through an issue report but has not responded yet.

Inappropriate user token revocation due to a logic error in the token revocation endpoint implementation in Cloudfoundry UAA v77.30.0 to v78.7.0 and in Cloudfoundry Deployment v48.7.0 to v54.10.0.

An authentication bypass vulnerability in the authorization mechanism of Weintek cMT-3072XH2 easyweb v2.1.53, OS v20231011 allows unauthorized attackers to perform Administrative actions using service accounts.

In broadcastIntentLockedTraced of BroadcastController.java, there is a possible way to launch arbitrary activities from the background on the paired companion phone due to a logic error in the code. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.

In onChange of BiometricService.java, there is a possible way to enable fingerprint unlock due to a logic error in the code. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.

In setHideSensitive of ExpandableNotificationRow.java, there is a possible contact name leak due due to a logic error in the code. This could lead to local information disclosure with no additional execution privileges needed. User interaction is not needed for exploitation.

In enableSystemPackageLPw of Settings.java, there is a possible way to prevent location access from working due to a logic error in the code. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.

In loadDataAndPostValue of multiple files, there is a possible way to obscure permission usage due to a logic error in the code. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation.