Header injection in the Subscribe to Comments WordPress plugin before 2.3.3 allows authenticated users with Author-level access or higher to insert CR/LF characters into the notification sender-name setting, appending arbitrary e-mail headers such as a hidden Bcc and spoofing the From address on outbound notifications. Publicly available exploit code exists, but CISA KEV does not list this issue. Impact is confined to notification e-mail rather than code execution on the WordPress host, and sites without untrusted Author-or-higher accounts are effectively not exposed.
Business-logic validation in Salon Booking System before 10.31.11 lets an unauthenticated visitor who holds a valid fixed-amount discount code submit a confirmed booking with a negative total. The issue is low severity (CVSS 3.7, AC:H) because exploitation depends on a currently valid code whose value exceeds the booking subtotal; publicly available exploit code exists, and the supplied data contains no CISA KEV confirmation. Sites that do not issue fixed-amount discount codes are not realistically exposed.
Webkul Aureus ERP through 1.6.0 fails to enforce sufficient password-strength requirements in the user creation/edit form implemented in UserForm.php, allowing accounts to be configured with weak passwords. Remote attackers without prior authentication (PR:N) can exploit this weakness through online password guessing or brute-force attempts, although exploitation is considered difficult (CVSS AC:H) and the direct impact is limited to low confidentiality. Publicly available exploit code exists, but there is no confirmed active exploitation (not in CISA KEV), and the CVSS 4.0 score is 2.9.
Cryptographically weak pseudo-random number generation in the PHP CMS Diem's dmString::random() helper can produce predictable values that may undermine tokens or identifiers derived from it. All rolling-release versions through commit 116974edfb9a5b8bd69cb13586dc62bcdbb485ad are affected; the flaw is remotely reachable by unauthenticated attackers but rated high complexity (CVSS 2.9), with impact limited to low confidentiality. Publicly available exploit code exists, and the vendor has not responded or released a fixed version.
Improper certificate validation in the Google Analytics Client bundled with the Diem PHP framework (dmCorePlugin/lib/vendor/gapi/gapi.class.php) allows an attacker positioned on the network path between a Diem server and Google's Analytics endpoint to intercept or tamper with outbound analytics traffic without detection. Unauthenticated remote attackers can exploit this, but the attack requires a man-in-the-middle position and the vendor itself describes exploitability as difficult (CVSS 4.0 AC:H); impact is limited to low integrity of analytics data with no availability or confidentiality effect. Publicly available exploit code exists (referenced in diem-project/diem issue #454), and no vendor-released patch has been identified because the project has not responded and ships as a rolling release without versioned artifacts.
A heap/stack buffer overflow in fastllm (versions up to and including 0.1.8.2) lets remote attackers corrupt memory by supplying a crafted GGUF model file whose string-length field is manipulated, triggering an out-of-bounds write inside GGUFBuffer::ReadString in third_party/gguf/gguf.cpp. Exploitation requires the victim to load or parse the malicious model (user interaction, CVSS UI:P/UI:R), and no authentication is needed; impact is limited to partial confidentiality, integrity, and availability loss. Publicly available exploit code exists (a POC archive is linked alongside the VulDB entry and upstream GitHub issue 759), but there is no CISA KEV listing, so active in-the-wild exploitation is not confirmed; the vendor has not yet responded to the report.
A path traversal (Zip Slip) flaw in kiwilan/php-archive through version 2.3.02 lets an attacker craft a ZIP archive whose entry names escape the intended extraction directory when ArchiveZip::extract() is called, allowing files to be written to arbitrary locations the PHP process can reach. The library is a dependency, not a network service, so exploitation requires the host application to pass an attacker-influenced archive to extract() and to run with filesystem permissions that make the overwritten paths meaningful. Publicly available exploit code exists (a proof-of-concept gist is referenced alongside the VulDB entry), but there is no evidence of active exploitation in the wild, and the vendor did not respond to the disclosure.
OS command injection in JumpServer through 4.10.19 lets an authenticated user with permission to create or edit a Host asset store a crafted IP/HOST address containing Jinja2 template syntax, which the Template Rendering component later evaluates and can use to run arbitrary OS commands. Exploitation requires a valid low-privileged account and a passive interaction step where the poisoned address is rendered, so it is realistically reachable by insiders or attackers who already hold credentials rather than by anonymous remote users. Publicly available exploit code exists (a public gist is referenced) and the vendor has fixed the issue in v5.0.0, with the fix also slated for an upcoming v4 release.
Open redirect in aircheng-org iWebShop-5 through 5.16 lets remote attackers craft a link that sends victims to an attacker-controlled site via the callback argument of IController::redirect in plugins/_userInfo/_userInfo.php. No authentication is required, but the victim must interact with the crafted URL (UI:P), and publicly available exploit code exists. The impact is limited to browser redirection for phishing or trust abuse, with no server-side code execution or data disclosure.
Unsafe deserialization in hankcs HanLP through 2.1.0-beta.64 allows unauthenticated remote attackers to achieve code execution when a victim loads an attacker-supplied model or config in which the classpath argument passed through hanlp.load() has been manipulated, reaching the reflection logic in plugins/hanlp_common/hanlp_common/reflection.py. Exploitation is not automatic against an idle installation: it requires the user to obtain and load an untrusted model, so exposure is concentrated in pipelines that consume third-party or externally sourced models rather than those loading only first-party artifacts. Publicly available exploit code exists (a PoC archive is linked from the VulDB entry), but there is no CISA KEV confirmation of active exploitation; the project was notified via GitHub issue #1969 and has not responded at time of analysis.
Double free in Artifex MuPDF (through 1.28.5) lets a crafted PDF crash the document parser by triggering the Signature Locking code path that reaches free_char_list in source/pdf/pdf-xref.c. The flaw is remotely reachable but requires the victim to open a malicious file (UI:P in the CVSS 4.0 vector), so exploitation hinges on user-initiated file opening rather than a network-facing service. Publicly available exploit code exists - the Ghostscript bug tracker hosts a working proof-of-concept attachment - though there is no CISA KEV entry confirming active exploitation in the wild; impact per all assessed metrics is availability-only (process crash / DoS), with no code execution, data disclosure, or integrity loss indicated.
Unsafe reflection in Unsloth's FastSentenceTransformer class (versions up to 0.1.501-beta) lets an attacker who supplies a crafted model, tokenizer, or configuration artifact steer class selection inside `_is_transformer_module_ref`, giving limited confidentiality, integrity, and availability impact. The CVSS vectors mark PR:N, so no credentials are needed, but UI:R/UI:P means the victim must load or process the attacker-influenced artifact - this is not exploitation of an exposed network service running with default configuration. Publicly available exploit code exists; the issue is fixed in Unsloth 0.1.902-beta.
Arbitrary Python object instantiation in ESPnet's espnet3 Inference Model Handler allows an attacker who supplies a crafted model bundle to execute code on the machine that loads it, with assessed impact of high confidentiality, integrity and availability loss. The bundle's inference config carries an attacker-chosen `_target_` value that Hydra's instantiate() resolves and calls, including targets nested inside arguments, and pre-fix ESPnet applied no namespace restriction to what could be built. Exploitation requires the victim to obtain and load the untrusted bundle (user interaction, no authentication required), publicly available exploit code exists, and ESPnet versions up to 202610.post2 are affected with an upstream fix merged as PR #6829; note that the published CVSS 4.0 base score of 2.1 reflects a low-impact/threat-adjusted rating and is materially lower than the assessed CVSS 3.1 vector of AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H.
Code injection in the DeepPavlov conversational AI framework (versions up to and including 1.7.0) allows an attacker who can supply or influence a pipeline configuration file to have arbitrary code executed when that configuration is resolved. The flaw sits in the resolve() function of deeppavlov/core/common/params.py, which turns config reference strings into imported Python objects; per the independent assessment the realistic attack path is local and requires victim interaction (CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H), i.e. the victim must load a crafted config rather than being reachable autonomously over the network. Publicly available exploit code exists (a proof-of-concept archive is linked from the GitHub advisory trail), but there is no CISA KEV listing and no evidence of exploitation at scale, so the real-world priority remains low despite the high CIA impact once a malicious config is loaded. The vendor was notified through GitHub issue #1713 and has not responded.
Code injection in Llama-Chinese's Gradio chat demo (examples/chat_gradio_no_merge.py) lets unauthenticated remote attackers run arbitrary code when an operator loads an attacker-influenced value for base_model_name_or_path. Exposed deployments of the example script are the realistic attack surface; the project uses a rolling release with no disclosed version and has not responded to the issue report. Publicly available exploit code exists, but there is no CISA KEV confirmation of active exploitation.
Manipulating the user_id argument in iWebShop-5 up to 5.16's send_message_mobile function (controllers/simple.php) can abuse weak password-recovery logic and potentially allow unauthorized password recovery for another account. The flaw is remotely reachable with low impact, publicly available exploit code exists, and independent assessment rates it a moderate, low-priority issue rather than an urgent one.
Reflected cross-site scripting in the Doctor Registration component (doctorregistration.php) of subhajitkhan's online-clinic-management-system allows remote attackers to inject script into a victim's browser session by manipulating the doc_name parameter, which the application echoes back without encoding. No authentication is required of the attacker, but exploitation depends on a victim loading a crafted request - the assessed vector is UI:R, so this is not a zero-click flaw - and impact is confined to the victim's session with no confidentiality or availability loss. Publicly available exploit code exists and the flaw has been published by VulDB, but it is not confirmed actively exploited (CISA KEV); real-world risk is assessed as LOW despite the MEDIUM 5.3 base score, chiefly because of the required user interaction and the project's negligible deployment base.
Deserialization of untrusted data in HanLP through 2.1.0-beta.64 allows an attacker who supplies a crafted vectors.vocab file to execute arbitrary code when a victim loads it with load_word2vec_as_vocab_tensor. The attack can be delivered remotely but requires the victim to open or load the malicious model file, so user interaction is the key prerequisite; no authentication or special configuration is needed. Publicly available exploit code exists, but the supplied intelligence does not confirm CISA KEV active exploitation.
An open redirect in the Diem PHP framework's Sign-In component (dmCorePlugin dmUserPlugin) allows remote attackers to redirect users to arbitrary external sites after authentication by manipulating the Referer argument in BasedmUserActions.class.php. Exploitation is unauthenticated but requires the victim to follow a crafted link (UI:P), and the impact is limited to integrity of the redirect destination, with no server compromise or data disclosure. Publicly available exploit code exists, though no confirmed active exploitation has been reported; the vendor has not responded to the issue report and follows a rolling release model, so no fixed version is specified.
SQL injection in itsourcecode Online Cake Ordering System 1.0 allows authenticated remote attackers to manipulate the 'cancel' argument in confirm.php, potentially exposing or altering limited data. A public exploit is available, but the vulnerability requires a low-privileged session and has low impact, and the product's negligible production presence keeps real-world risk low.
Code injection in RVC-Project Retrieval-based-Voice-Conversion-WebUI through 2.3.260718 lets unauthenticated remote attackers who can reach an exposed Gradio interface manipulate the ckpt_path argument passed to change_info_ in webui.py and execute attacker-controlled code, although user/operator interaction is required and default deployments bind to localhost. Publicly available exploit code exists, but there is no CISA KEV listing and no EPSS score was provided, and the project has not responded to the issue report. Practical risk is concentrated on instances explicitly exposed via Gradio --listen/--share or a reverse proxy/tunnel, where the lack of default authentication makes the code-injection path reachable.
Integer overflow in Tencent ncnn's YUV420SP-to-RGB conversion API (ncnn::yuv420sp2rgb in src/mat_pixel.cpp) lets unauthenticated remote attackers who can get a target application to process crafted YUV image dimensions or strides trigger an undersized buffer allocation and subsequent out-of-bounds writes. The flaw affects ncnn builds up to commit 44b02dabdcf97499281a4692ebd65bad73dcf8b6 and requires the application to process attacker-controlled YUV420SP data (UI:P); realistic outcomes are denial of service or bounded heap corruption with Low confidentiality, integrity, and availability impact rather than reliable remote code execution. Publicly available exploit code exists, and the vendor has not responded to the upstream issue report.
Out-of-bounds reads in Tencent ncnn's MatMul layer can disclose process memory or crash inference workloads when a crafted model drives matmul_transb with inconsistent dimensions. Affected builds are those at or before commit 44b02dabdcf97499281a4692ebd65bad73dcf8b6, and exploitation requires the victim to load and run an attacker-supplied model (UI:P); no KEV entry exists, but publicly available exploit code exists. The issue is genuine but moderate severity, with impact bounded to memory disclosure and possible denial of service rather than code execution.
Improper authorization in Donetick's chore history deletion handler (versions up to and including 0.1.82) lets any authenticated member of a Circle delete chore history belonging to other members by manipulating the AssignedTo argument passed to CanDeleteHistory. The flaw is limited in scope: the attacker must already hold an account inside the victim's Circle, so it is not reachable by unauthenticated outsiders and does not cross circle boundaries, and the impact is confined to integrity and availability of history records with no data disclosure. Publicly available exploit code exists and the issue is documented in the upstream repository (issue #711) and GitHub advisory GHSA-c545-2whr-g2v6, though there is no evidence of confirmed active exploitation (CISA KEV) at the time of analysis; the maintainer disputes the IDOR framing, characterizing the behavior as a bug affecting trusted collaborators rather than a security vulnerability.
SQL injection in Dav-ee Dairy Management System allows authenticated low-privileged users to manipulate the f_no or farmer parameter in payment/pay.php, potentially reading or altering payment and farmer records. The flaw stems from improper use of mysqli_real_escape_string, and publicly available exploit code exists; active exploitation is not confirmed by CISA KEV. Exploitation requires a valid low-privilege session and access to the payment feature, which limits exposure to authenticated staff accounts rather than unauthenticated internet users.
SQL injection in Dav-ee Dairy Management System (farmers/add.php) lets an authenticated low-privileged user inject arbitrary SQL through the f_no, f_id, f_name, f_locallity, f_ac, or f_phone parameters, which are passed unsanitized into mysqli_query. The product is not versioned, so all deployments of the identified commits are presumed affected, and publicly available exploit code exists. Real-world risk is LOW despite a public exploit: exploitation requires a valid low-privileged session with access to the add-farmer form, and the product has niche deployment; there is no indication of confirmed active exploitation (no CISA KEV listing).
Cross-site request forgery in aircheng-org iWebShop-5 through 5.16 allows an attacker to change an administrator password when a logged-in admin is induced to load attacker-controlled content; the affected code is admin_repwd_act in controllers/system.php. Publicly available exploit code exists (GitHub issue #9), but CISA KEV does not confirm active exploitation, and impact is limited because the attacker must rely on a live admin session and gains no direct access. The vendor has not responded and no fixed version is identified.
Open redirect in antoinersx clawhost through version 0.0.45 allows remote attackers to craft reverse-proxy URLs that send victims to attacker-controlled sites by manipulating the token argument handled in reverseProxy.ts. Exploitation requires user interaction - a victim must follow the crafted link - and impact is limited to integrity (phishing/credential-harvesting), consistent with CVSS 5.3 v4.0 and an assessed CVSS 3.1 score of 4.3. Publicly available exploit code exists, but there is no confirmed active exploitation (not CISA KEV).
Hard-coded MQTT broker credentials in D-Link EAGLE PRO AI routers (E15, M32, R15, R32) let remote attackers who can reach the broker service authenticate with publicly documented credentials and publish or subscribe to MQTT topics, potentially tampering with device and mesh communications. Publicly available exploit code and credential evidence exist, but exploitation depends on network reachability to the MQTT service, which is typically LAN- or mesh-side rather than internet-facing; the assessed impact is Low across confidentiality, integrity, and availability.
Code injection in Doctor-Dignity's MLC Chat module (_get_lib_module in python/mlc_chat/chat_module.py) allows an attacker who controls the model_lib argument to execute arbitrary code when a victim loads the attacker-supplied model library; publicly available exploit code exists, but there is no CISA KEV entry confirming active exploitation. The independent assessment rates real-world priority LOW because the attack is local (AV:L) and requires user interaction (UI:R) to load a malicious model configuration, rather than reaching an exposed network service.
Heap/stack buffer overflow in ztxz16 fastllm through 1.8.2, triggered inside SafeTensorItem::CreateBuffer (src/model.cpp) when a crafted safetensors file supplies manipulated data_offsets values during model weight loading. An attacker who convinces an operator to load an untrusted model file can cause out-of-bounds memory writes, yielding limited confidentiality, integrity, and availability impact; publicly available exploit code exists (VulDB submission with a poc.zip attachment) but there is no evidence of active exploitation in the wild. The vendor has not responded to the upstream issue report, so no fix has been published, and exploitation requires user interaction - the victim must explicitly load the malicious weights - which keeps this a moderate, low-priority supply-chain/untrusted-model-file issue rather than a remotely reachable server compromise.
Code injection in AXA Group NLP.js 4.27.0 allows remote attackers to execute arbitrary JavaScript via the Evaluator.walkFunction condition-expression pipeline when an embedding application passes attacker-controlled input into the condition argument. Public exploit code exists, but the flaw is not reachable in a default standalone installation; it requires the host application to expose the condition-expression evaluation path, and the project has not responded to issue #1380, so no vendor-released patch was identified at time of analysis. Although the published CVSS 4.0 score is 2.1 and the RCE/code-injection tags draw attention, practical priority is low for typical deployments because exploitation depends on that application-level exposure.
Cross-site WebSocket hijacking and request-forgery in BlackBeltTechnology pi-agent-dashboard through 0.8.0 lets any web page the victim visits open an unauthenticated WebSocket to the loopback dashboard (ws://127.0.0.1:8000/ws), receive the session snapshot, spawn a terminal, and blind-POST to /api/* and /auth/logout using the victim's ambient trust, because the gateway validates neither Origin nor a CSRF gate on those paths. Exploitation requires the dashboard to be running locally and the user to browse an attacker-controlled page on the same host (UI:P/UI:R); no credentials are needed (PR:N), but the pre-patch impact is limited to low integrity (VI:L / CVSS 4.0 base 5.3 Medium) since CORS only ever hid responses, not requests. Publicly available exploit code exists (a GitHub issue tracking the flaw and an upstream fix PR), and the vendor shipped a patched release; no CISA KEV entry has been recorded, so active in-the-wild exploitation is not confirmed.
Path traversal in the H3C Magic NX15 router (firmware V100R017) allows a user who already holds administrative rights on the device's management interface to write files outside the intended import location by supplying directory-traversal sequences in the fileName argument of the importprofile function at /api/upload. The flaw is network-reachable and needs no user interaction, but the CVSS vector specifies PR:H, so it is a post-authentication issue that presupposes privileged router credentials rather than an unauthenticated remote compromise; impact is limited to low confidentiality, integrity and availability effects, which is reflected in the low base score. Publicly available exploit code exists - a GitHub repository documenting the arbitrary file write against NX15R017 has been published - but exploitation has not been confirmed as active (no CISA KEV listing), and the overall risk is low-priority despite remote reachability.
Code execution in DeepPavlov through 1.7.0 can occur when a deployment loads a pipeline configuration that lets an attacker control the preprocessor argument passed to MultiTaskPipelinePreprocessor.__init__, causing attacker-supplied code to run with the privileges of the DeepPavlov process. Exploitation requires the ability to supply or alter the loaded model/pipeline configuration and low privileges, so it is not a drive-by unauthenticated remote attack; a user or service must load and execute the malicious configuration. Publicly available exploit code exists, but there is no CISA KEV confirmation of active exploitation, and the issue remains unresolved upstream.
SQL injection in Dav-ee dairy-management-system's employee deletion handler (employees/delete.php) allows a high-privileged authenticated user to supply a malicious e_payroll_no value that is passed unsanitized into mysqli_query. The issue is remotely reachable with low complexity and no user interaction, but the PR:H requirement limits exploitation to privileged sessions, and impact is assessed as low across confidentiality, integrity, and availability. Publicly available exploit code exists, but no CISA KEV entry or other evidence of active exploitation was provided; affected releases cannot be pinned because the project uses a rolling release model, and no vendor patch has been identified.
Stored cross-site scripting in SourceCodester Inventory System 1.0 is a genuine-but-low-priority issue: an authenticated low-privilege user can inject script through the Product_Name, Product_Code, or Vendor_Name parameters when saving a purchase via /api/transactions_handler.php?action=save_purchase, and the payload executes when another user views the affected purchase record. Publicly available exploit code exists, but there is no CISA KEV confirmation of active exploitation, and the product's limited deployment base reduces practical risk. The independent assessment scores this as CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:L/I:L/A:N, requiring a valid account and victim interaction.
Cross-site scripting in SourceCodester Inventory System 1.0 allows an authenticated low-privileged user to inject script through the Product_Id or Product_Name argument in /api/products_handler.php?action=save, with execution occurring when another user later views the Product Management page. Exploitation requires valid credentials and victim interaction, publicly available exploit code exists, and the CVSS 4.0 score is 5.1; impact is limited to browser-session integrity with no confidentiality or availability loss.
Cross-site scripting in the dmWidget link-rendering component of Diem CMS allows authenticated low-privilege users to inject script through the text argument of dmBaseLinkTag::text in dmCorePlugin/lib/view/html/link/dmBaseLinkTag.php, with execution occurring in a victim's browser after the victim views or interacts with the crafted link. Deployments tracking Diem's rolling release at or before commit 116974edfb9a5b8bd69cb13586dc62bcdbb485ad are potentially affected, and publicly available exploit code exists; the reporter scores the issue CVSS 4.0 5.1, while the independent assessment places it at CVSS 3.1 5.4-equivalent impact with PR:L/UI:R. Exploitation is limited by the need for an authenticated account permitted to supply the widget's text argument and by the requirement for victim interaction.
Use-after-free in Mini-XML 4.0.5 and 4.0.6 is a low-priority, high-noise finding rather than a genuine emergency: when a consuming application passes attacker-controlled XML into mxml_load_data and the malformed-input error path is reached, freed memory can be reused and the process can crash, producing denial of service rather than confirmed code execution. Publicly available exploit code exists, and a community patch plus a POC archive have been disclosed, but the upstream project has not responded and no official vendor patch is confirmed. The independent assessment is remote with user/application interaction required (CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:L); the CNA's CVSS 4.0 vector is lower (2.0, PR:L/UI:P), reflecting that exposure depends on an application exposing Mini-XML parsing to untrusted input.
OS command injection in the FusionPBX Dialplan Editor allows an authenticated user with dialplan-edit rights to run arbitrary operating-system commands through manipulated values processed by app/dialplans/resources/classes/dialplan.php. The issue affects FusionPBX through 5.4.4, 5.5.7, and 5.6.5; publicly available exploit code exists, and no vendor patch or response was referenced at time of analysis. Because dialplan management is an administrative function, exploitation requires high privileges, but multi-tenant or delegated-admin deployments broaden the exposure.
Path traversal in Diem through 5.1.3 lets an authenticated user with high backend privileges abuse the dmCodeEditor action's dmOs::join function to access files outside the intended directory; the issue is rated CVSS 4.0 5.1 and publicly available exploit code exists, though exploitation is gated by the high-privilege prerequisite. The project was notified via GitHub issue #451 and has not responded, and no vendor-released patch is identified.
Cross-site scripting in the Diem PHP CMS administrative console lets a user who already holds high-privilege console access inject script through the dm_command argument processed by the append() function in dmAdminPlugin/modules/dmConsole/actions/actions.class.php, with the payload executing in the browser of a console operator who views the crafted page. Diem is delivered as a rolling release with no version metadata, so current checkouts at or around commit 116974edfb9a5b8bd69cb13586dc62bcdbb485ad remain unpatched, publicly available exploit code exists, and the project has not responded to the report. This is a low-to-moderate real-world priority despite the medium CVSS 4.0 base score of 4.8, because the vulnerable path sits behind high-privilege console authentication and the assessed impact is limited to low integrity with no confidentiality or availability effect.
Cross-site scripting in the Diem admin plugin's Version Check Service lets a high-privileged authenticated attacker supply markup that the html handler in dmAdminCtrl.js renders without proper escaping, executing script in a victim administrator's browser when the affected view is opened. The impact is limited to low integrity per the independent CVSS:3.1 assessment (C:N/I:L/A:N), and exploitation also requires victim interaction, so the issue is network-reachable but gated by administrative privileges and user action. Publicly available exploit code exists, but there is no CISA KEV confirmation and overall risk is low, especially because Diem uses a rolling release with no vendor-issued fixed version.
Token-less deployments of MemTensor's memmy-agent Memory Service expose POST /api/v1/admin/shutdown to unauthenticated remote requests, letting an attacker shut the service down and cause a low-impact denial of service. All versions up to 1.0.9, 1.1.1 and 1.2.2 are listed as affected (CPE: cpe:2.3:a:memtensor:memmy-agent:*), and the issue is rated only 2.3 under CVSS 4.0 (with our own assessment of 3.1/AV:N/AC:H/PR:N/UI:R/A:L) because exploitation is gated by a victim visiting an attacker-controlled page, by specific browser cross-origin behavior, and by the admin token being left unconfigured. Publicly available exploit code exists (the upstream GitHub issue and VulDB entry), but the practical risk is limited to internet-exposed instances running in the default, token-less state - the endpoint is designed for local use and the cross-origin behavior was reportedly tightened in later changes.
Quadratic processing in markdown-it (prior to 14.3.3 and 15.0.3) allows attacker-controlled Markdown to block the event loop or freeze a browser main thread when HTML parsing is enabled. Only deployments instantiated with the non-default 'html' option or the 'commonmark' preset are exposed; the shipped default disables inline HTML, and HTML blocks are unaffected - the payloads must sit inside a paragraph. Both a public advisory and upstream commits/patched releases exist, but no public exploit code and no confirmed active exploitation were identified at time of analysis, and EPSS data was not supplied; this is a low-priority defect whose severity is capped by the required non-default configuration.
Cross-origin resource sharing (CORS) misconfiguration in IBM Concert 1.0 through 3.0.0 can allow attacker-controlled web content to carry out privileged actions in the context of a victim's authenticated session when the allowed domain is not limited to trusted domains. The vendor rates attack complexity High and impact limited, and the CVSS vector indicates no prior privileges are required (PR:N) even though practical exploitation depends on a victim with an active Concert session loading malicious web content. IBM has released a fix; no public exploit was identified at time of analysis.
Improper certificate validation in IBM Concert 1.0 through 3.0.0 allows an attacker who already holds a man-in-the-middle position to present forged or invalid TLS certificates that the product may accept, yielding low confidentiality and integrity impact with no availability effect. The provided CVSS 3.7 vector requires high attack complexity and user interaction, and the supplied vectors conflict on whether low-level privileges (PR:L) are required, so authentication requirements are not consistently confirmed from the available data; no public exploit code or CISA KEV entry was identified at time of analysis. This is a low-severity issue and the signals are consistent in pointing to limited real-world risk.
IBM Concert 1.0 through 3.0.0 can expose highly sensitive information because it relies on weaker-than-expected cryptographic algorithms, allowing an attacker who obtains the protected data to attempt to recover plaintext. The CVSS vector (AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:N/A:N) describes a network-reachable, unauthenticated and no-interaction path, but exploitation is high-complexity and limited to confidentiality impact. No CISA KEV entry and no public exploit were identified at time of analysis, so this is a low-priority cryptography hardening issue rather than an actively exploited intrusion vector.
Server-side request forgery in alexta69 MeTube (up to 2026.09.28) allows a remote, unauthenticated attacker who can influence a playlist or channel feed to make the instance fetch an arbitrary thumbnail URL - including redirects - from its own network position, writing the internal HTTP response into the download directory that /download serves back. The flaw sits in DownloadQueue.__write_feed_metadata_sync (app/ytdl.py) and is only reachable when metadata/thumbnail writing is enabled (writeinfojson/writethumbnail); exploitation is rated highly complex (AC:H, CVSS 4.0 base 2.9) and impact is limited to internal-response disclosure with no code execution. Publicly disclosed proof-of-concept exploitation details exist (CVSS E:P), and the issue is fixed in release 2026.09.29; it is not listed in CISA KEV, so there is no confirmation of active in-the-wild exploitation.