Cross-site request forgery in the OAuth extension of Zimbra Collaboration Suite (ZCS) Network Edition lets an attacker bind an external account they control to a victim's mailbox, because the OAuth callback neither binds its state value nor its authorization code to the initiating Zimbra session and instead picks the destination mailbox from the browser's current authentication cookie. ZCS 10.1.16 and 10.1.20 are affected, and the attack requires a deployment where the OAuth extension and a provider are actually configured, a fresh authorization code obtained for the attacker's own provider account, and a signed-in victim induced to load the crafted callback (CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:L/A:N, 7.1). Successful linking creates a persistent attacker-controlled data source that imports contacts or events, and where bidirectional CalDAV synchronization is enabled, appointments later created or edited in the linked calendar are disclosed to the attacker; no public exploit code has been identified at time of analysis, and the case is prioritised for this week rather than treated as a drop-everything emergency given its Network Edition-only, pre-configured-provider and user-interaction preconditions.
Reflected cross-site scripting in the VikRentCar WordPress plugin (all versions through 1.4.8) lets a remote attacker run script in a victim's browser by luring them to a crafted link that reflects unsanitized input back into a page generated by the plugin. CVSS 3.1 rates it 7.1 (AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:L); the scope change reflects that the injected script executes in the context of the hosting WordPress site rather than the vulnerable component alone. No confirmed active exploitation (not in CISA KEV) and no public exploit code were identified at time of analysis; Patchstack reported the issue and the vuln.today decision feed places it at 'this_week' priority.
Stored cross-site scripting in the VikRentCar WordPress plugin through version 1.4.8 lets an attacker store a crafted payload that executes when a victim views the affected page. The CVSS 3.1 score is 7.1 (AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:L), so the vector indicates network reachability and no required privileges, but exploitation depends on user interaction with the stored content. No public exploit is identified at time of analysis, and the available source does not specify the vulnerable input field or the exact privilege needed to inject the payload.
Privilege escalation in Tandoor Recipes before 2.7.0 lets any authenticated user obtain admin-level authorization inside a space where they only hold guest or user rights, by priming a per-user permission cache while their own self-created space is active. The flaw is in has_group_permission(), which caches group names under GROUP_CACHE_<user id> with no active-space component and re-sets the 30-second TTL on every request, so cached admin groups from space A keep answering checks after switch_user_active_space() moves the session into victim space B - and group revocations or demotions never take effect while the user stays active. Rated CVSS 4.0 7.1 (AV:N/AC:L/AT:N/PR:L/UI:N/VC:L/VI:H/VA:N); exploitation needs a multi-space deployment, an authenticated account, and - on multi-worker installs using the default per-process LocMemCache - the escalated request must land on the worker that primed the entry. Not listed in CISA KEV, but the vendor advisory documents a verified proof-of-concept harness against the real functions, so the technique is fully public and reproducible.
Memory-note lifecycle tooling in obsidian-tc 1.0.1 through 1.31.3 can overwrite the entire body of an Obsidian vault note it does not own, and can leave the SQLite memory graph half-committed when a filesystem step is refused. An authenticated caller holding memory-write scope (write:memory) that invokes create_entity, link_entities, rename_entity, unlink_entities or delete_entity on a (type, name) pair whose computed memory/<type>/<name>.md path is already occupied by a hand-authored note or a foreign orphan will have that note's body regenerated from SQLite, with only its frontmatter preserved; delete_entity still requires its own deletion permission and destructive confirmation, which this flaw does not bypass. The maintainer found and fixed the class in 1.31.4 (PR #978, GHSA-r3g7-r4hr-m258); no public exploit identified at time of analysis, and the issue is rated this_week priority rather than an emergency because exploitation depends on already-granted memory-write permissions.
obsidian-tc 1.13.0 through 1.31.3 lets an authenticated client complete destructive, dispatch-gated Obsidian vault operations (delete, cross-folder move, frontmatter replace, link rewrite) even after the human explicitly declined or cancelled the human-in-the-loop confirmation, or after the elicitation response was omitted or malformed. The server treats a validly bound requestState echoed on resubmission as proof of approval, but that state is minted before the elicitation leg is ever asked, so it only proves the resubmission belongs to the same call; exploitation requires a Streamable HTTP deployment with auth.jwtSecret configured, a 2026-07-28-era client declaring elicitation.form support, and a tool marked destructive:true or carrying a human-in-the-loop scope floor. Only client-driven HTTP deployments are affected - stdio or jwtSecret-unset installs fall back to the CLI elicit_token path and never offer the round trip - and there is no public exploit identified at time of analysis (no CISA KEV listing), though the advisory notes the underlying construction was reproduced directly by the maintainers.
Stored cross-site scripting in baserCMS (5.0.0 through 5.3.1 and 5.4.0 through 5.4.1) lets an unauthenticated visitor post a blog comment containing a shortcode such as [BcBaser.getGoogleMaps ...] that the CMS expands across the entire rendered page after HTML escaping, so an attacker-chosen URL is emitted as a script src and runs JavaScript on the site's own origin for anyone who opens that article. An administrator who views the poisoned article exposes their authenticated browser context, and the same expansion mechanism lets an attacker embed live server-side content in a comment, for example the Mail.getForm contact form complete with a valid CSRF token. A verified, step-by-step proof of concept against a clean 5.4.1 instance is published in the GitHub advisory; there is no CISA KEV listing, and the item is prioritised for remediation this week rather than as an emergency.
DOM-based cross-site scripting in the AivahThemes Hostme v2 WordPress theme (versions through 7.0) allows remote attackers to execute script in a victim's browser session on an affected site. The flaw is rated CVSS 7.1 (AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:L): no attacker authentication is required, but exploitation depends on a targeted victim visiting or clicking attacker-supplied content that the theme's client-side JavaScript renders into the DOM without neutralization, and the scope change means injected code can act beyond the vulnerable component. The exact vulnerable parameter and DOM sink are not described in the available sources; no public exploit code was identified at time of analysis and there is no CISA KEV confirmation, and the advisory reference does not state a fixed version.
Client-IP spoofing in dadrus Heimdall (versions up to and including 0.17.16) lets remote attackers inject arbitrary strings, malformed IP literals, RFC 7239 'unknown' values and obfuscated identifiers into Request.ClientIPAddresses whenever the optional trusted_proxies setting is enabled, because the Forwarded (for=) and X-Forwarded-For values are never checked for syntactic validity and the Forwarded parser splits on ',' and ';' without honoring quoted strings. The practical consequence is deployment-dependent: rules that consume ClientIPAddresses (for example a CEL authorizer using networks() for IP-range checks, or a Remote authorizer forwarding the client IP) can be made to evaluate against attacker-chosen values, and in proxy mode the injected values are passed upstream unchanged in X-Forwarded-For/Forwarded headers. Exploitation is gated on non-default configuration (trusted_proxies), on attacker influence over the forwarding headers, and for the access-control path on rules actually referencing ClientIPAddresses; no public exploit code was identified at time of analysis and the issue is not in CISA KEV, with remediation available in 0.17.17.
Local privilege escalation and potential host compromise in the Linux kernel's arm64 KVM nested virtualization (NV) support, caused by a use-after-free of the nested_mmus shadow page-table array. Affected builds are those containing the NV nested_mmus implementation introduced by commit 4f128f8e1aaac189f83d0f828bcdb2986d8d2e51 (from Linux 6.11 onward) and running on arm64 hosts that actually enable the experimental, non-default NV feature; fixed versions are 6.12.113, 6.18.56, 7.2.9 and 7.3. An authenticated, local actor with guest/VMM-level privileges (CVSS PR:L, AV:L) must drive a VCPU_INIT that relocates the array while another vCPU is actively executing an L2 context, a narrow concurrency window (AC:H), which yields high confidentiality, integrity and availability impact (CVSS 7.0). There is no KEV listing and no confirmed working public exploit; EPSS is 0.21% (11th percentile), consistent with a constrained, hard-to-trigger local issue rather than a mass-exploitation emergency.
An undocumented root-level backdoor in the DJI CrystalSky ZS600C 'assistant' daemon allows a physically present attacker with USB access to execute arbitrary OS commands as root on firmware 0.3.9.7 through 0.6.6.9. Sending the hard-coded string 'dji26656677' to /dev/ttyGS0 unlocks a mode in which each subsequent line is passed to system(), with no authentication or user interaction required once USB access is obtained. The physical-access prerequisite is the dominant limiting factor: devices kept in continuous custody are not exposed, while units that leave custody during shipping, loan, resale or shared use are at risk; no public exploit identified at time of analysis.
Linux kernel builds for T-Head C9xx RISC-V CPUs can silently skip the GHOSTWRITE errata mitigation when ERRATA_THEAD_GHOSTWRITE is enabled but ERRATA_THEAD_MAE is disabled, leaving xtheadvector active and sysfs incorrectly reporting "Not affected." The flaw is local-only and high-complexity: an attacker needs a low-privileged local account on the vulnerable hardware plus a kernel built with that non-default configuration, so practical exposure is narrow despite CVSS 7.0. A vendor patch is available; no public exploit has been identified at time of analysis and EPSS is 0.21%.
A use-after-free-style lifetime bug in the net/mlx5 hardware steering (HWS) code can leave a dangling reference in a miss table's list when a flow-steering table that uses a default miss table is destroyed, allowing a local user with the privilege to create and tear down steering tables to corrupt kernel memory. Only hosts running the mlx5 driver against NVIDIA/Mellanox ConnectX hardware with HWS enabled and a miss table registered via the mlx5hws_table_set_default_miss API are affected; the attack surface is local and gated on CAP_NET_ADMIN-equivalent privileges and specialized non-default hardware, with High attack complexity. No public exploit code or active exploitation has been identified at time of analysis, and EPSS is low (0.21%, 10th percentile); fixed kernels are available from the vendor.
Writing encrypted data on Linux kernel 6.19 through 7.2 can trigger a NULL pointer dereference in ext4_finish_bio, crashing the kernel from the ext4-rsv-conversion workqueue. The flaw appears when the ext4 'encrypt' feature is enabled on an already-mounted filesystem, which bypasses the mount-time check that normally disables large folios for encrypted inodes; the crash path is local, requires a privileged administrative action (toggling a filesystem feature on a live mount), and affects availability only. The vendor fix ships in 7.2.10 (commit 40554d1221e3d2ff2e0d192b74a05d89eb3620e7), 7.3 is not affected due to later refactoring, EPSS is low at 0.19% (8th percentile), and no public exploit or CISA KEV entry was identified at time of analysis.
Local apps on DJI CrystalSky ZS600C devices can invoke system-privileged actions-including a factory reset that wipes user data, shutdown, clock/timezone changes, and a firmware-apply hand-off-because the DJIUpdateService binder does not verify caller identity or permissions. The flaw affects firmware v0.3.9.7 through v0.6.6.9 and is reachable by an installed app with no permissions and no user interaction. No public exploit has been identified at time of analysis; the issue is access-gated to local code execution on the specific device but warrants this-week prioritization because successful abuse is destructive and disruptive.
Forwarded-header injection in Heimdall (dadrus/heimdall) proxy mode up to and including 0.17.16 lets an unauthenticated remote attacker place commas and semicolons into the Host header so that Heimdall copies them verbatim into the RFC 7239 Forwarded header it sends upstream, producing attacker-controlled for= entries that can appear to originate from trusted addresses such as 127.0.0.1 or 10.0.0.1. This is a conditional, chained access-control-bypass flaw rather than a direct compromise of the proxy: Heimdall itself suffers no confidentiality, integrity or availability impact, and impact only materializes when an upstream service parses Forwarded per RFC 7239 and makes IP-based trust or allowlist decisions on the for= value on its own rather than relying on Heimdall's access-control capabilities. No authentication is required, no public exploit code was confirmed at time of analysis (the advisory supplies a curl-based proof-of-concept reference), and version 0.17.17 fixes the issue by hardening Host, Forwarded and X-Forwarded-For handling.
OS command injection in the web management interface of the D-Link DAP1160L (hardware revision A1, firmware FW101WWb04) lets a remote attacker with a low-privilege authenticated session to the device's web UI inject shell metacharacters through the /htdocs/web/diagnostic.php endpoint and have them executed by the underlying operating system, yielding full confidentiality, integrity and availability impact (CVSS 3.1 base 8.8, AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H). Exploitation requires network reachability to the management interface plus valid low-privileged credentials; it is not exploitable by anonymous attackers, and devices whose admin interface is confined to a trusted LAN or management VLAN are correspondingly less exposed. The flaw is not in CISA KEV, EPSS is low at 0.84% (58th percentile), and a public PoC reference exists in a GitHub writeup; it is not verified as working exploit code.