Use-after-free in GPAC's MP4Box multimedia tool (all versions through commit f1219cde) allows a local low-privileged attacker to corrupt heap memory via the gf_list_count function in utils/list.c, resulting in application crashes and potentially limited code execution within the process. The fix released in abi-16.23 (commit 49dee5cad329) addresses multiple interrelated UAF and null-dereference conditions across the scenegraph manager, BIFS memory decoder, and AV1 reframer subsystems. A public proof-of-concept exploit exists; no active exploitation has been confirmed via CISA KEV.
Null pointer dereference in GPAC's MP4Box scenegraph engine crashes the application when processing malformed MPEG-4 BIFS scene data. The flaw resides in gf_node_list_add_child within base_scenegraph.c, reachable when node registration encounters a zero instance count or a null node pointer in gf_node_unregister() and gf_node_get_name_and_id(). A public proof-of-concept exploit exists and a vendor-released patch (abi-16.23) is available; no active exploitation is recorded in CISA KEV.
Null pointer dereference in GPAC's MP4Box tool crashes the process when parsing crafted MPEG-4/VRML scene graph data containing a malformed MFURL field, triggering the flaw inside gf_sg_mfurl_del() during scene graph cleanup. Local, low-privileged users can cause a denial of service by feeding MP4Box a specially crafted file; there is no confidentiality or integrity impact. A public proof-of-concept is available and the CVSS 4.0 vector confirms Exploit maturity of Proof-of-concept (E:P), making this straightforward for local attackers to reproduce, though impact remains limited to application crash.
SQL injection in LuxSoft LuxCal through version 5.3.4L allows unauthenticated remote attackers to extract data from the underlying database by sending crafted requests to two public-facing PHP endpoints: rssfeed.php and common/retrieve.php. The vulnerability is reachable without authentication, as RSS and data-retrieval endpoints are typically exposed to anonymous visitors. A researcher has published proof-of-concept details, making exploitation straightforward for any attacker with network access to a LuxCal instance. No public exploit identified at time of analysis beyond the researcher disclosure.
Unauthenticated password reset in jaychouchannel's Tourism-Management-System allows any remote attacker to overwrite an arbitrary user's credentials by supplying only a known username to the resetPass endpoint - no current password, token, or session is required. All three Java Spring controllers (UsersController, DaoyouController, YonghuController) share the vulnerable pattern, enabling forced account lockout or, if the API response includes the newly generated random password, full account takeover. Publicly disclosed exploit code exists and the upstream fix is available via commit 9cb6215.
Hard-coded API token in dromara orion-visor through version 2.5.7 allows remote unauthenticated attackers to bypass the ExposeApi authentication gate enforced by ExposeApiAspect.beforeExposeApi. The static token is now publicly documented in a proof-of-concept writeup on GitHub after the vendor failed to respond to the responsible-disclosure issue report. No public exploit confirmed in CISA KEV at time of analysis, but a POC is available, materially lowering the bar for exploitation.
SQL injection in SourceCodester School Registration and Fee System 1.0 allows unauthenticated remote attackers to manipulate database queries via the Status parameter in /bilal/normal/save_stud.php. The vulnerability enables unauthorized data access, modification, and potential full database compromise through crafted HTTP requests. A public proof-of-concept exploit has been disclosed via a GitHub issue, raising immediate exploitation risk for any exposed instances.
Reflected XSS in Kagi Smallweb's query string rendering allows injection of JavaScript into HTML form action attributes across three server-rendered templates. Jinja2's |safe filter was applied to the raw Werkzeug request.query_string value in flag_panel.html, mobile_more.html, and reactions.html, bypassing auto-escaping when unencoded double-quote characters appear in the query string. Exploitation is technically feasible but practically constrained: standard browsers percent-encode double-quotes as %22, so delivery requires raw HTTP clients such as curl or netcat rather than crafted URLs sent to victims. No public exploit or active exploitation (CISA KEV) has been identified; the fix is patch commit 00b68144.
Heap memory exhaustion in Zephyr RTOS mcumgr SMP settings handlers allows unauthenticated attackers to trigger a monotonic heap leak by sending settings read, write, or delete commands that the application access hook rejects, ultimately depleting the kernel heap until the device must be rebooted. The three affected handlers - settings_mgmt_read(), settings_mgmt_write(), and settings_mgmt_delete() in subsys/mgmt/mcumgr/grp/settings_mgmt/src/settings_mgmt.c - allocate buffers via k_malloc() but execute a direct return bypassing the end: cleanup label whenever CONFIG_MCUMGR_GRP_SETTINGS_ACCESS_HOOK rejects a request via MGMT_CB_ERROR_RC. The vulnerability is availability-only with no memory corruption or information disclosure; no public exploit code or CISA KEV listing exists at time of analysis.
Reflected or stored cross-site scripting in the Order Printing component of Exrick xmall allows remote attackers to inject malicious scripts into victim browsers via the order-print.jsp template. The affected file is xmall-manager-web/src/main/webapp/WEB-INF/jsp/order-print.jsp, which fails to sanitize user-controlled input before rendering it in HTML context. Successful exploitation requires victim interaction (e.g., a privileged user viewing or printing a crafted order), making this a realistic attack against administrative personnel in e-commerce deployments using this platform.
Stored cross-site scripting in quequnlong shiyi-blog up to version 1.2.1 allows unauthenticated remote attackers to inject arbitrary JavaScript via the body.content parameter of the Add Message API, which executes in the browser of any administrator who later views submitted messages in the blog-admin panel. The affected file blog-admin/src/views/message/message/index.vue fails to sanitize message body content before rendering it, meaning a crafted submission persists until an admin reviews it. No vendor patch exists - the maintainer has not responded to the Gitee issue report (IK5RF6), and no public exploit has been identified at time of analysis.
Race condition in vvbbnn00's WARP-Clash-API Save Account Job component allows authenticated low-privilege remote attackers to exploit improper synchronization during concurrent account-state operations, yielding limited confidentiality, integrity, and availability impacts. The project has been abandoned by its maintainer, who did not respond to disclosure, and operates on a rolling release model with no versioned releases - meaning no patch will be issued. A publicly available proof-of-concept exploit exists on GitHub, though the CVSS 4.0 score of 1.3 and high attack complexity rating indicate that reliable exploitation is difficult in practice.
Race condition in the `doUpdateLicenseKey` function of vvbbnn00/WARP-Clash-API allows authenticated remote attackers with low privileges to exploit concurrent state manipulation for limited confidentiality, integrity, and availability impacts. The affected product is an abandoned, unsupported open-source Cloudflare WARP license key management tool with no versioning scheme and an unresponsive maintainer. A public proof-of-concept exploit exists on GitHub, though the high attack complexity (AC:H) and requirement for prior authentication substantially limit real-world risk.
Stored cross-site scripting in Open-Generative-AI's studio history components (ImageStudio.js, CinemaStudio.js, LipSyncStudio.js, VideoStudio.js) allows an authenticated low-privileged user to inject arbitrary HTML and JavaScript into the history panel by crafting a malicious generation prompt or URL, which then executes in any victim's browser session when the history panel is viewed. Versions up to 1.0.11 and 2.0.0 are affected per CPE and VulDB reporting. No public exploit has been identified at time of analysis, and no KEV listing exists, but the technique is trivially reproducible given the unguarded innerHTML interpolation pattern visible in the patch diff.
Cross-site scripting in shiyi-blog's search component allows an authenticated attacker to inject malicious scripts via crafted post titles or summaries, which execute in the browsers of other users who trigger search results containing the poisoned content. The `highlightKeyword` function in `blog-web/src/components/Search/index.vue` renders the title and summary fields without sanitization, making DOM injection trivial for any low-privilege content creator. No public exploit code has been identified at time of analysis; the vendor was notified via responsible disclosure through a Gitee issue report.
Stored cross-site scripting in Mogu Blog v2 through version 5.2 allows authenticated low-privileged users to inject arbitrary JavaScript via the `sortName` parameter of the `BlogSortServiceImpl.addBlogSort` function, rendered unsanitized through the FreeMarker template `info.ftl`. Any user who subsequently views the affected blog sort page becomes a victim of the injected script. No patch exists at the time of analysis; the project was notified via a Gitee issue but has not responded.
Stored cross-site scripting in the comment rendering layer of bbs-springboot 3.0.0 allows an authenticated low-privileged user to inject arbitrary HTML and JavaScript that executes in the browsers of other users who view the affected comment sections. The root cause, confirmed by the patch diff, is the Vue.js frontend's use of the `v-html` directive in `ChildComment.vue` and `LatestComment.vue`, which renders comment content as raw HTML rather than escaped text. No public exploit code has been identified and the vulnerability is not listed in CISA KEV.
Cross-site scripting in shiyi-blog 1.0.0-1.2.1 allows authenticated remote attackers to inject arbitrary JavaScript through the `chat_msg` argument of the chat sendMsg endpoint, with execution triggered when another user loads the chat history via `SysChatMsgMapper.getChatMsgList`. The vulnerability resides in the Vue.js frontend component `blog-web/src/views/chat/index.vue` due to improper output encoding of user-supplied chat content. No vendor patch is available; the project has not responded to the responsible disclosure filed on Gitee.
Stored cross-site scripting in DataEase up to version 2.10.26 allows an authenticated attacker to inject arbitrary JavaScript through the `tooltip.backgroundColor` parameter of the Symbolic Map chart component, executing in the browser of any user who views the affected visualization. The injection point is the `buildTooltip` function in `symbolic-map.ts`, where user-controlled chart styling attributes are rendered without sanitization. No patch has been released; the vendor had not responded to the responsible disclosure issue at time of reporting, and no public exploit code has been identified.
Cross-site scripting in TDuckApp tduck-platform affects the Form Write View component (tduck-front/src/views/form/write/index.vue) through the unsanitized submitShowCustomPageContent parameter in all versions up to and including 5.3. An authenticated remote attacker can inject malicious script content into form submission custom page configurations, which then executes in victims' browsers when they load the affected form view. No vendor patch has been released; the project maintainers were notified via Gitee issue IK5RJD but have not responded.
Stored cross-site scripting in linlinjava litemall versions 1.4.0 through 1.8.0 allows a high-privileged attacker to inject malicious scripts via the `detail` argument processed by `AdminGoodsService.validate` in the Product Detail admin component. Exploitation requires admin-level credentials to write the payload and passive interaction from a second user viewing the poisoned product detail page. No public exploit has been identified at time of analysis, and the project maintainer has not responded to the responsible disclosure issue filed on Gitee.
Stored cross-site scripting in linlinjava litemall's admin topic management panel allows a high-privileged attacker to inject malicious scripts via the AdminTopicController.validate function, which are then executed in the browsers of other users who view the affected topic page. Versions 1.5.0 through 1.8.0 are affected. No vendor patch is available; the project has not responded to the responsible disclosure filed in Gitee issue IK5SUU. No public exploit code or CISA KEV listing has been identified at time of analysis.
Stack buffer overflow in alsa-lib through 1.2.16.1 allows local low-privilege attackers to crash any process consuming ALSA control-element identifiers by supplying a name= field of 64 or more characters via saved state files or command-line arguments. The off-by-one condition in __snd_ctl_ascii_elem_id_parse() (ctlparse.c:216-241) writes one byte past a 64-byte stack buffer, corrupting adjacent stack memory and causing process termination. No public exploit is identified at time of analysis; modern stack-canary mitigations further limit exploitation beyond DoS.
CyberPanel before version 2.4.4 contains a logic omission in the `checkOwnership()` access control function that allows authenticated low-privilege users to bypass domain ownership verification. The `plogical/acl.py` module's `checkOwnership()` function lacked a `return 0` in its denial branch, causing it to implicitly return `None` instead of an explicit denial signal when a requesting user does not own the target domain. An authenticated user can exploit this to access domain-level information belonging to other panel users. No public exploit or CISA KEV listing exists at time of analysis.
CyberPanel versions before 2.4.4 fail to record audit log entries when authenticated panel users interact with the child domains management interface, allowing configuration manipulation to proceed without any forensic trail. The upstream fix commit also reveals a companion access-control defect: a missing return path in the ACL ownership check (checkOwnership() in acl.py) and an ORM filter regression in fetchChildDomainsMain that broadened the child domain result set beyond intended alias-excluded entries. Together, these gaps allow an authenticated low-privilege user to alter domain list state silently, undermining post-incident accountability. No public exploit has been identified at time of analysis.
Identity spoofing in Forgejo's ActivityPub federation allows a remote attacker who controls any single valid ActivityPub actor and keypair to submit HTTP-Signature-valid activities falsely attributed to arbitrary actor identities on a target instance. Affected versions span 13.0.0 through 16.0.4, but only when federation is explicitly enabled via the non-default setting `[federation] ENABLED = true`. The vendor explicitly confirms this does not permit account takeover or content modification - impact is limited to false attribution of federated activities. No public exploit has been identified at time of analysis.
Authorization bypass in Apache Storm Logviewer 3.0.0 exposes daemon logs and tenant log file listings to any filter-authenticated user, entirely disregarding the configured `logs.users` and `logs.groups` access controls. A boolean short-circuit in the daemon log authorization path discards the authorizer's decision whenever the daemon-log flag is set, and the daemon log page and download endpoints are reached without consulting an authorizer at all; additionally, the listing endpoints `/listLogs` and `/searchLogs` accept a user argument but never apply it. On a multi-tenant Storm cluster, any user who passes the servlet filter can read nimbus.log, supervisor.log, and other daemon logs-which contain co-tenant topology names, job owners, and configuration fragments-and enumerate all worker log file names. No public exploit identified at time of analysis.
Absolute path traversal in the Zettlab D6 Ultra NAS firmware before version 1.7.0 permits authenticated network-adjacent users to navigate outside their designated personal folder and read files from arbitrary directories on the device. The vulnerability stems from insufficient sanitization of user-supplied file paths, allowing the personal-folder boundary to be bypassed using absolute path references. No public exploit code or CISA KEV listing has been identified at time of analysis; the fix is confirmed in version 1.7.0.
Command injection in LB-LINK AC1900_AZ2 router firmware V1.0.2 allows adjacent-network actors to inject OS commands through shell metacharacters passed to the bs_SetLimitCli_info function within libshare.so, triggered via a POST request to /goform/set_LimitClient_cfg. The vulnerability is notable for an internal contradiction: the CVSS vector assigns PR:H (administrative privileges required), yet the description explicitly frames the attack scenario as one where the actor lacks administrative access but can reach the endpoint - suggesting the endpoint may be accessible to lower-privilege or unauthenticated callers in certain deployments. No patch has been identified and no public exploit is confirmed at time of analysis beyond the researcher's GitHub write-up.
Link spoofing in Floodlight SDN controller (commit 71fe8a7) enables adjacent-network attackers to inject false topology information, causing the controller to misclassify a boundary port as a non-boundary port and disrupting legitimate host communication. The vulnerability targets OpenFlow-based topology discovery, requires adjacent network positioning and low privileges (AV:A/PR:L), and carries a low CVSS base score of 3.0 with a scope change (S:C) reflecting cross-component impact on the controller's global network view. No public exploit code and no active exploitation have been identified at time of analysis.
Out-of-bounds read in FFmpeg's WMA audio encoder (libavcodec/wmaenc.c) before version 9.0 allows a process crash when encoding WMA audio with crafted extradata parameters. The root cause is missing required padding bytes in the extradata buffer allocation path, permitting a read operation to extend past the buffer boundary (CWE-125). With a CVSS score of 2.9 and impact limited to availability, no public exploit identified at time of analysis, and no CISA KEV listing, this is a low-severity issue with narrow real-world exposure.
Heap-based buffer over-read in TinyEXIF before 1.1.0 allows an attacker who can supply a crafted JPEG image to crash applications that use the library for EXIF metadata parsing. EntryParser::Fetch methods compute SubjectArea IFD entry offsets without validating them against the actual buffer length, producing out-of-bounds reads when processing malformed tags. No public exploit exists and no CISA KEV listing is present; impact is limited to availability and assessed as Low severity.
Out-of-bounds read in FFmpeg's MOV container parser (mov_read_iacb in libavformat/mov.c) crashes the process when handling IACB atoms containing insufficiently padded extradata, affecting all releases before 9.0. The vendor-assigned CVSS score of 2.9 (Low) with AV:L/AC:H reflects a local, high-complexity attack limited to a process crash with no confidentiality or integrity impact. No active exploitation has been identified; an upstream fix is available in FFmpeg 9.0 via commit 8439e020 and PR 22988.
Terminal escape sequence injection in procs (the Rust-based `ps` replacement by dalance) through version 0.14.12 enables a local attacker to embed ANSI, CSI, or OSC control sequences in process argv that are rendered unfiltered into other users' terminals when they run procs. The Command column in `src/columns/command.rs` only stripped literal newlines and tabs, leaving ESC bytes and derived sequences intact; a victim running procs on the same host will have those sequences interpreted by their terminal emulator, enabling clipboard overwrite via OSC 52, screen manipulation via CSI, or hyperlink spoofing. No public exploit exists beyond the fix-commit unit test, and the CVSS 4.0 score of 2.4 (Low) accurately reflects the narrow, local-only, limited-integrity impact.
OS command injection in webgjc web_robot versions 2.4.0, 2.5.0, and 2.8.0 enables remote attackers to execute arbitrary operating system commands by supplying malicious input through the case_name parameter handled by the controller_listen or controller_recover functions in py/web.py. A publicly disclosed proof-of-concept exploit exists, the vendor failed to respond to responsible disclosure, and no patch has been released. The reported CVSS 4.0 base score of 5.3 reflects conservative impact ratings that appear inconsistent with the true scope of OS command injection on an unpatched host.
Improper privilege assignment in PHPGurukul Bank Locker Management System 1.0 allows remote authenticated attackers to bypass access controls by manipulating the `UserType` parameter in `sidebar.php`. The server fails to validate the supplied privilege level against the authenticated session, permitting lateral or vertical privilege escalation within the application. A public proof-of-concept exploit has been released per the VulDB disclosure, though the CVSS 4.0 base score of 2.1 reflects limited actual impact across confidentiality, integrity, and availability dimensions.
Privilege escalation via mass assignment in lenve vhr 1.0-SNAPSHOT allows an authenticated remote user to manipulate restricted fields - including role or privilege-level attributes - through the HrInfoController.updateHr endpoint by injecting unexpected parameters alongside the Password argument in HrMapper.xml. The vulnerability stems from improper privilege management (CWE-266), enabling a low-privilege HR system user to elevate their own account permissions without authorization. A public proof-of-concept has been disclosed; no vendor patch is available and the vendor has not responded to the disclosure.
Stored cross-site scripting in Xuxueli xxl-job up to version 3.5.0 permits an authenticated low-privileged user to inject malicious script content via the name or author parameters of the /jobinfo/insert endpoint, which is later rendered unsanitized to other users who view the job listing interface. The CVSS 4.0 score of 2.0 reflects the constrained real-world impact: exploitation requires a valid account and passive victim interaction, with impact limited to low integrity effect and no confidentiality or availability consequence. A public proof-of-concept is available, but no vendor patch exists as the vendor did not respond to coordinated disclosure.
Stored JavaScript injection through the Home Page editor in Grav 1.7.50.2 permits an authenticated administrator to embed script content that executes in a victim's browser when the page is viewed. However, NVD's own description includes a formal dispute: Grav's administrator role already grants unfettered access to modify templates, install plugins, and upload arbitrary executable content, meaning this JavaScript injection path adds no incremental attack capability beyond what the role inherently permits. NVD assigned a base score of 1.8 (AV:L/AC:H/PR:H/UI:R), and no active exploitation or public exploit code has been identified at time of analysis.
Missing authorization in Chengdu Qilu Technology Ludashi 6.1026.4715.714 allows a local high-privileged attacker to bypass protection controls enforced by the ProtectFilter64.sys kernel driver. The MessageNotifyCallback function within the Message Dispatch Handler processes messages without adequate authorization checks, enabling unauthorized manipulation of integrity and availability controls. A public proof-of-concept exploit has been released; the vendor was unresponsive to coordinated disclosure.
Information disclosure in Chengdu Qilu Technology Ludashi 6.1026.4715.714 allows a local high-privileged attacker to read arbitrary physical memory by sending a crafted IOCTL with a manipulated PhysicalAddress argument to the kernel driver ComputerZ_x64.sys. The vulnerable function sub_11008 exposes physical memory read capability to user-mode callers without adequate address validation, a classic pattern in poorly designed Windows kernel drivers that can expose sensitive in-memory data such as credentials or cryptographic material. A public exploit has been published; no vendor patch exists as the vendor did not respond to coordinated disclosure.