🚨 CVE-2026-45363
ruby-jwt is a Ruby implementation of the RFC 7519 OAuth JSON Web Token standard. Prior to 2.10.3 and 3.2.0, JWT.decode(token, '', true, algorithm: 'HS256') accepts an attacker-forged token because OpenSSL::HMAC.digest('SHA256', '', payload) returns a valid digest under an empty key and no empty-key precondition exists in the HMAC algorithm. The same path is reached when a keyfinder block or key_finder: argument returns an empty string, nil, or an array containing nil for an unknown key, affecting HS256, HS384, and HS512 verification through JWT.decode and JWT::EncodedToken#verify_signature!. This issue is fixed in versions 2.10.3 and 3.2.0.

🎖@cveNotify

🚨 CVE-2026-78030
DBI versions before 1.653 for Perl load arbitrary modules via unvalidated dbm_type and dbm_mldbm attributes in DBD::DBM.

DBD::DBM passes the dbm_type and dbm_mldbm connect attributes to require without checking that the value names a module. require treats a path-shaped string as a literal filename and does not consult @INC, so the attribute chooses the file that Perl loads and runs.

The MLDBM::Serializer:: prefix that DBD::DBM prepends to dbm_mldbm is not a boundary: only the :: separators are rewritten to /, so a value containing / traverses out of the serializer directory. The value is also assigned to $MLDBM::Serializer, which MLDBM requires the same way when it ties the table.

A caller that lets an untrusted party influence either attribute, for example through a DSN fragment or a parameter that selects a storage backend, runs the file-scope code of whatever module the value names.

For example,

my $dsn = "dbi:DBM:f_dir=/var/db;dbm_type=../../Untrusted.pm"
my $dbh = DBI->connect( $dsn );

Note that DBD::Gofer forwards connect attributes to the server side, and DBI::ProxyServer checks only that a DSN starts with a driver prefix.

🎖@cveNotify

🚨 CVE-2026-82560
Pod::Text versions before 6.1.1 for Perl allow CPU and memory exhaustion formatting a POD document whose =over nesting drives the margin to the output width.

Each =over adds its indent to the margin, which wrap() subtracts from the output width to get the space available for text. When that space reaches zero, the line-splitting substitution matches the empty string, and the loop consumes no input while appending the margin padding on every pass.

Formatting an attacker-supplied POD document never returns, and the output grows until memory is exhausted.

🎖@cveNotify

🚨 CVE-2026-93954
A security vulnerability has been detected in grimmory-tools grimmory up to 3.3.3/3.4.1. Affected is the function AppSettingController.getAppSettings of the file backend/src/main/java/org/booklore/controller/AppSettingController.java of the component Settings API Endpoint. Such manipulation leads to incorrect authorization. The attack can be launched remotely. The exploit has been disclosed publicly and may be used. The name of the patch is 2b66ca6df8110f6b512e030b54c16b9fbe318f17. Applying a patch is advised to resolve this issue. PR #2558, merged as 53abc8b, moved the OIDC secret into a dedicated setting, but did not by itself restrict GET /api/v1/settings.

🎖@cveNotify

🚨 CVE-2026-93955
A vulnerability was detected in grimmory-tools grimmory up to 3.3.3/3.4.1. Affected by this vulnerability is the function streamFileToResponse of the file backend/src/main/java/org/booklore/controller/KoboController.java of the component Download Endpoint. Performing a manipulation of the argument bookId results in authorization bypass. The attack may be initiated remotely. The exploit is now public and may be used. Issue #2431 is closed as completed, but its only comment states that the issue “has already been reported elsewhere.” No fixing commit or pull request is identified there.

🎖@cveNotify

🚨 CVE-2026-93956
A flaw has been found in olivier-ls PHP-FTS up to 1.1.2. Affected by this issue is the function SearchEngine::buildHighlights of the file src/SearchEngine.php of the component Search Engine. Executing a manipulation of the argument Query can lead to cross site scripting. The attack may be launched remotely. The exploit has been published and may be used. Upgrading to version 1.1.3 can resolve this issue. This patch is called 0b2fae333d6b022da7ed4c43e2d41aa03f91dff3. The affected component should be upgraded. The vendor was contacted early, responded in a very professional manner and quickly released a fixed version of the affected product.

🎖@cveNotify

🚨 CVE-2026-93988
QloApps through 1.7.0 contains a path traversal vulnerability in the getEmailHTML action of admin/ajax.php that allows authenticated back-office users to read arbitrary files. Attackers can supply relative path sequences in the email parameter to bypass directory restrictions and access sensitive files including database credentials and configuration data.

🎖@cveNotify

🚨 CVE-2026-93989
vLLM through 0.29.0 fails to properly validate bad_words token indices against the model's generation output width in SamplingParams.update_from_tokenizer(). Attackers can supply out-of-bounds token indices that corrupt logits memory of concurrent requests, causing different in-flight HTTP requests to return incorrect tokens.

🎖@cveNotify

🚨 CVE-2026-93990
Expat through 2.8.4 fails to validate low surrogates following high surrogates in UTF-16 input, allowing malformed UTF-16 sequences to be accepted. Attackers can craft UTF-16 encoded XML with lone high surrogates that consume following code units, hiding markup characters from the parser and enabling XML injection attacks.

🎖@cveNotify

🚨 CVE-2026-93991
Argo Workflows versions 4.1.0 through 4.1.3 contain an authorization bypass vulnerability in ListArchivedWorkflows that fails to apply cluster-scoped access review when the metadata.namespace field selector uses the NotEquals operator. Attackers with namespace-scoped list permissions can use a negated namespace field selector to retrieve archived workflows from all other namespaces, exposing spec arguments, parameter values, and annotations.

🎖@cveNotify

🚨 CVE-2026-93992
Gopeed through 2.0.0-beta.3 contains a path traversal vulnerability in archive extraction that allows attackers to write arbitrary files outside the extraction directory. Attackers can craft malicious archives with entries containing directory traversal sequences that bypass validation, enabling file write operations when users download and extract archives with AutoExtract enabled.

🎖@cveNotify

🚨 CVE-2026-93993
Mistral Vibe before 2.25.5 contains a remote code execution vulnerability in the worktree creation process that executes git hooks before trust validation. Attackers can supply a repository with a crafted post-checkout hook that executes arbitrary shell commands with the privileges of the user running Vibe.

🎖@cveNotify

🚨 CVE-2026-94054
Exim before 4.100.1, when Proxy-Protocol is used with an attacker-controlled proxy, has an out-of-bounds write.

🎖@cveNotify

🚨 CVE-2026-94055
Exim before 4.100.1, when certain non-default TLS settings are used with GnuTLS, has a use-after-free.

🎖@cveNotify

🚨 CVE-2026-94056
Exim before 4.100.1, when Proxy-Protocol is used with an attacker-controlled proxy, allows attackers to read certain uninitialized data from stack memory.

🎖@cveNotify

🚨 CVE-2026-94057
Exim before 4.100.1 allows SMTP smuggling in which the received message does not match any sent message, and instead depends on crafted data sent after a rejection during DATA processing.

🎖@cveNotify

🚨 CVE-2026-86754
Snipe-IT before 8.7.0 fails to properly gate Laravel Passport's OAuth client management routes, allowing any authenticated user to register OAuth clients with attacker-controlled redirect URIs. Attackers can trick administrators into approving consent screens, then exchange authorization codes for bearer tokens inheriting full admin API permissions lasting up to 40 years.

🎖@cveNotify

🚨 CVE-2026-86759
Snipe-IT versions before 8.7.0 fail to authorize the POST /hardware/history endpoint, allowing any authenticated user to reassign arbitrary assets and modify audit logs. Attackers can submit a CSV file to reassign assets across companies and inject fraudulent audit trail entries, compromising inventory integrity and accountability.

🎖@cveNotify

🚨 CVE-2026-86764
Snipe-IT through 8.6.4 (fixed in 8.7.0) does not enforce the components.view permission on the authenticated endpoint GET /api/v1/hardware/<asset-id>/assigned/components. The endpoint authorizes only assets.view on the parent asset before returning linked component details; the components.view check is applied only to the response's available_actions.view flag and not to the returned data. As a result, an authenticated user holding only assets.view can enumerate component IDs, names, assigned quantities, and notes that are otherwise protected — the direct GET /api/v1/components/<id> endpoint correctly returns 403 Forbidden for such users.

🎖@cveNotify

🚨 CVE-2026-86769
Snipe-IT versions before 8.7.0 contain an improper ownership management vulnerability in the consumables checkout API endpoint that records the checkout target user's id in the created_by column instead of the authenticated caller's id. Authenticated attackers with consumables.checkout permission can perform checkouts that result in misattributed audit trail entries in the consumables_users pivot table, obscuring which operator performed the action.

🎖@cveNotify

🚨 CVE-2026-86774
Snipe-IT versions before 8.7.0 contain a broken access control vulnerability in AssetModelPolicy where the files() method cascades from assets.files permission, allowing authenticated users to upload and delete file attachments on Asset Model records without the required models.files permission. Attackers with only assets.files permission can mutate shared model file attachments across company boundaries and bypass the dedicated models.files permission intended to restrict file management on the shared Asset Model catalog.

🎖@cveNotify