🚨 CVE-2026-33625
LMDeploy is a toolkit for compressing, deploying, and serving large language models. Versions 012.1 through 0.12.2 contain a code injection vulnerability in `lmdeploy/pytorch/config.py` line 620 that allows an attacker to execute arbitrary Python code by publishing a malicious HuggingFace model with a crafted `quantization_config.quant_dtype` value. When a user loads the model with lmdeploy, the `quant_dtype` is passed to `eval(f'torch.{quant_dtype}')` without any validation. Version 0.12.3 contains a patch.

🎖@cveNotify

🚨 CVE-2026-77386
Kyoo is a self-hosted media server focused on movies, series, and anime. Prior to 5.1.0, an unauthenticated attacker could initiate the OIDC login flow with an attacker-controlled redirectUrl. The login handling in auth/oidc.go stored that URL with the opaque login state, and /auth/oidc/logged/{provider} appended the provider, token, and error values before redirecting the victim's browser without validating the destination. Because the one-use token was not bound to the browser session that initiated login, an attacker who induced a victim to complete OIDC authentication could capture the token at the attacker-controlled destination and exchange it through /auth/oidc/callback/{provider} for the victim's Kyoo session. This issue is fixed in version 5.1.0.

🎖@cveNotify

🚨 CVE-2026-81180
SysReptor is a fully customizable pentest reporting platform. Prior to 2026.61, authenticated users of SysReptor Professional can upload image files whose formats cause image processing to invoke Ghostscript, allowing embedded PostScript to operate in the shared temporary directory. An attacker can combine that behavior with a race involving GnuPG configuration files in temporary subdirectories to cause GnuPG to copy attacker-controlled Python code into the application code directory. The injected code executes with the privileges of the SysReptor application process after a worker restart. The Community edition is not affected. Version 2026.58 contains a partial mitigation, and this issue is fully fixed in version 2026.61.

🎖@cveNotify

🚨 CVE-2026-85058
Moquette is a lightweight Java MQTT broker. Prior to 0.18.1, PostOffice.publishWill publishes a client-controlled Last Will message through publish2Subscribers without invoking the authorizator.canWrite check used by normal PUBLISH paths. When anonymous access is enabled and topic ACLs restrict writes, a remote client can set an ACL-protected topic as the Last Will Topic during CONNECT and perform an abnormal client disconnect, causing the broker to inject attacker-controlled messages into a topic for which the client lacks write permission. This issue is fixed in version 0.18.1.

🎖@cveNotify

🚨 CVE-2026-91127
File Viewer is a browser-native viewer for Office, PDF, CAD, archive, and other files in private and internal web applications. Prior to @file-viewer/doc 2.3.1 and msdoc-viewer 0.2.2, the legacy DOC renderer emitted document-controlled hyperlink targets into generated HTML after character escaping but without restricting URL schemes. A crafted legacy DOC file could place javascript:, vbscript:, data:, or another unsafe scheme in a rendered link, and script could execute in the embedding application's origin when a user clicked the link. The fix blocks external document links by default, allows only HTTP(S), mail, telephone, safe relative URLs, and internal bookmarks when external links are explicitly enabled, and applies mount-boundary sanitization as defense in depth. This issue is fixed in @file-viewer/doc 2.3.1 and msdoc-viewer 0.2.2.

🎖@cveNotify

🚨 CVE-2026-92702
Cocos AI is a confidential computing system for running AI workloads inside trusted execution environments. In versions up to and including 0.8.2, the intra-handshake attested TLS (aTLS) AMD SEV-SNP verification path does not enforce attestation freshness when the expected reportData value is nil, empty, or omitted, leaving the SEV-SNP policy ReportData unset so the verifier accepts unrelated or stale Evidence not bound to the current connection. A relying party that uses this path without an expected reportData as a trust or authorization decision can be induced to trust an unintended attestation context; a supplied non-empty reportData is still validated. The issue is fixed in version 0.9.0.

🎖@cveNotify

🚨 CVE-2026-92745
A flaw was found in cockpit-machines. This vulnerability allows a local attacker with the ability to inspect process metadata to disclose a sensitive Red Hat Subscription Management (RHSM) offline token. The token is exposed when it is passed as a command-line argument to a helper script during the token validation process. Successful exploitation could lead to the compromise of confidentiality, as the exposed token can be used to request access tokens.

🎖@cveNotify

🚨 CVE-2019-25776
Weaver E-cology contains an unauthenticated SQL injection vulnerability that allows remote attackers to execute arbitrary SQL queries by submitting malicious input through the userIdentifiers GET parameter in the mobile plugin endpoint. Attackers can bypass space-based filter controls by wrapping SQL keywords in parentheses to perform UNION-based injection and extract sensitive data including administrator credential hashes from the database. Exploitation evidence was first observed by the Shadowserver Foundation on 2022-07-28.

🎖@cveNotify

🚨 CVE-2026-93854
In OpenStack Blazar before 17.0.1, the V2 lease API does not enforce object-level authorization on its update and delete operations (PUT /v2/leases/{lease_id} and DELETE /v2/leases/{lease_id}). The policy authorize() wrapper attempts to load the target lease to build the authorization target from its owner, but it looks up the lease under the keyword "lease_id" whereas the controller methods name the parameter "id" (and the wsme_pecan.wsexpose wrapper delivers it positionally). The lookup returns None, and thus authorization falls back to the requesting user's own project_id/user_id instead of the target lease owner. Any authenticated user who knows a lease ID can therefore modify or delete leases belonging to other users and projects, bypassing the intended ownership check.

🎖@cveNotify

🚨 CVE-2026-57226
Suricata is a network Intrusion Detection System, Intrusion Prevention System and Network Security Monitoring engine. Prior to 7.0.17 and 8.0.6, HTTP SWF decompression with the non-default swf-decompression feature and an unsafe decompress-depth can use the configured depth when allocating in src/util-file-decompression.c instead of limiting the allocation to the Flash file's actual data requirement. A crafted SWF response can therefore trigger an integer-related heap buffer overflow and crash Suricata; the default disabled feature and default depth are not affected. This issue is fixed in versions 8.0.6 and 7.0.17.

🎖@cveNotify

🚨 CVE-2026-61722
FluidSynth is a software synthesizer based on the SoundFont 2 specifications. From 2.5.0 until 2.5.6, the native DLS parser validates articulation chunks using the unsigned expression cbsize + connblocks * 12 without first ensuring that the multiplication and addition fit in 32 bits. A crafted DLS file can supply a large connblocks value that wraps the expression and bypasses the chunk-size check, after which the parser performs approximately one billion 12-byte iterations beyond the chunk boundary. The excessive processing and invalid reads can cause denial of service. Builds with the CMake option enable-native-dls set to OFF do not expose the parser. This issue is fixed in version 2.5.6.

🎖@cveNotify

🚨 CVE-2026-61820
pg_partman is a PostgreSQL extension that manages partitioned tables by time or ID. Prior to 5.5.0, inherit_template_properties() manually surrounds primary-key column names from pg_attribute.attname with double quotes without escaping embedded double-quote characters. A partman_user who owns a template table can create a crafted column name that breaks out of the generated ALTER TABLE ADD PRIMARY KEY identifier when the background worker applies the key to a child partition. The generated SQL then executes with pg_partman_bgw.role privileges, which default to PostgreSQL superuser, permitting database-wide compromise and operating-system command execution as the PostgreSQL service account. The crafted catalog identifier persists until removed and can trigger again during later partition creation. This issue is fixed in version 5.5.0.

🎖@cveNotify

🚨 CVE-2026-76900
CordysCRM is an open source AI-powered customer relationship management system that supports private deployment. In version 1.7.3, ApprovalResourceService.sendWebHook reads WebHookConfig.webHookUrl from stored approval-node configuration and passes it through ApprovalFlowService.updateApprovalPostField to HttpClientUtils without the SSRF validation used by the optional testConnect path. A user with PROCESS_SETTING_ADD can configure an internal URL through POST /approval-flow/add and cause the server to request it when POST /approval-action/approve executes the approval action, enabling cloud metadata access, internal network reconnaissance, and interaction with reachable internal services. This issue is fixed in version 1.7.4.

🎖@cveNotify

🚨 CVE-2026-81657
IBM Guardium Data Protection 12.2 could allow a remote unauthenticated attacker to execute arbitrary code on the system due to the deserialization of untrusted data.

🎖@cveNotify

🚨 CVE-2026-91202
A flaw was found in cockpit-files. A low-privileged local user can exploit this vulnerability by crafting a directory containing a symbolic link (symlink) and then using the privileged "Paste as owner" function. This allows for arbitrary file ownership changes outside the intended pasted directory, leading to a compromise of data integrity. In some cases, this could also lead to reduced confidentiality if the new ownership grants unauthorized read access. Exploitation requires user interaction to select a non-original owner during the paste operation.

🎖@cveNotify

🚨 CVE-2026-93839
LightLLM through 1.2.0 contains an authentication bypass vulnerability in the /pd_register WebSocket endpoint that allows unauthenticated attackers to register arbitrary nodes by supplying crafted JSON without peer address validation. Attackers can disclose full user prompts routed to their socket, trigger denial of service by replacing legitimate nodes, or make the PD Master issue requests to internal network addresses.

🎖@cveNotify

🚨 CVE-2026-93870
Cotonti through 1.0.0 fails to validate anti-CSRF tokens in the ratings plugin AJAX handler, allowing attackers to forge ratings on behalf of authenticated users. Attackers can craft malicious pages that auto-submit POST requests to modify stored rating data when visited by logged-in users.

🎖@cveNotify

🚨 CVE-2026-57223
Suricata is a network Intrusion Detection System, Intrusion Prevention System and Network Security Monitoring engine. Prior to 7.0.17 and 8.0.6, the Windows service installation and parameter-update logic in src/win32-service.c can pass an unquoted service ImagePath to CreateServiceA. When Suricata is installed below a path containing spaces and an earlier path component is writable by a local low-privileged attacker, Windows can execute an attacker-controlled program as LocalSystem, resulting in local privilege escalation. This issue is fixed in versions 8.0.6 and 7.0.17.

🎖@cveNotify

🚨 CVE-2026-57227
Suricata is a network Intrusion Detection System, Intrusion Prevention System and Network Security Monitoring engine. From 7.0.0 until 7.0.17 and 8.0.6, the MQTT parser in rust/src/mqtt/mqtt.rs permits repeated PUBREC or PUBREL messages to be appended to one transaction without a limit. Crafted MQTT traffic can grow transaction state indefinitely, consuming CPU and memory and causing slowdown or denial of service. This issue is fixed in versions 8.0.6 and 7.0.17.

🎖@cveNotify

🚨 CVE-2026-63448
Suricata is a network Intrusion Detection System, Intrusion Prevention System and Network Security Monitoring engine. Prior to 7.0.17 and 8.0.6, the SMB parser can retain force-completed transactions on flows where Suricata sees payload in only one direction, including async-oneside flows, because cleanup waits for inspection in the unseen direction. The transaction creation paths in rust/src/smb can exceed the intended SMB_MAX_TX bound, and cleanup repeatedly scans the growing list. Sustained one-directional SMB traffic can therefore cause unbounded per-flow state and CPU and memory exhaustion. This issue is fixed in versions 8.0.6 and 7.0.17.

🎖@cveNotify

🚨 CVE-2026-85271
Open edX Platform enables the authoring and delivery of online learning at any scale. From Redwood until Ulmo and Verawood.1, the add_additional_attributes_to_notifications function in openedx/core/djangoapps/notifications/email/utils.py assigns notification content without sanitizing discussion-title values produced by get_notification_content in openedx/core/djangoapps/notifications/base_notification.py. An enrolled student can place CSS-capable markup in the post_title value supplied by lms/djangoapps/discussion/rest_api/discussions_notifications.py. Digest and batched-email rendering then passes that value through openedx/core/djangoapps/notifications/templates/notifications/digest_content.html as safe HTML, allowing email-open tracking and content spoofing or phishing when another learner opens a CSS-rendering client. The immediate-email path is not affected because it strips title markup and renders a separately sanitized body. This issue is fixed in Ulmo and Verawood.1.

🎖@cveNotify