CVE-2026-108140 is a path traversal vulnerability in Veeam Backup and Replication that allows an authenticated user with the Restore Operator role, who can control backed-up NAS share contents, to write files to arbitrary paths on the backup server. This flaw affects versions 13.0 up to but not including 13.0.3 and 13.1 up to but not including 13.1.2. The vulnerability is fixed in versions 13.0.3 and 13.1.2.
CVE-2026-105477 is a local privilege escalation vulnerability in Veeam Agent for Linux versions 13.1 up to but not including 13.1.2. It affects only installations managed by a Veeam Backup & Replication instance. The issue is due to an untrusted search path (CWE-426), which could allow an attacker with local access to escalate privileges.
CVE-2026-82984 is an improper access control vulnerability in Nextcloud Server versions 32.0.0 through 33.0.0. When a team is configured as "visible to everyone," the web UI prevents non-members from viewing the team, but the resources API endpoint can still be accessed by any authenticated user. This allows enumeration of the team's internal resources, revealing their existence, names, and identifiers without granting access to the actual contents.
CVE-2026-77167 is a denial-of-service vulnerability in Nextcloud server affecting versions 22.0.0 through 34.0.0. An attacker can abuse the federated share endpoint to write arbitrary rows into the oc_share_external database table, which persist indefinitely due to lack of automatic cleanup. This leads to three distinct denial-of-service conditions that self-renew every 24 hours as long as the malicious rows remain.
CVE-2026-77173 is an improper access control vulnerability in Nextcloud Files Confidential affecting versions 3.3.0 through 5.0.0. It allows a user to create a .docx or .odt file with a '#' character in the filename containing confidential information, such as credit card numbers, without the file being marked as confidential.
CVE-2026-67400 is a low-severity vulnerability in Nextcloud Richdocuments where a missing token type check allows retrieval of additional WOPI configuration information using a user or guest token.
CVE-2026-77171 is an improper access control vulnerability in Nextcloud Collectives version 4.0.0 where a public shared collective link allowed recipients to create pages. This flaw could lead to unauthorized content creation by link recipients.
A vulnerability in Nextcloud Server versions 32.0.0 through 34.0.0 allows app passwords to bypass password confirmation on sensitive endpoints. This improper privilege management issue can lead to unauthorized actions without the usual password verification step.
CVE-2026-87897 is a server-side request forgery (SSRF) vulnerability in Nextcloud Whiteboard versions prior to 1.0.0. It allows a low privileged attacker to access internal Chrome Devtools Protocol APIs and leak JWT authentication tokens of other users. Exploiting this flaw can grant the attacker read/write access to the victim's Whiteboard session. The vulnerability has a medium severity score of 4.6 and was fixed in version 1.0.0.
A vulnerability in the Nextcloud Tables app allows users with read-data access to enumerate all share recipients and their permission levels for a table via a share-listing API endpoint. This exposure does not include the table's data itself, only metadata about sharing. The flaw affects versions from 0.8.1 through 2.0.0 and has a low severity rating.
CVE-2026-82986 is an improper access control vulnerability in Nextcloud Tables versions 0.6.4 through 1.0.0. It allows a user who is a member of a group to bypass group restrictions set by an administrator on link sharing, enabling them to create share links for tables without having the required permission.
CVE-2026-54462 is a resource exhaustion vulnerability in the knative.dev/pkg webhook admissionHandler prior to version 0.0.0-20260531000007-52dbd5ece63f. The webhook handler decodes attacker-supplied AdmissionReview requests without limiting the request body size, allowing a pod-network client to send an oversized POST request that causes memory exhaustion and pod termination. This affects all downstream consumers using the shared webhook server, including Knative, Tekton, KEDA, Sigstore policy-controller, and Kpack. The issue is fixed in version 0.0.0-20260531000007-52dbd5ece63f.
Starting around October 20, 2026, unprotected Cairn AI pentesting harnesses have been targeted by automated campaigns from unknown actors. These attackers insert malicious tasks to manipulate AI agents, leading to host system compromise through profiling, credential theft including AI model API keys, Poseidon malware deployment, and unauthorized SSH key installation. Two distinct threat clusters, AIA0009 and AIA0010, conduct synchronized attacks focusing on persistence/malware deployment and exploration/credential extraction respectively. The campaigns use AI-driven automation to attack AI operations, including container escape attempts and reconnaissance. Poseidon RAT is deployed with Bunny.net CDN-based command and control infrastructure.
This is a credential phishing campaign targeting professionals through fake business bid invitations. It uses password-protected PDF attachments to evade automated security scans. The PDFs contain a fake OneDrive button that redirects victims to a fraudulent document-sharing site, prompting them to enter their email and then credentials on a counterfeit Microsoft login page. The campaign exploits trusted business workflows and familiar brand interfaces to bypass technical and human defenses.
The qs library version 6.7.0 through 6.16.0 has a vulnerability in its stringify function when using the arrayFormat 'comma'. If an array element has a non-callable own toString property, qs.stringify throws an uncaught TypeError. This can be triggered remotely by an unauthenticated attacker controlling input that is later serialized with this option, potentially crashing the Node.js process and causing a denial of service. There is no impact on confidentiality or integrity, and no prototype pollution occurs.
GFI KerioConnect before version 10.0.8 Patch 2 contains a stored cross-site scripting (XSS) vulnerability in its Webmail component. This flaw allows remote attackers to execute JavaScript in an authenticated user's Webmail session by uploading a specially crafted SVG email attachment that bypasses sanitization of event-handler attributes. Exploitation leverages the temporary-download endpoint's weak Content-Security-Policy sandbox, enabling script execution when the attachment is opened. This can lead to theft of user credentials and mailbox contents. The vulnerability is rated medium severity with a CVSS score of 6.2 and has been observed exploited in the wild by Proofpoint Threat Research. A fix is available in version 10.0.8 Patch 2.
The undici WebSocket client, when using HTTP/2 (enabled by default with allowH2: true), crashes the Node.js process if it connects to a malicious server that responds with a failed WebSocket handshake over HTTP/2. This occurs because the client aborts the request too late, leaving an open HTTP/2 stream that emits an unhandled error when the server closes the connection, causing a denial of service. Versions 8.11.0 and 8.11.1 are affected. A fix is available in version 8.11.2. Workarounds include disabling HTTP/2 for WebSocket connections or avoiding untrusted servers until patched.
Authlib versions up to 1.7.2 accept an empty HMAC key for JWS/JWT HS256/HS384/HS512 signature verification, allowing attackers to forge tokens with arbitrary claims. The empty key produces a deterministic digest that anyone can reproduce, bypassing signature verification. This vulnerability affects OAuth 2.0 resource servers and OIDC relying parties using Authlib if the verification key can be empty at runtime. The issue is fixed in Authlib 1.9.1.
A vulnerability in rustls versions 0.23.44, 0.23.45, and certain 0.24.0-dev releases causes buffered application data to be sent to the server before the client sends the mandatory 'ech_required' alert when Encrypted Client Hello (ECH) is rejected. This violates RFC 9849 requirements that no application data be sent before the alert. The issue affects clients using EchMode::Enable that send application data before handshake completion and connect to servers rejecting their ECH config. The flaw was fixed in rustls 0.23.46 and 0.24.0-dev.2.
A vulnerability in rustls allows a network attacker to poison the client's per-server-name key-exchange record by sending an unauthenticated ServerHello selecting the X25519 group. This causes subsequent ClientHellos for that server name to carry only an X25519 share, potentially leading the handshake to complete with classical key exchange only, losing quantum-resistant confidentiality. No authentication bypass or key compromise occurs. The issue is fixed in rustls versions 0.23.46 and 0.24.0-dev.2.