MISP contains an authorization flaw in the handling of remote pull and feed operations. When a non-internal pull or a feed processes incoming data and encounters an existing sharing group, the application did not verify that the remote server's organization was actually a member of that sharing group before allowing the group's metadata to be updated. The captureRemote mechanism previously stored only a boolean flag indicating whether the link was internal, and the authorization check in captureSGExisting relied solely on that flag. For non-internal pulls and feeds, the check was insufficient: a remote server whose organization was not a member of the target sharing group could still cause the group record to be modified (for example, altering the local flag, the modified timestamp, or other group properties). A feed, which has no associated server organization at all, could likewise trigger an update to an existing sharing group's metadata. The security impact is an integrity violation: an authenticated remote entity with sync or feed privileges can alter sharing group attributes that govern how data is distributed among organizations, potentially changing the local flag or other access-relevant properties of groups it does not belong to. The vulnerability requires an authenticated user with the ability to initiate a pull or feed (a sync user or feed administrator) and a target sharing group that exists on the local instance. No unauthenticated access or user interaction is required.
The Gentlemen ransomware group listed Nvidia as a victim on its data leak website, but Nvidia has found no indication that its own systems or networks were compromised. Nvidia is investigating a possible incident involving a third-party partner but has not confirmed any breach or data exposure related to its infrastructure. The ransomware group uses a double-extortion model and has targeted multiple sectors since mid-2025. The claim by the ransomware group does not establish that Nvidia was breached or that any data was stolen.
CVE-2026-133161 is a vulnerability in pydantic-ai-harness where enabling the strip_ansi option in ToolOutputLimits causes inefficient regular expression processing. This leads to quadratic time complexity when removing ANSI escape sequences from tool output, allowing an attacker to craft large outputs that can block the event loop for extended periods. The issue affects versions 0.4.0 through 0.69.0. A fix is available in version 0.70.0 and later.
OFFIS DCMTK 3.5.4 through 3.7.1 contains an integer overflow in DJCodecEncoder::encodeTrueLossless size checks that allows attackers to trigger a heap out-of-bounds read. Attackers can supply a crafted DICOM file with large Columns, Rows and SamplesPerPixel values that, when compressed with true-lossless JPEG, crashes the application.
OFFIS DCMTK version 3.7.1 has a heap buffer overflow vulnerability in functions handling color-by-plane DICOM images compressed with the true-lossless JPEG codec. The flaw arises from incorrect computation of row offsets using rows instead of columns, allowing crafted images with more rows than columns to corrupt heap memory. This can cause application crashes and potentially enable code execution.
modified eCommerce Shopsoftware through 3.3.0 contains a stored cross-site scripting vulnerability that allows authenticated administrators to inject scripts via the unsanitized products_model field. Attackers with product editing permission can store JavaScript that executes in browsers of other administrators, storefront visitors, and customers viewing carts and order history.
Tiptap @tiptap/core 2.0.0-alpha.0 before 3.30.4 contains a cross-site scripting vulnerability that allows attackers to inject DOM attributes via a __proto__ key passed to mergeAttributes(). Attackers can supply untrusted JSON attribute data whose __proto__ replaces the merged object's prototype, causing DOMSerializer.renderSpec to render injected event handlers that execute JavaScript.
Tiptap @tiptap/core 3.7.0 before 3.30.5 contains a regular expression denial of service vulnerability in parseAttributes and parseShortcodeAttributes that allows attackers to exhaust CPU through quadratic backtracking. Attackers can supply Markdown with long crafted attribute strings parsed by createBlockMarkdownSpec, createAtomBlockMarkdownSpec or createInlineMarkdownSpec to stall browser pages or server processes.
Cozy Cloud cozy-stack versions 1.6.62 through 1.6.68 contain an authorization bypass vulnerability in the shared-drive move handler. This flaw allows authenticated users who own a shared drive to manipulate the move source by specifying another instance's domain and directory ID, enabling unauthorized access to and deletion of files in the victim's file tree. The vulnerability is fixed in version 1.6.69.
WatchAlert versions 2.0.0 through 3.9.20 have an insecure direct object reference vulnerability in the /api/w8t/datasource/dataSourceGet endpoint. Authenticated users with datasource view permission can bypass authorization controls to access other tenants' datasources. This allows exposure of sensitive credentials including basic auth passwords, AWS and Alibaba Cloud keys, and Kubernetes kubeconfigs.
HotGo 2.18.6 through 2.19.0 contains an information disclosure vulnerability that allows authenticated backend users to read other administrators' private letters because the messageIds helper omits the receiver filter. Low-privilege users can call GET /admin/notice/pullMessages, which is exempt from role permission checks via exceptAuth, to retrieve titles, contents and senders of private letters.
The npm package 'requester-lb' contains malicious code that compromises any computer on which it is installed or running. The compromise is severe enough that all secrets and keys stored on the affected system should be rotated immediately from a different, uncompromised machine. Removal of the package alone may not eliminate all malicious software introduced by its installation.
flowintel allows users with editor-level access to author case and task fields, including aggregated tags. When these records are exported as CSV, the field values are written verbatim into the output file.
Because no sanitisation is applied to user-supplied text before CSV serialisation, an attacker can store values that begin with spreadsheet formula triggers (such as =, +, -, @, or their fullwidth Unicode equivalents, or leading control/whitespace characters). When a victim opens the exported CSV in a spreadsheet application (e.g. Microsoft Excel, LibreOffice Calc), the stored text is interpreted and executed as a live formula rather than displayed as literal text.
This can lead to unintended data manipulation within the spreadsheet, potential data exfiltration via formula functions, or execution of downstream commands depending on the spreadsheet application and its configuration.
The vulnerability requires the victim to actively open the CSV file in a spreadsheet application; the flowintel application itself is not directly compromised.
Affected: <4.1.0
MISP contains an authorization flaw in the sharing group capture logic. When a sharing group is re-captured with a newer modified timestamp and an updated roster, the application unconditionally overwrote the type, nationality, sector, and contacts fields of every organisation listed in the roster, including the host organisation. This behaviour was reachable by any user holding the perm_sync role, and additionally by any external server pulled from or any feed fetched, because those operations execute with the privileges of the site administrator who initiated them. As a result, a low-privileged sync user or a malicious/compromised external feed or pull source could silently corrupt organisation records on the receiving MISP instance without any site-admin or internal-sync authorization. The vulnerability does not disclose data or affect availability; the impact is limited to integrity of organisation metadata (type, nationality, sector, contacts). The fix restricts organisation detail updates to internal sync operations (pulls over links flagged as internal, or senders holding perm_sync_internal) and site administrators; all other sync, pull, and feed operations may only match existing organisations by UUID or create new ones, without modifying their stored details.
MISP contains an authorization bypass in its server-to-server proposal synchronization mechanism. Two related flaws were identified:
1. The events/pushProposals endpoint located the target event by UUID without enforcing any access-control check. A sync user with perm_sync could push shadow-attribute proposals to any event on the local instance, including events with restricted distribution (e.g., distribution level 0, visible only to the owning organisation). The stored proposal also carried the sender-supplied event_uuid and event_org_id rather than the local event's actual identity, corrupting the data model.
2. When accepting an edit proposal (old_id != 0), the code resolved the target live attribute by UUID alone (Attribute.uuid = shadow.uuid) without scoping the lookup to the proposal's own event. A malicious sync user could therefore craft a proposal whose uuid matched an attribute in a different event; upon acceptance, the wrong attribute would be modified or deleted. The same unscoped lookup existed in the ShadowAttribute::acceptProposal path.
Preconditions: an authenticated user with sync permission (perm_sync) on a MISP instance that participates in server-to-server synchronisation. The attacker must know or enumerate the UUID of a target event or attribute.
Security impact: unauthorised read access to restricted events, unauthorised insertion of proposals onto events the user cannot see, and unauthorised modification or deletion of attributes belonging to other events. This compromises the confidentiality and integrity of threat-intelligence data across organisational boundaries within a MISP deployment.
In Eclipse Dataspace Components (EDC) – Connector (`org.eclipse.edc:decentralized-claims-*`, `org.eclipse.edc:jwt-verifiable-credentials`, `org.eclipse.edc:ldp-verifiable-credentials`) there's a credential verification bypass in Decentralized Claims Protocol (DCP) presentation handling: a participant can present credentials appearing to be issued by any trusted issuer without holding them.
CVE-2026-123894 is a vulnerability in VMware Tools for Windows involving improper verification of SAML token cryptographic signatures. This flaw allows a signature bypass, potentially enabling unauthorized actions. The issue affects VMware Tools versions 13.x.x.x prior to 13.1.5.0 and 12.3.x prior to 12.5.5. VMware has released fixes in versions 13.1.5.0 and 12.5.5 to address this vulnerability.
CVE-2026-123893 is a local privilege-escalation vulnerability in VMware SDDC Manager and VCF Installer. It affects versions 5.x, 9.0.x.x, and 9.1.x.x prior to fixed releases. The flaw allows an attacker with local access and high privileges to escalate their privileges further. The vulnerability has a CVSS 3.1 base score of 6.7, indicating a medium severity level. Fixed versions are 5.2.5, 9.0.2.0100, and 9.1.1.0200.
CVE-2026-123892 is an authenticated OS command injection vulnerability in VMware SDDC Manager caused by improper sanitization of user-supplied input. It affects versions 5.x, 9.0.x.x, and 9.1.x.x prior to fixed releases. The flaw allows an attacker with authenticated access to execute arbitrary OS commands, potentially leading to full system compromise. VMware has released patches in versions 5.2.5, 9.0.2.0100, and 9.1.1.0200 to address this issue.