FlowIntel versions prior to 4.1.0 contain an authorization bypass vulnerability in the case serialization logic. Authenticated users with access to any case linked to private cases can view metadata (ID, title, description) of those private cases without proper permissions. This exposure occurs across multiple access paths including browser views and REST API endpoints, bypassing the intended per-case access control model.
FlowIntel versions prior to 4.1.0 have a missing authorization vulnerability in the task file deletion functionality. Authenticated users with basic case membership roles (e.g., editor or queuer) can delete evidence files associated with privileged tasks in restricted states, which should be limited to Admin, Case Admin, or Queue Admin roles. This flaw affects the core deletion logic, browser interface, and REST API, allowing unauthorized deletion of files, timeline corruption, and case history alteration.
FlowIntel prior to version 4.1.0 contains an authorization bypass vulnerability that allows non-approver case members to modify tasks that should be restricted and locked pending approval. This includes operations such as self-unassignment, reordering tasks, linking/unlinking MISP objects and attributes, creating or deleting MISP objects or attributes linked to restricted tasks, and executing modules on these tasks. The flaw bypasses the required approver permission checks, undermining the intended approval workflow.
A vulnerability exists in Sunnet CTMS up to version 10.11 in the External Training Application Form Feature. It involves improper control of resource identifiers due to manipulation of the 'memberids' argument in the /cltcms/api/external-training/ext-course file. The flaw allows remote exploitation. The vendor has not responded to the disclosure. The vulnerability has a CVSS score of 6.3, indicating medium severity.
flowintel prior to version 4.1.0 contains an authorization bypass vulnerability in its task completion workflow for privileged cases. Users with lower-privilege roles such as editor or queuer could mark tasks as completed without the required approver review from Admin, Case Admin, or Queue Admin roles. This bypass undermines the mandatory approval workflow and allows unauthorized modification of task completion status. The vulnerability can be exploited through multiple methods including the browser UI, REST API, changing task status to "Finished", creating tasks with the completed flag set, and using UUID aliases or Matrix delegation.
A vulnerability exists in shader-slang slangpy up to version 0.43.1 in the tinyexr::DecodePixelData function within the EXR Image Loader component. This flaw allows an attacker to cause an out-of-bounds read, potentially leading to information disclosure or application instability. The vulnerability can be exploited remotely and requires user interaction. It is rated as medium severity with a CVSS score of 6.3.
Postgres Enterprise Manager (PEM) versions prior to 10.6.1 do not validate webhook URLs, allowing users with alert configuration privileges to specify file:// URLs. When such a webhook is triggered, the PEM agent, running as root by default, reads the referenced local file and stores its contents in the PEM database's webhook error record, accessible to any PEM user. Starting from PEM 10.6.1, webhook URLs are restricted to http and https schemes.
In Postgres Enterprise Manager (PEM), any PEM user could previously invoke the pem.send_webhook() function, which operates with elevated privileges to queue payloads for configured webhooks. This allowed users to send forged notifications using stored authentication tokens or certificates of those webhooks. Starting from PEM version 10.6.1, direct calls to this function by PEM login roles are blocked, and webhook notifications are sent only through internal PEM mechanisms.
A weakness in SqueezeAILab SqueezeLLM's CUDA Matvec Kernel component allows remote attackers to cause an out-of-bounds write via manipulation of the full_row_indices argument in the vecquant4matmul_spmv_hybrid_nuq_perchannel function. The product does not use versioning, so affected versions cannot be precisely identified. Public exploit code is available. The vendor has not yet responded or provided a fix.
A vulnerability in Postgres Enterprise Manager (PEM) allows users with the SQL Profiler role to read and overwrite files outside their designated storage directory via the SQL Profiler save and load functions. This can lead to overwriting the PEM web application's configuration database, causing a denial of service for all users. Starting from PEM version 10.6.1, this issue is mitigated by restricting file operations to the user's storage directory.
A vulnerability in Sonos Tract up to version 0.23.4 allows remote attackers to cause an out-of-bounds read via manipulation of inputs/outputs in the InferenceRulesOp::rules function in the onnx/src/ops/math/gemm.rs file. This issue is fixed in version 0.23.5. The vulnerability has a low severity score and does not impact confidentiality or integrity but may affect availability.
In Postgres Enterprise Manager (PEM) versions prior to 10.6.1, the function pem.deploy_job_template() does not verify that the caller has access rights to the servers, groups, and agents targeted for job template deployment. This allows a user with the scheduled tasks role (pem_manage_schedule_task) to deploy and schedule SQL or shell job steps on agents belonging to other teams, bypassing intended access controls. Starting with PEM 10.6.1, access checks are enforced for every server, group, and agent involved in deployment.
Postgres Enterprise Manager (PEM) versions prior to 10.6.1 store encrypted passwords for PEM agents and the PEM backend superuser using a default encryption key common to all installations. This allows any PEM user to decrypt and read these passwords. Starting with PEM 10.6.1, this issue is resolved by preventing PEM users from reading these passwords.
Poste.io contains a critical remote code execution vulnerability in its DMARC report processing. Emails sent to dmarc-reports@<domain> are saved without verifying the sender. A root-level script unpacks attachments using unsafe shell commands, allowing command injection via crafted attachment filenames or archive contents. An unauthenticated remote attacker able to send email to the server's port 25 can execute arbitrary commands as root inside the Poste.io container. This vulnerability is fixed in version 2.5.22.
CVE-2026-14915 is a vulnerability in Tegsoft İletişim Merkezi that allows unauthorized exposure of sensitive system information via a query system. It affects versions before latest-202611161510. The vulnerability has a medium severity with a CVSS score of 5.3, indicating that it can be exploited remotely without privileges or user interaction to disclose information.
A path traversal vulnerability was discovered in the workspace directory handling and Git integration components of Google Cloud Dataform, BigQuery Studio, and Colab Enterprise. This flaw allows an authenticated remote attacker to read private files and create arbitrary files across tenant repositories by exploiting crafted workspace operations and Git integration requests. The vulnerability was patched on October 10, 2026, and no customer action is required.
In Postgres Enterprise Manager (PEM), the pem_agent role inherited by every PEM agent login was a member of the pg_read_server_files and pg_write_server_files roles on the PEM backend database. This allowed an attacker controlling any monitored host and its PEM agent certificate to read and write files on the PEM server host as the postgres OS user, including the PEM certificate authority key. This vulnerability enables the attacker to issue certificates for other agents and impersonate them. Starting from PEM version 10.6.1, agents no longer hold these roles, mitigating the issue.
The BuddyBoss Addons plugin for WordPress versions up to and including 1.2.1 contains an authentication bypass vulnerability. This flaw allows unauthenticated attackers who know a user's email address to link their own social login identity to that user's account, including administrator accounts, enabling account takeover. Exploitation requires the BuddyBoss Platform plugin with the Social Login module and a selected provider enabled. On sites using the BuddyBoss App JWT component, the vulnerability can yield an application token immediately.
Postgres Enterprise Manager (PEM) versions prior to 10.6.1 have a row-level security policy flaw on pem.job and pem.jobstep tables that fails to restrict PEM agents to their own jobs. An attacker controlling one monitored host and its PEM agent certificate can manipulate jobs across all agents, including the PEM server agent. This can lead to execution of SQL and shell job steps with elevated database credentials, potentially granting superuser access to all monitored databases and the PEM backend. Starting from PEM 10.6.1, agents are prevented from creating or modifying jobs, mitigating this issue.
A vulnerability in erupts erupt up to version 2.0.2 allows remote attackers to cause uncontrolled memory allocation by manipulating the 'height' argument in the UPMS CAPTCHA Endpoint. This issue affects the function EruptUserController.createCode and can lead to denial of service due to resource exhaustion. The vulnerability is fixed in version 2.0.3.