Termix versions 1.8.0 through 2.5.1 expose server-stats API routes that lack per-host authorization, allowing any authenticated user to read host online/offline state and lastChecked timestamps for hosts they cannot access and to clear the instance-wide SSH connection pool. Because GET /status/:id accepts an attacker-supplied numeric host identifier, a low-privileged account can enumerate host IDs (IDOR), while POST /clear-connections disrupts other users' active sessions and pooled connections. Unauthenticated requests remain blocked and no public exploit code has been identified at time of analysis; CVSS is 5.4 (AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:L) and the vendor fixed the issue in 2.5.1.
Code injection in DbGate (a web-based database GUI) versions up to 7.2.5-beta.5 lets an authenticated user who can create or edit a JSON Runner or scheduled script inject text into the comment.text or script.schedule fields that escapes the generator's `//` comment wrapper and reconstitutes a `// @require <module>` or `// @schedule` directive, causing the server to load and execute attacker-supplied code. The flaw is remote, requires low privileges (PR:L) and no user interaction, and is limited to unpatched releases; it is fixed by upgrading to 7.2.5-beta.6 (upstream commit 70e7b6b58e464d7a015ba16e8d7574b420ee4877, PR #1527). There is no confirmed active exploitation (not in CISA KEV) and no public exploit identified at time of analysis, though the EPSS-style exposure is mitigated mainly by the requirement for an authenticated account with runner-editing rights. This issue is explicitly distinct from CVE-2026-47668.
TREK, a self-hosted collaborative travel planner, returns complete trip itinerary data - day plans, assignments, notes, place names, coordinates, addresses, descriptions and prices - from GET /api/shared/:token even when the trip owner has turned off the share_map setting. The server-side getSharedTripData function in server/src/services/shareService.ts never enforces that toggle; only the web client hides the map, so anyone who queries the JSON endpoint directly with the valid random share token can read the location and route information the owner explicitly chose to withhold. The exposure is read-only (no data modification is possible), affects all versions prior to 3.3.0, and requires possession of the unguessable token rather than anonymous scanning; a vendor patch is available in 3.3.0, and no public exploit code has been identified at time of analysis.
Information disclosure in OpenStack Swift deployments that run the shipped default proxy pipeline: the tempurl middleware does not reject an attacker-supplied X-Copy-From header on signed PUT requests, so anyone holding a valid, unexpired PUT TempURL for a single object can make the copy middleware pull another object from the same account into the destination path and then read that data back using a GET TempURL for the destination. The impact is bounded rather than systemic - copies across account boundaries are rejected, deployments without the copy middleware in the pipeline are unaffected, and the CVSS score is 5.3 with a PR:L vector reflecting the need to hold a legitimate signed TempURL capability. No public exploit code has been identified at time of analysis, and CVE-2026-97149 is not reported as confirmed actively exploited (CISA KEV).
Playlist enumeration in Black Candy through 3.2.1 allows any authenticated user to retrieve the names of every playlist hosted on the instance, including those belonging to other accounts. The playlist search code paths (SearchController and Search::PlaylistsController) do not scope results to the session user, so blank or targeted query parameters return cross-user playlist names; exploitation requires a valid account (CVSS 4.0 PR:L, independently assessed CVSS 3.1 5.3 with C:L/I:N/A:N) and network reachability to the web interface. Exposure is confined to playlist names - no playlist content, credentials, integrity, or availability impact - and no public exploit has been identified at time of analysis.
Heap out-of-bounds write in libconfini 1.16.4 causes deterministic denial of service in applications that follow the library's bundled load_ini_buffer.h allocation pattern, which sizes the INI buffer to exactly ini_length bytes while strip_ini_cache() unconditionally writes a NUL terminator at ini_source[ini_length]. Any non-empty INI input is sufficient to trigger the resulting single-byte heap corruption, so exploitation requires no special payload or timing. The CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L, 5.3) frames the issue as unauthenticated and remotely reachable, but real-world network exposure depends entirely on whether the host application actually parses attacker-supplied INI content; no public exploit code was identified at time of analysis, there is no CISA KEV entry, and EPSS is only 0.15% (3rd percentile), supporting a low-priority availability assessment with no demonstrated code execution.
Unauthenticated path traversal in Docmost versions prior to 0.80.1 allows a remote attacker to read files from local storage by supplying attacker-controlled fileName path segments to the public avatar and logo image endpoint, which resolves them outside the intended image directory. Disclosure is constrained to local-storage objects whose final basename satisfies the route's UUID check, so arbitrary system files cannot be read; the assessed impact is confidentiality-only (CVSS 5.3) and the flaw requires a deployment using the local filesystem storage driver rather than external/object storage. No public exploit code has been identified at time of analysis and the issue is not confirmed as actively exploited, though the low attack complexity and lack of authentication make prompt upgrading to 0.80.1 advisable.
The MongoDB Python Driver (PyMongo) versions 3.9.0 through 4.18.1 can be induced to open connections to local Unix domain sockets on the application host when an attacker who already holds write access to the encryption key metadata in the database alters a key management (KMS) endpoint value so that it ends in '.sock' - a string the client-side field level encryption (CSFLE) code path mistakes for a local socket path instead of a remote host. Exploitation requires the opt-in, non-default CSFLE feature to be in use and an authenticated attacker with database-side write access to the key vault, and it is only reachable over the network in the sense that CVSS scores the vector AV:N/PR:L; the transmitted bytes are restricted to the beginning of a TLS handshake, so no attacker-chosen content reaches the socket and data exfiltration or arbitrary remote-host reachability is not achievable. No public exploit code has been identified at time of analysis and the issue is not in CISA KEV; despite the CWE-918 (SSRF) classification, this is a low-priority, low-impact defect. Vendor-released patch: 4.18.2.
Termix 2.4.0 through 2.5.0 permits an attacker who already holds a victim's authenticated session and knows the account password to disable TOTP two-factor authentication or regenerate (and thereby invalidate) the victim's backup codes through POST /users/totp/disable and POST /users/totp/backup-codes, degrading the account to single-factor authentication. The defect is a regression of the fix originally shipped for CVE-2026-45749: a 2.4.0 refactor made verifyTotpReauth accept a bcrypt-validated password as an interchangeable substitute for a TOTP code, so no authenticator or valid second factor is needed. Scored CVSS 5.3 (AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:H/A:N) and tracked as EUVD-2026-86120, this is a conditional security-degradation flaw rather than a high-urgency issue - the attacker cannot reach the endpoints without an active authenticated session (PR:L) and must separately know the plaintext password (driving AC:H), and the outcome is a weakened authentication posture (integrity impact only) with no confidentiality or availability impact. Vendor-released patch: 2.5.1; no public exploit code or confirmed active exploitation was identified at time of analysis.
Authenticated path traversal in DbGate's files-style API endpoint lets a low-privileged remote user read files outside the intended data directory by manipulating the filePath/uri argument passed to fs.readFile. The issue affects self-hosted web/server deployments running versions up to 7.2.5 and 7.3.1-premium-beta.1 where the packages/api controller is network-reachable; single-user desktop/Electron installs without an exposed API are largely out of scope. No public exploit code has been identified at time of analysis and the flaw is not in CISA KEV, but the vendor was contacted during disclosure and did not respond, so no official fix is confirmed. Overall this is a moderate-risk information-disclosure bug (assessed CVSS 3.1 6.5, AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N) whose real impact depends on which files the DbGate process can read.
A path traversal flaw (CWE-22) in HAVELSAN Liman MYS versions 2.3.2 up to but not including 2.3.4-1124 allows remote, unauthenticated attackers to read files outside the intended web-accessible directory, resulting in limited confidentiality exposure. Per the CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N, base 5.3), no credentials, privileges, or user interaction are required, so any reachable instance of an affected version is directly exposed to attempts; the impact is confined to partial file disclosure with no write, integrity, or availability effect. No public exploit code was identified at the time of analysis and there is no CISA KEV listing, so exploitation is not confirmed to be active, but the low attack complexity and lack of authentication make the issue practically attractive against internet-exposed Liman MYS deployments. This is a genuine but modest-priority issue.
Workspace ADMIN accounts in Docmost prior to 0.90.1 can mint OWNER-level accounts through the workspace invitation flow, because the invite path never enforces a role ceiling that would stop an administrator from granting privileges above their own. An attacker who already holds the ADMIN role can invite an external email address with the OWNER role; once that invitation is accepted, the resulting account holds the highest privilege level in the workspace, enabling a persistent backdoor OWNER account or the elevation of a colluding external user. Exploitation is access-gated rather than anonymous - it requires an authenticated HIGH-privilege (ADMIN) account and the invitee must accept the emailed invitation (PR:H, UI:R) - and no public exploit code or confirmed active exploitation was identified at time of analysis.
Stored cross-site scripting in Aureus ERP through 1.6.0 allows an authenticated user to upload an SVG file containing embedded JavaScript that is written to the application's public disk and served from the application origin, so the script executes in a victim's browser session when the uploaded file's direct URL is opened. Successful exploitation enables theft of session cookies and CSRF tokens, giving the attacker the ability to act as the victim within the ERP; independent assessment rates this CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:L/I:L/A:N (vendor score 5.1), reflecting a cross-site scripted response that requires a second user to open the crafted file URL. Exploitation prerequisites are non-trivial - the attacker must hold an authenticated account with access to a file-upload field and the victim must navigate to the stored file - and no public exploit identified at time of analysis, so this is best handled as a standard patch-cycle item rather than an emergency.
Stored client-side template injection in Krayin CRM through 2.2.6 allows an authenticated user with rights to create or edit web forms to plant Vue.js double-brace template expressions in the form description field, which the Vue template compiler renders without sanitization and which ultimately resolves to arbitrary JavaScript execution in the application origin for any user who views the crafted form. Exploitation requires an authenticated foothold (CVSS:4.0 PR:L / CVSS:3.1 PR:L) plus a separate victim passively rendering the form (UI:P / UI:R), so it is a stored, second-order attack rather than a self-contained remote exploit; impact is confined to client-side script execution in other users' sessions (SC:L/SI:L) with no direct server-side compromise. Severity is moderate - CVSS:4.0 5.1 and the independent assessment's CVSS:3.1 5.1 - and no public exploit code was identified at time of analysis; the issue was reported by VulnCheck and is documented in a public CVE disclosure repository and vendor advisory.
Stored client-side template injection in Krayin CRM through 2.2.6 lets an authenticated user who can create or edit products plant Vue.js template expressions in the product name field, which are then compiled and executed as JavaScript in the browser of any other user who opens the affected product record. Exploitation requires a valid CRM account with product create/edit rights (PR:L) and completes only passively, when a victim views the record (UI:P); impact is confined to the victim's browser and the application's origin, with no server-side compromise or availability effect. No public exploit code was identified in the available intelligence at time of analysis, and there is no evidence of confirmed active exploitation, so this is a genuine but moderate-priority finding rather than a high-severity one.
Stored client-side template injection in Krayin CRM through version 2.2.6 lets any authenticated user who can create or edit a person record plant Vue.js double-brace expressions in the name field, which the client-side Vue template compiler later evaluates in the browser of any other user who opens that record. Successful exploitation results in arbitrary JavaScript execution in the Krayin application origin, allowing theft of data or session material available to the viewing user, but it requires a low-privileged authenticated account to persist the payload and a second user to view the poisoned record, keeping the practical risk at medium severity (vendor CVSS 4.0 5.1; independent assessment CVSS 3.1 5.1). Impact is confined to the victim's browser on the CRM origin with no direct server-side code execution, and there is no public exploit identified at time of analysis and no confirmed active exploitation.
Root-level attackers inside an IBM PowerVM logical partition can corrupt the partition firmware runtime boot configuration stored in NVRAM, leaving that partition unable to boot until an operator manually deletes and recreates the partition configuration. The issue affects PowerVM firmware levels FW1120.00 through FW1120.01, FW1110.00 through FW1110.31, FW1060.00 through FW1060.81, and FW950.00 through FW950.H3, and requires pre-existing root privileges within the target partition (CVSS 3.1 AV:L/PR:H, base score 5.1) with no remote or unauthenticated path to exploitation. There is no public exploit identified at time of analysis and no confirmed active exploitation; the practical risk is post-compromise sabotage or persistence against the attacker's own partition rather than an initial-access vector.
Denial of service in IBM PowerVM Hypervisor firmware affects logical partitions running affected FW1120, FW1110, FW1060, or FW950 levels: an attacker who already holds root inside an LPAR can send a crafted RTAS firmware-to-OS request and crash that partition, making it unavailable. The vulnerability is local-only, requires high privileges, and is contained to the attacker's partition, so other LPARs on the same managed system are not affected. No public exploit was identified at time of analysis, and the CVSS 3.1 score is 5.1 (AV:L/AC:L/PR:H/UI:N/S:U/C:N/I:L/A:H).
Stored client-side template injection in Krayin CRM through 2.2.6 lets an authenticated user who can create or edit leads embed Vue.js double-brace template expressions in a lead's title field, which then execute as arbitrary JavaScript in the browser of any user who opens the poisoned record. The injected expressions reach the client-side Vue template compiler, allowing prototype-chain traversal to the Function constructor and code execution in the application origin, so every subsequent viewer of the lead is affected until the record is sanitized. Exploitation requires valid CRM credentials with lead privileges plus a victim who views the record, and is confined to the browser origin with no server-side code execution; no public exploit code and no confirmed active exploitation (not listed in CISA KEV) were identified at time of analysis, and this is assessed as a moderate, not top-tier, priority.
Reflected cross-site scripting in Paessler PRTG Network Monitor before version 26.2.120.1449 allows an unauthenticated remote attacker to craft a URL targeting a non-existent .htm resource, causing the 403 'Forbidden Path' error page to echo the requested path without encoding and execute arbitrary JavaScript in the context of a victim's active PRTG web session. Exploitation requires user interaction - the victim must click the crafted link while already authenticated to PRTG - and because the PRTG session cookie is not protected with the HttpOnly attribute, a successful attack can read and exfiltrate that cookie to enable session hijacking. No public exploit code or confirmed active exploitation has been identified at time of analysis, and Paessler has released version 26.2.120.1449 to address the flaw.
Trusted-HTML injection in Discourse chat excerpt renderers lets an authenticated user who can both upload files and post chat content embed HTML markup in a filename, which is then rendered unescaped in chat channel lists, chat summary emails, pinned message bars, reply previews, thread previews and other excerpt surfaces. The CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:C/C:N/I:L/A:N, score 5.0) confirms the attacker must be authenticated and that only content integrity is affected, with scope change reflecting that the injected markup renders in contexts beyond the uploader's own message; no public exploit code has been identified at time of analysis and JavaScript execution was not demonstrated under the default Content Security Policy. The practical risk is low and is driven almost entirely by sites that have disabled or relaxed the default CSP, which the advisory notes have increased exposure to script-level attacks. Fixed versions are 2026.1.8, 2026.6.3, 2026.7.2 and 2026.8.0.
An authenticated administrator can read files outside the configured log directory in IBM ContextForge MCP Gateway 1.0.0 through 1.0.8 by abusing a prefix-only path check on the Admin API log-download endpoint GET /v1/admin/logs/file. Because the confinement test uses str.startswith(), a crafted filename that resolves into a sibling directory sharing the LOG_FOLDER string prefix is accepted, exposing .log, .jsonl and .json files - but not arbitrary filesystem content, since only those extensions and prefix-adjacent paths are reachable. Exploitation requires a high-privileged (PR:H) admin account, and no public exploit identified at time of analysis; the C:H confidentiality rating overstates practical risk given the admin prerequisite and the narrow set of reachable files.
Authentication bypass in the Qt VNC Server module allows a remote, unauthenticated attacker to join a shared application session without knowing the configured VNC password by connecting with a modified client that violates the RFB protocol during the handshake, exposing the confidentiality and integrity of the shared session (no availability impact, VA:N). Exposure is conditional: a reachable Qt VNC Server instance must be actively sharing an application session and using the module's built-in password authentication, which is why the independent assessment rates this a genuine but conditional MEDIUM risk (assessed CVSS 3.1 AV:N/AC:H/PR:N/UI:R, C:H/I:H/A:N) rather than an urgent one. No public exploit code was identified at time of analysis (CVSS 4.0 E:U, no CISA KEV listing), and a vendor fix is available.
Local root access inside a guest logical partition (LPAR) on IBM PowerVM systems running affected partition firmware allows an attacker to send a crafted request to the partition firmware runtime and crash it, with possible memory corruption. Rated CVSS 5.1 (CVSS:3.1/AV:L/AC:L/PR:H/UI:N/S:U/C:N/I:L/A:H), the flaw affects firmware lines FW1120.00 through FW1120.01, FW1110.00 through FW1110.31, FW1060.00 through FW1060.81 and FW950.00 through FW950.H3; no public exploit has been identified at time of analysis, and the realistic outcome is denial of service to the partition firmware runtime rather than confirmed cross-partition code execution. The guest-root prerequisite makes this a bounded, low-priority risk relevant mainly to multi-tenant or compromised LPARs on shared Power hardware.
Local privilege escalation in Dell Rugged Control Center (RCC) versions prior to 5.2.206 allows a low-privileged attacker with local access to bypass authorization and gain elevated privileges. Exploitation requires an existing low-privileged local account on the host; no public exploit code has been identified, and real-world risk is modest as the flaw cannot be reached remotely and primarily affects multi-user or shared rugged devices.
Label-filter isolation in Dozzle 10.6.5 and earlier is bypassed through the SSE events endpoint (GET /api/events/stream), so any authenticated user - including one explicitly constrained by a users.yml 'filter:' label to a single container scope - receives live CPU/memory/network/disk telemetry and lifecycle metadata for every container on every monitored host. The flaw requires authorization to be enabled (e.g. DOZZLE_AUTH_PROVIDER=simple) and at least one account using per-user label filters, which is the standard multi-tenant or environment-segregated configuration; the attack needs only valid credentials (PR:L) and no special shell/actions/download role or victim interaction. Impact is bounded to metadata and telemetry (container name, image, full label map, and resource counters) rather than log contents, environment variables, or exec/attach capability, which is consistent with the assessed CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N rating of 4.3 (CWE-200). A lab-only proof-of-concept script is published in the vendor advisory demonstrating the leak; there is no evidence of active exploitation (not in CISA KEV).
Authenticated Termix users can abuse the POST /credentials/:id/deploy-to-host endpoint to reach other users' credential and host records, because the server resolves attacker-supplied credentialId and targetHostId integers without verifying ownership. On instances running 1.7.0 through 2.5.0, this yields limited information disclosure (whether a record exists and its authType) and lets an attacker cause a victim host that authenticates by key to make an outbound SSH connection attempt using the attacker's public key; encrypted passwords and private keys are never exposed, keeping impact consistent with the assessed CVSS 3.1 base score of 4.3 (AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N). No public exploit identified at time of analysis, and the vendor fixed the authorization check in version 2.5.1.
An authorization bypass through a user-controlled key (CWE-639) exists in Delta Electronics DIAEnergie versions before 1.11.00.022, allowing a low-privilege authenticated user to reference objects or records belonging to other users and read data they are not authorized to access. The flaw is network-reachable (AV:N) and requires no victim interaction (UI:N), but exploitation is gated behind a valid low-privilege session (PR:L), so realistic abuse is limited to malicious insiders, legitimate users exceeding their privileges, or attackers who have already obtained credentials; impact is bounded to confidentiality (C:L) with no integrity or availability effect, giving a CVSS 3.1 base score of 4.3. No public exploit code has been identified at time of analysis and the issue is not confirmed actively exploited (CISA KEV), making this a genuine but low-priority weakness rather than a high-severity emergency.
Insufficient input sanitization in HCL Sametime allows remote, unauthenticated attackers to expose limited information to a victim who interacts with attacker-controlled content processed by the product's web interface. According to the assessed vector (CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:L/I:N/A:N, score 4.3), exploitation cannot occur autonomously - the attacker must socially engineer a target into opening or interacting with crafted input - and the resulting impact is read-only information disclosure with no integrity or availability effect. No public exploit code or confirmed active exploitation has been identified at time of analysis, and HCL's advisory (KB0132652) does not specify which feature, input field, or configuration is affected.
Scoped OAuth MCP tokens issued by TREK before 3.3.0 can pull complete trip summaries - trip metadata, itinerary days, accommodations, and the email addresses of trip members - through the get_trip_summary tool even when the token was never granted the trips:read scope. The issue is an authenticated (CVSS PR:L), low-severity least-privilege bypass: an MCP client holding only an unrelated capability such as weather:read receives data it was never consented to read, defeating the OAuth scope boundary for the token owner's own trips and those of their collaborators. Cross-user trip authorization still holds, so the flaw is not a path to arbitrary users' trips, and no public exploit code was identified at time of analysis; vendor fix is TREK 3.3.0.
Authenticated workspace members in Docmost prior to 0.90.1 can read confidential sync-block content from private spaces via the transclusion/sync-block lookup API, which fails to enforce private space membership before resolving the source page. This authorization bypass through user-controlled key (CWE-639) is rated CVSS 4.3 (AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N), reflecting low confidentiality impact. No public exploit code has been identified at time of analysis.
Limited information disclosure affects IBM PowerVM Hypervisor firmware on specific service-pack levels (FW1120.00-FW1120.01, FW1110.00-FW1110.31, FW1060.00-FW1060.81, and FW950.00-FW950.H3), where an out-of-bounds read can expose a small amount of adjacent memory content to an attacker. Rated CVSS 4.3 (CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N), the flaw requires no authentication or user interaction but is not reachable across the internet: the attacker must already sit on the same adjacent network segment and be able to reach a hypervisor firmware interface that processes malformed input. Impact is confined to partial confidentiality loss with no integrity or availability effect, and no public exploit identified at time of analysis; IBM has published a vendor fix advisory.
Rendering of attacker-controlled iframes in Discourse is possible on sites that configure wildcard patterns in the allowed_iframes setting, allowing an authenticated user with posting privileges to embed an iframe whose origin is not actually allowlisted. The flaw is a bypass of the wildcard origin check: a crafted URL places the allowlisted suffix after a URL authority separator, so Discourse's text-based match succeeds while the browser resolves a different, attacker-controlled origin, and the iframe is then rendered through a post or a Onebox oEmbed response. Rated CVSS 4.3 (CWE-346, origin validation error) with only a low integrity impact and no confidentiality or availability impact; exploitation requires both a wildcard iframe configuration and posting privileges (PR:L), and no public exploit was identified at time of analysis.
Impact, prerequisites, and exposure for CVE-2026-91769 cannot be determined from the available intelligence; the only confirmed datum is that Ubuntu's security team is tracking this identifier against a package in the Ubuntu archive (reported by vendor:ubuntu). No description, CVSS score or vector, CWE classification, CPE string, affected version range, or configuration prerequisite was supplied, so it is not possible to state whether the flaw is remotely or locally reachable, what privilege an attacker needs (authentication requirements not confirmed from available data), or how severe the consequence is. Exploitation status is unconfirmed: there is no CISA KEV listing and no public exploit identified at time of analysis, but this reflects absence of reporting rather than evidence of safety. Defenders should treat CVE-2026-91769 as an unassessed Ubuntu package issue until the Ubuntu Security Notice and NVD record are published.
Login CSRF in the LoginRadius backend of Python Social Auth (social-core) before 5.0.0 lets an attacker complete a victim's browser session authentication against an attacker-controlled LoginRadius identity, because the OAuth state parameter is never validated during the callback. Only applications that specifically enable the LoginRadius backend are affected; per the assessed vector (AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:L/A:N, CVSS 4.3) the attacker must possess a valid LoginRadius token and induce user interaction, and the victim's own account is not compromised - they are simply logged in as the attacker, with no confidentiality impact. No public exploit code was identified at time of analysis and there is no CISA KEV listing; the vendor-released fix is social-core 5.0.0, which enforces callback state validation for LoginRadius.
Docmost 0.70.0 through 0.80.0 allows an authenticated low-privileged user with edit rights on an exportable page to disclose attachments belonging to restricted pages in the same space by embedding a forged attachmentId and exporting that page with the includeAttachments=true option. The export flow resolves the attacker-supplied attachment identifier against storage without re-validating ownership, so the restricted file is bundled into the returned ZIP even though a direct download would be denied (CWE-639 IDOR). This is a genuine but low-severity read-only authorization flaw (CVSS 3.1 base 4.3), exploitation is limited to attachments within a space the attacker can already access, no public exploit code has been identified at time of analysis, and the issue is fixed in version 0.80.1.
An out-of-bounds heap read in Espressif ESP-IDF's BlueDroid A2DP sink path (btc_a2dp_sink_handle_inc_media()) affects versions 5.5.5, 6.0.1, and 6.1 on builds where Classic Bluetooth (BR/EDR) and A2DP sink support are explicitly enabled. A paired BR/EDR audio source within Bluetooth radio range can send a malformed A2DP media packet whose layout omits the timestamp field, causing the sink to read a UINT32 past the valid buffer and leak limited adjacent heap contents (CVSS 4.3, AV:A/AC:L/PR:N/UI:N/C:L/I:N/A:N). There is no CISA KEV listing and no public exploit identified at time of analysis, and the reporter explicitly states that arbitrary memory disclosure and code execution are not established, despite third-party tags implying RCE.
Authenticated WordPress users holding Contributor-level access or higher can retrieve administrator-configured Instagram Graph API access tokens from any site running the Spectra Legacy - Gutenberg Blocks plugin at version 2.20.0 or earlier, because the plugin's editor_assets function injects the uag_insta_linked_accounts option into the uagb_blocks_info JavaScript object without performing a capability check. The exposure is gated on a specific deployment state: the separate Spectra Pro plugin must be active and an administrator must already have linked an Instagram account, so sites without that integration have no token to leak and are not meaningfully at risk. Rated CVSS 4.3 (AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N, CWE-200), this is a real but low-impact, configuration-dependent information disclosure; no public exploit code identified at time of analysis and no confirmed active exploitation (CISA KEV).
Missing capability checks in the MailerLite - Signup forms (official) WordPress plugin, in all versions up to and including 1.7.21, allow authenticated attackers holding Contributor-level access or higher to create or delete arbitrary signup forms on the affected site. The flaw sits in the forms() method of the AdminController class, which is reachable without verifying that the caller has the capability normally required to manage plugin forms, so any Contributor account is sufficient and no further user interaction is needed. Impact is limited to data integrity - attackers can add or remove form definitions but cannot read data, escalate privileges, or cause site-wide denial of service - and this is a low-severity, routine-patching issue rather than an emergency; no public exploit identified at time of analysis and no CISA KEV listing.
Information disclosure in the Keycloak Admin REST API lets an authenticated delegated administrator holding only basic group/role search privileges enumerate every group assigned to a given role, including groups they are supposed to be unable to see. The flaw (CWE-862, missing authorization) affects the Admin REST API endpoints that return role-to-group mappings in Keycloak, Red Hat Build of Keycloak, and Red Hat Single Sign-On 7, and is rated CVSS 3.1 4.3 (AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N) - a read-only confidentiality impact with no integrity or availability effect. It requires valid, privileged-but-scoped credentials and is only meaningful where an organization actually partitions group visibility across delegated admins; there is no public exploit identified at time of analysis and the issue is not confirmed actively exploited (no CISA KEV listing).
Keycloak's Level of Authentication (LoA) step-up enforcement can fail open: when a relying client requires a higher authentication level for a user who already holds an active session at a lower level, a logic error in session re-evaluation causes Keycloak to issue a token carrying the lower acr/LoA claim instead of forcing the required step-up authentication. The impact is a bounded assurance downgrade rather than unauthenticated access - the attacker must already be authenticated (CVSS PR:L) - but it may grant access to sensitive resources whose authorization decisions trust those claims, with low confidentiality and integrity impact and no availability impact (CVSS:3.1 4.2, AV:N/AC:H/UI:N/S:U/C:L/I:L/A:N). Exploitation depends on a deployment that actually configures per-client step-up/LoA and on a specific session-state and re-evaluation timing; no public exploit code has been identified at time of analysis and this CVE is not listed in CISA KEV.
Discourse's moderator review queue can expose the contents of private messages to moderators who are not participants in those conversations, because reviewables generated by the optional Discourse AI plugin were not filtered to the audience permitted to access the underlying private-message topic. Affected deployments are those running releases prior to 2026.1.6, 2026.5.2, 2026.6.1, and 2026.7.0 with the Discourse AI feature producing reviewables; exploitation requires an authenticated moderator account, the AI plugin, and a pre-existing AI reviewable tied to a PM the moderator is not a participant in, so an attacker cannot reliably force the required state. Confidentiality impact is high (and a moderator may additionally close the PM topic or delete a post if the relevant reviewable actions are available), but real-world risk is low-to-moderate given the high privilege required, the dependence on optional functionality, and the opportunistic precondition; no public exploit identified at time of analysis.