2026-11-03
An infinite loop in the ShapeRecord::loadFromFile parser of phpMyAdmin ShapeFile (versions up to and including 4.0.1) allows remote attackers to stall or hang the importing process by supplying a crafted ESRI .shp file, consuming CPU until the process is killed or times out. No authentication is required, but the crafted shapefile must actually be fed into the SHP import/parse path for the condition to fire, so exposure is limited to deployments that accept externally supplied shapefiles. Impact is confined to availability (VA:L) with no confidentiality or integrity effect; a public PoC reference exists, but it is not verified as working exploit code, and there is no confirmed active exploitation (no CISA KEV listing).
Uninitialized-memory disclosure in the MongoDB C Driver's legacy GridFS file handling allows an authenticated database user with ordinary write access to a shared GridFS storage area to set up conditions under which a separate application later appends to the affected file, causing bytes left over in the application's process memory - potentially including credentials - to be written into records that same user is permitted to read. Because the trigger depends on that second application's append behavior and on the legacy code path being in use, the leak is opportunistic rather than guaranteed and the issue is a moderate, authenticated information-disclosure risk rather than an urgent widespread threat. No public exploit code and no confirmed active exploitation were identified at time of analysis; the vendor tracks the defect as CDRIVER-6508.
Path traversal in the datadir plugin of ruTorrent before 5.3.20 lets an authenticated user escape the configured topDirectory jail and have the web server create directories and move torrent data to locations outside it. The root cause is that correctDirectory() compares the submitted datadir as written and does not resolve symbolic links, while the downstream mkdir/rename/copy operations do follow them; an attacker who can place a symlink under their own top directory and then submit a datadir path through it can redirect torrent data to any path writable by the web server or rtorrent, such as the web root. Exploitation requires a valid ruTorrent account with datadir plugin access and the ability to create that symlink; no public exploit code was identified at time of analysis, and the vendor fix is in 5.3.20.
Client-side use-after-free in FreeRDP's urbdrc USB redirection channel lets a malicious RDP server crash a connecting client that runs with /usb enabled while a matching local USB device is being announced. FindChannelById returns a raw channel pointer without taking a reference, and the control-channel CLOSE_REQUEST path frees that channel without holding the device manager's loading_lock, so the libusb announcement thread can read channel->Write from freed memory. Triggering requires the victim to initiate the connection with USB redirection active and to win a narrow timing window (AC:H/AT:P); a PoC harness shipped with the vendor advisory demonstrates the use-after-free under AddressSanitizer, but it is not an end-to-end RDP reproduction and is not verified as working exploit code, and no active exploitation has been reported.
Information disclosure in MongoDB Database Tools versions 3.1.4 through before 100.20.1 allows an authenticated user who can modify GridFS file metadata to set the stored filename to '-' and cause a later authorized mongofiles get_id retrieval without --local to write the file contents to standard output instead of a local file. The attacker only gains access to content they cannot read directly if a higher-privileged operator performs that retrieval and the captured standard output is readable by the attacker, making exploitation highly conditional; no public exploit was identified at time of analysis.
Denial of service in the Eclipse Vert.x Redis client (io.vertx:vertx-redis-client) 4.0.0 through 4.5.35 and 5.0.0 through 5.2.2 lets a malicious, compromised or man-in-the-middle Redis server exhaust the client's heap and crash the JVM. Two triggers are documented: a server that streams bytes on an EOL-terminated line (such as a simple string) without ever sending the terminating CRLF, which grows the read buffer without bounds because no configurable line limit exists, and a crafted RESP bulk string with a missing or corrupt CRLF trailer (for example `$3\r\nfooXX`) that drives the parser into an infinite loop, re-parsing the same buffer and re-emitting the same response to the consumer until memory is exhausted. The flaw is remotely reachable without authentication (CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H, 5.9) and affects only availability - confidentiality and integrity are not impacted - with the significant precondition that the client must be talking to an attacker-controlled or compromised endpoint; a public proof-of-concept reference exists, but it is not verified as working exploit code, and there is no CISA KEV confirmation of active exploitation. Fixed versions are 4.5.36 on the 4.x branch and 5.2.3 on the 5.x branch.
µD3TN daemon crashes and denial of service can result from a use-after-free in the Bundle Protocol version 6 (BPv6) fragmenter cleanup path. The flaw is reachable by remote, unauthenticated attackers per the CVSS vector, but triggering it requires the node to be processing BPv6 bundles and a memory allocation to fail during cleanup, so reliable exploitation depends on memory pressure or resource constraints. No public exploit code has been identified at the time of analysis and active exploitation has not been confirmed; the CVSS 5.9 score reflects availability-only impact with high attack complexity.
Denial of service in µD3TN allows an unauthenticated, remote sender of BPv6 (Bundle Protocol version 6) traffic to crash the bundle node by driving the fragmenter's block-duplication code path into a double free, but only if a heap memory allocation fails at that exact point (CVSS 5.9, CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H, availability-only). Because that allocation-failure precondition is not attacker-controlled, practical exploitation generally requires first inducing memory pressure or exhaustion on the target node, which is why the vector carries AC:H. No public exploit code and no CISA KEV entry were identified at time of analysis, and the vuln.today priority is this-month remediation rather than emergency response.
WP Fastest Cache, a WordPress caching plugin, writes the full response of Core's wp-activate.php to a publicly readable cache file in versions up to and including 1.5.3, ignoring that endpoint's no-store/no-cache headers, its wp-activate-* cookie, and the WP_INSTALLING state. Because WordPress prints the new account's plaintext username and twelve-character password on that activation screen, unauthenticated attackers who can request the cached file can read the credentials of the most recently activated Multisite registrant. The exposure is heavily gated: the site must be a Multisite subdirectory installation with both public user and site registration enabled, and a real registrant must have just followed their activation link; the vulnerability is not known to be actively exploited (no CISA KEV entry) and no public exploit code was identified at time of analysis.
A divide-by-zero in NVIDIA nvJPEG (CWE-369) lets an attacker crash applications that decode a crafted JPEG image through the library, producing a denial of service with no code execution or data exposure. The CVSS 3.1 vector (AV:L/AC:L/PR:N/UI:R, A:H) requires the malformed image to reach a locally running nvJPEG consumer and a user or application to actively process it, so exposure is limited to environments that feed untrusted images into nvJPEG-based decoding such as image services and ML preprocessing pipelines. No CISA KEV entry, no confirmed public exploit, and no patch version are present in the supplied intelligence; the NVIDIA product-security advisory for 2026 (5910) is the root reference, and this is a moderate-severity availability issue rather than a top-priority emergency.
Opening and traversing a crafted HDF5 file can crash the application because H5HL_unprotect() in HDF5 through 2.2.0 may pass a NULL local-heap pointer to H5AC_unpin_entry(). The bug is a CWE-476 NULL pointer dereference: if the metadata cache destroys a corrupted local heap entry and H5HL__prfx_dest() clears heap->prfx or heap->dblk, the final unprotect has nothing to unpin, causing an assertion failure in debug builds and a crash in release builds. Impact is limited to local denial of service with user interaction and no confidentiality or integrity loss; the supplied CVSS 4.0 vector has E:P (proof-of-concept maturity), but no CISA KEV listing or confirmed working exploit code was provided.
Open redirect in the TYPO3 'mail' extension's click tracking middleware allows an unauthenticated remote attacker to craft a tracking link on a trusted TYPO3 site that forwards victims to an arbitrary external destination, turning the site's own domain into a phishing lure. It is rated CVSS 5.3 (medium) by the vendor and CVSS 4.0 5.3 by the reporting source, with the assessed vector CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:N/I:L/A:N - the practical impact is limited to integrity of the redirected-to context, so risk is social engineering rather than direct compromise. No public exploit code or PoC is referenced in the available intelligence and the issue is not in CISA KEV; the vendor priority label places it in the this-month remediation window.
Saving the backend module configuration of the third-party TYPO3 extension "mail" writes the submitted settings into page TSconfig without restricting them to the module's own configuration keys, allowing arbitrary page TSconfig injection. The flaw affects installations running the extension (CPE cpe:2.3:a:typo3:extension_"mail" with no version bound) and is reachable by a backend user who holds a valid account and has been granted access to the mail module - no victim interaction is required and attack complexity is low, but permissions are required, so it is not unauthenticated. Impact is integrity-only and low (assessed CVSS:3.1 AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:L/A:N); no public exploit code has been identified at time of analysis, no EPSS score or CISA KEV entry was supplied, and the vuln.today priority decision is this_month rather than an emergency.
The TYPO3 extension "mail" (EXT:mail) performs server-side fetches of content referenced by an "external URL" mail - the message body, its stylesheets and its embedded images - without validating the destination host, allowing server-side request forgery against internal network services. Exploitation requires a valid TYPO3 backend account that holds access to the mail module, so the attacker is an authenticated operator/insider rather than an anonymous outsider; no user interaction is needed and the behaviour is the extension's default. Impact is limited to low-confidentiality disclosure of what internal hosts return, with no integrity or availability effect, and an independent assessment rates the issue moderate with a this-month priority; no public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV.
HTTP 500 responses generated by RESTEasy's default exception mapper embed the unhandled exception's localized message, so a remote unauthenticated client that can induce such an exception receives whatever text that message carries (CWE-209). The behavior was reproduced on RESTEasy 7.0.2.Final and, per the CPE data, flows into a wide set of Red Hat products that bundle RESTEasy, including Red Hat Build of Quarkus and Keycloak, JBoss EAP 8 and its expansion pack, Fuse 7, Satellite 6, Red Hat Build of Apache Camel 4 for Quarkus 3, Red Hat Build of Apicurio Registry 3 and Red Hat Build of Debezium 3; complete affected and fixed version ranges remain unconfirmed. No public exploit was identified at time of analysis, and the confidentiality impact is capped by the actual content of the leaked message rather than by arbitrary memory or file exposure. Priority is 'this_month' rather than immediate.
Error messages returned by the Exchange RPC in Octo STS 0.11.1 and earlier disclose the contents of trust policies to callers whose OIDC token does not satisfy them. Remote, unauthenticated callers who hold any validly signed OIDC token from an accepted issuer (for example a GitHub Actions token from a repository they own) and who know or guess a target owner/repo plus identity name can read the policy's issuer, subject and audience patterns, claim_pattern regexes, and app pin details - information that frequently names internal repositories, branches, environments or cloud account identifiers, even when the policy lives in a private repository. The flaw is confidentiality-only: no token minting or trust-policy bypass is possible; the vendor issued a fix in v0.11.2 (commit 88c6432), and no public exploit was identified at time of analysis.
Axxon One deployments before version 3.2.0 on Windows and Linux expose their full user and role inventory to any authenticated low-privilege account: the security configuration API's user configuration request returns the complete lists of users and roles regardless of the caller's permissions. Because the flaw only requires valid credentials for an ordinary account and no user interaction, it lowers the barrier to privilege mapping and follow-on targeting, though it grants no modification or disruption capability. There is no public exploit code or confirmed active exploitation, and the available signals rate this as a low-urgency, non-emergency information disclosure to be scheduled for this month's patch cycle; upgrading to 3.2.0 removes the exposure.
Incorrect authorization in the Health Monitoring service of AxxonSoft AxxonNet on Linux (versions before 3.28.0) lets a remote attacker who already holds a shared report token - a credential intended only for read-only report access - create, modify, or delete monitoring alerts through the monitoring API. The flaw is a scope-enforcement failure rather than an authentication bypass in the strict sense: a valid low-privilege token is a hard prerequisite (PR:L, AV:N), and the resulting impact is limited to alert integrity and availability, with no confidentiality loss and no path to code execution. Severity is medium (CVSS 4.0 base 5.3; independently assessed CVSS 3.1 5.3, S:U/C:N/I:L/A:L); there is no CISA KEV listing and no public exploit identified at time of analysis, so the practical risk is suppression or forgery of health alerts by someone who already has report-level access.
Path traversal (CWE-22) in isomorphic-git up to and including 1.38.5 lets a malicious or man-in-the-middle-controlled non-GitHub remote write attacker-chosen files outside the working-tree root when a victim clones or checks out that repository, and overwriting shell rc, configuration or startup files can escalate the write into code execution. Affected are Node.js and browser deployments using any host-backed fs provider on Linux, macOS or Windows; the issue is unauthenticated (PR:N in the vector) but requires user-initiated clone/checkout of untrusted content (UI:R) and a non-GitHub remote, because GitHub runs fsck on push and rejects '..' entries, and escalation further depends on a writable security-relevant parent-path file existing. A public PoC reference demonstrating the escape write exists; it is not verified as working exploit code, there is no CISA KEV listing, and the overall priority is moderate rather than an emergency.
Authorization bypass in NetBox (2.9.0 up to but not including 4.7.4) lets authenticated users who hold a bulk add-component permission such as dcim.add_interface or dcim.add_modulebay act on parent devices and virtual machines that their ObjectPermission constraints should hide from them. By submitting arbitrary primary keys in the 'pk' list, or '_all' together with a GET filter, the attacker can view out-of-scope device details and create components on those parents; the flaw is in BulkComponentCreateView, which failed to apply the 'view' permission restriction to the parent queryset. Impact is limited to confidentiality and integrity at low severity (CVSS 5.3 / assessed CVSS:3.1 5.3, PR:L, no availability impact), it requires a valid account and the add-component permission, and it matters chiefly in multi-tenant or role-separated deployments using constrained ObjectPermissions; no public exploit code was identified at time of analysis and the issue is not in CISA KEV, with a vendor fix released in 4.7.4.
Stored cross-site scripting in buku's optional bukuserver web component (versions 5.0 through 5.1.1) lets an unauthenticated remote attacker plant a bookmark URL that executes JavaScript in the browser of any user who later opens the bookmark list page. Because the injected script runs same-origin with the bukuserver UI, it can issue API requests carrying the viewer's credentials, but the realized damage is bounded to those same-origin actions and requires a separate victim to load the affected page (passive user interaction). No CISA KEV entry and no public exploit identified at time of analysis; this is a genuine but moderate-priority web vulnerability rather than a high-urgency emergency.
Missing authorization in the ClickCease Click Fraud Protection WordPress plugin, versions through 3.3.0, allows remote unauthenticated attackers to reach a plugin feature that lacks an access-control/capability check, resulting in a bounded integrity impact with no confidentiality loss and no availability impact. The CVSS 3.1 base score is 5.3 (Medium) with vector AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:N, so no privileges and no user interaction are required, but the scope of what an attacker can change is limited (I:L). No public exploit code has been identified at time of analysis, and the issue is not listed in CISA KEV; the specific endpoint or action missing the check is not described in the available sources.
AnyWhere Elementor, a WordPress plugin by WPVibes, ships with a missing authorization check (CWE-862) in versions up to and including 1.3.0, letting unauthenticated remote requests reach functionality that should have been capability-gated. Per the CVSS vector (PR:N/UI:N) no credentials or user interaction are needed, but the exposure is limited to partial confidentiality loss (C:L) with no integrity or availability impact, giving a moderate CVSS 5.3. No public exploit code or active exploitation is confirmed, and EPSS/KEV data were not provided; the trigger endpoint is not described in available sources, so real-world urgency is low-to-moderate rather than immediate.
Missing authorization in the weDevs Dokan Lite WordPress plugin (Dokan through 5.3.2) lets unauthenticated remote attackers reach functionality that should be access-controlled and read restricted data from the site. Rated CVSS 5.3 with confidentiality-only, low-magnitude impact (C:L/I:N/A:N), it needs no credentials and no user interaction against any site running an affected version. No public exploit was identified at time of analysis and there is no confirmed active exploitation; the priority label for this issue is 'this_month'.
Missing authorization in the Dokan Lite WordPress plugin through 5.3.2 permits unauthenticated network attackers to access limited information guarded by incorrectly configured access controls, with low confidentiality impact and no integrity or availability effect. No public exploit is identified at time of analysis, and active exploitation is not confirmed in the supplied data.
Missing authorization in WPMU DEV Forminator through 1.58.3 allows remote unauthenticated attackers to make limited integrity-impacting changes on WordPress sites running the plugin, per the CVSS 3.1 vector AV:N/AC:L/PR:N/UI:N with C:N/I:L/A:N. No public exploit was identified at time of analysis, and the issue is assessed as a moderate-severity, low-priority concern rather than an emergency; the specific endpoint or feature that triggers the flaw is not described in the available sources.
Missing authorization checks in the WC Lovers WCFM Membership WordPress plugin, in all versions up to and including 2.12.0, allow unauthenticated remote attackers to reach plugin functionality that should be gated by an access-control decision, with the official vector (CVSS 3.1 5.3, AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:N) describing only limited integrity impact and no confidentiality or availability effect. The flaw is classified CWE-862 (Missing Authorization), and the specific feature, endpoint or action that lacks the check is not described in the available sources. No CISA KEV entry and no public exploit code were identified at time of analysis, and the vulnerability is rated 'this_month' in the priority decision signal rather than act-now urgency.
Unauthenticated remote attackers can bypass intended access-control restrictions in the WordPress plugin Uncanny Automator through version 7.7.1, because a function or endpoint lacks a proper authorization check (CWE-862). The CVSS 3.1 score is 5.3 (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:N), so the worst confirmed outcome is limited integrity impact with no confidentiality or availability effect; the specific endpoint, feature, or configuration that exposes the missing check is not described in the available sources. No public exploit code and no confirmed active exploitation were identified at time of analysis, and vendor advisory details are available through Patchstack.
Unauthenticated information disclosure affects the WebberZone Better Search WordPress plugin through version 4.5.2, where a missing authorization check (CWE-862) lets network attackers reach functionality that should be access-controlled, exposing limited data (CVSS 5.3, C:L/I:N/A:N). Per the CVSS vector (PR:N/UI:N), no credentials or user interaction are required, but the impact is confined to partial confidentiality loss with no modification or denial of service. No public exploit code or CISA KEV entry was provided; the trigger endpoint is not described in the available sources, and the issue should be scheduled for patching rather than treated as an emergency.
Stored cross-site scripting in TeXlyre 0.11.0 through 0.13.4 lets an attacker run arbitrary JavaScript in the TeXlyre origin by supplying a malicious LSP or typesetter recipe whose free-form 'icon' field carries an HTML event handler such as <img onerror=...>; the payload is stored in the user's settings and re-executes each time the tool-settings card is rendered. Exploitation requires the victim to actively import the recipe via Settings → Import (file, URL, or pasted JSON), so the attack depends on social engineering around TeXlyre's recipe-sharing feature rather than unattended remote triggering. A public PoC reference exists in the vendor advisory; it is not verified as working exploit code, the issue is not listed in CISA KEV, and success yields read/exfiltration of IndexedDB-stored project documents plus access to in-memory decrypted Git backup tokens.
Missing authorization in Dromara domain-admin before 1.6.79 allows authenticated users to read member metadata from groups they do not belong to by calling POST /api/getGroupUserList with other group_id values. Attackers can enumerate sequential group IDs to disclose member usernames, leader and edit-permission flags, and owner usernames; exploitation requires a valid low-privileged account. No public exploit was identified at time of analysis, and the vendor has released version 1.6.79 with a group-membership permission check.
DbGate before 7.3.3 does not enforce LOGIN_PERMISSIONS connection restrictions on the /connections/test and /connections/save endpoints, so an authenticated user who is supposed to be limited to specific database connections can instead submit arbitrary connection definitions. That lets the user probe internal database hosts through the test endpoint and overwrite stored connection definitions through the save endpoint, both with limited (Low) confidentiality and integrity impact. Exploitation requires a valid low-privilege account in a multi-user deployment that actually enforces per-connection restrictions; no public exploit code has been identified at time of analysis, and the issue is rated medium severity (CVSS 5.3) with a moderate rather than urgent priority.
Cross-site scripting bypass in svg-sanitizer (Composer package enshrined/svg-sanitize) 1.0.0 and earlier lets an attacker who can supply or upload SVG content place a javascript: URL inside an animateTransform values attribute, which survives sanitization and overwrites an anchor's href, executing arbitrary JavaScript when a victim clicks the link. The bypass only fires in Apple Safari, which uniquely allows animateTransform to target href, and requires the victim to click the affected link (UI:R), so this is a genuine but moderate-risk stored-XSS sanitizer flaw rather than an emergency; the vendor demo tool accepted the published payload. No confirmed active exploitation or KEV listing is reported, and the upstream security advisory documents a public proof-of-concept payload; upgrading to 1.1.0 removes the bypass.
SVG Sanitizer 0.5.0 through 1.0.0 fails to strip remote URLs expressed with CSS escape sequences inside url() in SVG presentation attributes, so a sanitized SVG can still cause a browser to make an outbound request when rendered. The advisory shows that fill="url(//evil.example/x)" is removed while fill="url(\2f\2f evil.example/x)" is not, which can enable data exfiltration depending on how the application uses the SVG. This is a medium-severity issue (CVSS 4.0 base 5.3) with a this-month remediation priority; CISA KEV status is not indicated in the supplied data, and a public PoC reference exists in GHSA-3vh3-jf65-656w but is not verified as working exploit code.
Frappe HRMS before 15.61.0 allows any authenticated employee to read other employees' expense advance records by querying the Expense Claim Advances API, which fails to enforce ownership-based authorization (CWE-639). The exposure is limited to read access of sensitive financial data belonging to coworkers; there is no integrity or availability impact and an attacker must already hold valid employee credentials on the instance. A public advisory exists, no public exploit code has been identified at time of analysis, and this is tracked as a moderate-severity information-disclosure issue rather than an emergency.
Froxlor versions through 2.3.14 allow an authenticated customer to create MySQL databases through the Mysqls.add API endpoint even when the administrator has explicitly hidden or disabled the MySQL feature via panel.customer_hide_options, because add() omits the authorization check that get(), update(), and delete() enforce. The result is a policy violation rather than a breach: the customer provisions unauthorized databases only within their own existing quota, with no code execution or access to other tenants' data. This is a low-to-moderate real-world risk and should be treated as a policy/hardening fix rather than an urgent emergency; no public exploit code is confirmed in the available data, and exploitation requires both a valid customer account and the MySQL feature already being hidden.
Froxlor 2.3.0 through 2.3.14 lets an authenticated customer who holds a valid API key overwrite the Let's Encrypt-managed certificate and private key of a domain they own, because the Certificates.add and Certificates.update API commands never check the letsencrypt flag that the UI and read paths enforce. The attacker-supplied key pair is stored in domain_ssl_settings and pushed into the webserver configuration by a REBUILD_VHOST task, degrading or breaking HTTPS for that domain's visitors until the next ACME renewal and desynchronizing the panel from acme.sh; the original LE private key stays unreadable, so the exposure is an integrity/availability gap on the write path. Exploitation is bounded to the attacker's own domains and requires API access to be enabled for the customer account; a proof-of-concept with reproduction steps is documented in the published advisory, and no CISA KEV entry or EPSS score accompanies this CVE.
AzuraCast's station waveform endpoints (GET and POST /api/station/{station_id}/waveform/{media_id}) are missing the Permissions middleware that every sibling station-scoped media route carries, so any authenticated user - even one with no station assignment at all - can read another station's cached waveform JSON and overwrite it with arbitrary content. Successful abuse gives limited cross-station data disclosure and limited tampering with the Visual Cue Editor waveform data, not access to media files, credentials, or arbitrary filesystem paths (CVSS 3.1: AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:N; CVSS 4.0 base 5.3). Authentication of the lowest privilege tier is required and the attacker must reach the API with a target station_id and media_id; VulnCheck and the upstream advisory provide a PoC reference that boots the real application and introspects its compiled route table to prove the missing middleware - it is not verified as working end-to-end exploit code, and there is no KEV listing or active-exploitation reporting.
SurrealDB 3.1.0 through 3.2.x mishandles port-specific network deny rules, so an authenticated user who can call the http::* functions can reach an internal service that a --deny-net HOST:PORT or IP:PORT rule was meant to block simply by naming the same target through a different alias (a hostname that resolves to the denied IP, or the bare IP of a denied hostname). The advisory demonstrates this against the v3.2.0 release binary with published reproduction steps, so a public proof-of-concept reference exists; there is no CISA KEV entry indicating active exploitation. Scope is bounded: it is an egress-policy bypass that only matters to deployments relying on port-bearing deny rules, and bare-IP deny rules are unaffected; it is fixed in 3.3.0.
Incorrect authorization in SurrealDB 3.1.0 through 3.2.x lets an authenticated record user recover restricted field values by inference: the default streaming execution planner selects FULLTEXT, compound B-tree and HNSW indexes built on fields whose field-level SELECT permission is not FULL, then drops the index-covered predicate instead of re-evaluating it against the permission-reduced document. Affected are deployments that combine restrictive field-level SELECT permissions with indexes on those fields; direct SELECT output continues to hide the values, but result-set membership and ordering act as a confirmation or nearest-neighbour oracle, so the impact is low-severity, confidentiality-only disclosure (CVSS 4.0 5.3 / CVSS 3.1 AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N, CWE-863/CWE-200). A public PoC reference exists (a reproduction script with recorded results in the GitHub advisory); it is not verified as working exploit code, and there is no CISA KEV listing.
FreeRDP clients up to and including 3.32.1 can be terminated by a malicious or compromised RDP server that sends a crafted ALPHA WIRETOSURFACE_2 graphics-pipeline command with zero width and zero height. The session crash is client-side and requires the victim to connect to the attacker's server with RDPGFX active (AUTODETECT enables it), so exploitation depends on a user-initiated connection rather than an exposed listening service. Confirmed termination was observed on builds compiled with verbose WINPR assertions, where an element-size assertion aborts the process; on builds without that assertion the advisory describes a following division by zero as a plausible but not executed termination path. No CISA KEV listing exists; the advisory supplies a handler-level reproducer and documents a preserved live server test, but this is not verified general-purpose exploit code.
Denial of service in FreeRDP clients at version 3.32.1 and earlier lets a malicious or compromised RDP server terminate the connecting client by sending a three-byte zero-output ZGFX payload (E0 20 00) on the RDPGFX dynamic channel after session activation. The decompressor returns success while leaving its output pointer null and its size zero, and rdpgfx_on_data_received - which rejects only negative decoder status - forwards that null pointer into Stream_StaticConstInit, whose WINPR_ASSERT(buffer) aborts builds compiled with verbose WinPR assertions, the upstream CMake default. A component-level reproduction and a documented live xfreerdp test (client exit 134) are public; only assertion-enabled builds are affected, the verified impact is limited to client availability with no memory corruption, disclosure or code execution, and no CISA KEV listing or confirmed active exploitation is reported (priority label: this_month).
FreeRDP clients up to and including 3.32.1 can be driven into excessive memory allocation by a malicious RDP server when serial device redirection is enabled, because the serial channel fails to cap the server-supplied OutputBufferLength carried in an IRP_MJ_DEVICE_CONTROL request before allocating and echoing it back. Only availability is affected: the vendor measured roughly 4.01 GiB peak RSS and about 3.8 seconds of handler time from a single IOCTL_USBPRINT_GET_1284_ID request, with no confidentiality or integrity impact observed. Exploitation requires the victim to connect out to an attacker-controlled server and to have serial redirection configured with a redirected device successfully opened; no public exploit is identified beyond the reproduction harness published in the vendor advisory, and there is no CISA KEV confirmation of active exploitation.
Drive redirection in FreeRDP 3.32.1 and earlier can be silently disabled by a malicious RDP server: a single truncated IRP_MJ_CREATE device I/O request terminates the drive worker while leaving its error state at CHANNEL_RC_OK, so the RDPDR channel stays up and every subsequent server request is queued in IrpQueue with no consumer until memory is exhausted or the session ends. Only clients that initiate a connection to an attacker-controlled or compromised RDP server with drive redirection enabled (/drive or drivestoredirect) under the default asynchronous RDPDR dispatch are affected, and the impact is availability loss confined to that session, with no confidentiality or integrity impact (assessed CVSS:3.1 AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:L; upstream rates it 5.3). There is no CISA KEV listing and no reported in-the-wild exploitation; a public proof-of-concept verification harness accompanies the vendor advisory, and the issue is fixed in 3.32.2.
FreeRDP clients up to and including 3.32.1 can be crashed by a malicious or compromised RD Gateway that sends a PKT_TYPE_EXTENDED_AUTH_MSG (type 0x0003) immediately after the HTTP/WebSocket upgrade, before tunnel authentication completes. In the default configuration (GatewayHttpExtAuthSspiNtlm=FALSE) the HTTP authentication context is freed before tunnel packets are processed, so rdg_process_packet dispatches the extended-auth packet with rdg->auth NULL and a zero-length blob reaches credssp_auth_is_complete(NULL), producing a null-pointer dereference; the advisory documents an abort (exit 134) on an assertions-enabled build and a null-pointer read crash (exit 139) on an assertions-off build, with no memory-corruption or data-access effect demonstrated. The attacker does not reach the client directly: the victim must initiate a connection to the hostile gateway (CVSS 4.0 AV:N/PR:N/UI:P, our assessed CVSS 3.1 AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:L, availability-only impact). A public proof-of-concept reproduction is published in the GitHub security advisory GHSA-8rjm-m7xg-mg6m and has been independently re-verified to crash affected clients; the issue is not in CISA KEV and no in-the-wild exploitation is reported, and the vulnerability is fixed in 3.32.2.
Memory exhaustion in FreeRDP's RD Gateway client (versions through 3.32.1) lets a malicious or compromised gateway force roughly 2 GiB of client-side allocation before tunnel authentication, tying up memory while the session waits on body bytes that never arrive. A single eight-byte RDG header declaring packetLength 0x7FFFFFFF slips past the INT_MAX check in rdg_receive_packet and reaches Stream_EnsureCapacity, where WinPR rounds the request to 0x80000000 and zero-fills the expanded stream. The attack is unauthenticated (CVSS PR:N) but narrow: the victim must be running xfreerdp/FreeRDP in RD Gateway mode (/gateway over HTTP or WebSocket) and must initiate a connection to the hostile endpoint, impact is availability-only, and if the host refuses the allocation the connection fails cleanly; there is no CISA KEV entry, and the GHSA advisory publishes a full reproduction harness that was executed and measured at 2.032 GiB resident memory.
Heap buffer over-read in FreeRDP's winpr PNG encoder (save_png_to_buffer) allows a malicious or compromised RDP server to embed a limited amount of client heap memory into a clipboard image and disclose it to a local application. FreeRDP 3.x through 3.32.1 built with WINPR_UTILS_IMAGE_PNG and clipboard redirection enabled is affected when the server offers a 16 bpp BI_BITFIELDS CF_DIB and a local application requests that clipboard content as image/png. The impact is limited to low confidentiality, with no integrity or availability effect and no code execution; no active exploitation or confirmed public exploit code is identified, although the vendor advisory includes a public reproduction harness that triggers the over-read under ASan.
Remote denial of service against FreeRDP client processes in versions 3.0.0 through 3.32.1: an attacker who operates (or intercepts with an accepted certificate) the Remote Desktop Gateway a user connects through can send a single crafted RPC-over-HTTP RTS PDU whose authentication padding contradicts the lengths declared in the header, tripping a reachable assertion in rts_read_auth_verifier and aborting the client. The abort is only reachable in builds compiled with WITH_VERBOSE_WINPR_ASSERT=ON, which is not the default for release-style builds; with the option disabled the same packet is rejected through the ordinary connection-error path with no crash, and FreeRDP server, proxy and shadow roles never enter this parser. No confidentiality or integrity impact and no memory corruption is involved, exploitation requires the victim to initiate the gateway connection (PR:N but user interaction required), and no in-the-wild exploitation has been confirmed; a public PoC reference exists in the form of an inline reproducer published in the upstream advisory, and it is not independently verified here as working exploit code.
A hostile RDP server can exhaust the file descriptor table of a FreeRDP client by repeatedly sending IRP_MJ_CREATE requests for a redirected parallel device, leaking one local descriptor per request until descriptor allocation fails and the client process crashes. FreeRDP versions 3.0.0 through 3.32.1 are affected (fixed in 3.32.2), and only clients that explicitly enable parallel-device redirection to an openable local path are exposed - with redirection disabled the vulnerable code path is unreachable. The CVSS vectors supplied by the vendor and by our assessment both mark this unauthenticated and network-reachable but user-interaction dependent (UI:P, PR:N), availability-only, and bounded to the single client process; the advisory ships a tested local relay PoC, so a public PoC reference exists for reproducing descriptor growth in a lab, but no active exploitation (CISA KEV) or EPSS signal is present in the available data.
FreeRDP clients built with the optional WITH_AAD=ON flag can be aborted by a crafted ARM gateway response: an HTTP 200 JSON reply that omits the redirectedServerCert field (or supplies it as a non-string) drives a NULL buffer into the target-certificate parser, tripping a WINPR_ASSERT in the stream initializer and terminating the client with SIGABRT. Only non-default AAD-enabled builds (>= 3.0.0 through 3.32.1) that also link WinPR JSON are affected; the advisory states the default AAD-off build does not compile the ARM gateway route and is unaffected. Impact is client-side availability loss only with no confidentiality or integrity effect, and exploitation requires the user to point the client at a malicious or malformed gateway. A public proof-of-concept fake-gateway reproducer is included in the vendor advisory, but it is a loopback-only test harness rather than general-purpose exploit code, and no confirmed active exploitation or KEV listing exists; the vendor addressed the flaw in 3.32.2.