Authenticated horizontal privilege escalation in Vvveb CMS before 1.0.8.7 allows any administrator - including one deliberately denied the 'view_other_admin' and 'edit_other_admin' capabilities - to read and overwrite another administrator's REST API tokens by changing the admin_id GET parameter on admin/auth-token. The stolen token can then be replayed against the REST API with a Bearer header to act as the victim administrator, so the impact escalates to full account takeover of a higher-privileged admin. Exploitation requires a valid authenticated administrator session (CVSS PR:H, no user interaction), the vendor states the issue is fixed in 1.0.8.7, and a public PoC reference with reproduction steps exists in the GitHub security advisory; it is not verified as working exploit code, and there is no KEV listing indicating active exploitation.
Account impersonation in pgAdmin 4 server mode before 9.19 lets a user who successfully authenticates against a configured LDAP directory or Kerberos realm be logged into a different pgAdmin account that happens to share their username - including the built-in internal administrator created at first setup - inheriting that account's role, registered servers and saved database passwords. A second variant occurs when internal authentication is listed before LDAP or Kerberos in AUTHENTICATION_SOURCES: Flask-Security's internal form validation resolves the typed email to an account and retains it even though the password did not match, so a Kerberos or LDAP user can type another directory user's principal or email and be logged in as that user. Exploitation requires valid credentials for the directory or realm pgAdmin is configured to use (authenticated, PR:L) plus a username collision across sources; desktop mode and versions 9.19 and later are not affected, and no public exploit code was identified at time of analysis.
Stored cross-site scripting in Chamilo LMS 2.0.0 through releases before 2.0.1 lets an account holding at least ROLE_STUDENT plant HTML in six user-editable profile fields (firstname, lastname, competences, openarea, teach, diplomas), which are rendered unescaped on the anonymously reachable GET /user/{username} page, so the payload executes in the browser of any visitor, including administrators, and can lead to account takeover via session, JWT or localStorage exfiltration. Exploitation requires a student-level account (self-registration, where enabled, lets an otherwise anonymous attacker obtain that role) plus victim interaction - a visitor must follow the public profile link - and no public exploit identified at time of analysis.
Pickle-analysis bypass in Trail of Bits fickling through 0.1.12 and vendor master commit d011709 allows arbitrary code execution when a victim uses fickling.load() or check_safety() as a safety gate on a crafted pickle, with CVSS 8.8 (AV:N/AC:L/PR:N/UI:R). The crafted pickle uses a dotted __qualname__ such as optparse.Values.read_file, which fickling records and renders as a plain import and never analyzes as a call, so OvertlyBadEvals does not flag it and check_safety() returns LIKELY_SAFE with zero findings while pickle resolution reaches a code-executing class method. No public exploit identified at time of analysis, and exploitation requires the victim to actively run fickling against the malicious pickle; the advisory states the bypass remained present on vendor master at commit d011709 as of 2026-10-30 and the vuln.today priority is this_week.
fickling through 0.1.12 misclassifies malicious pickle payloads as safe: its OvertlyBadEvals analysis pass skips any call whose name was added to the likely_safe_imports set purely because the source module appears in sys.stdlib_module_names, so check_safety() can return LIKELY_SAFE with zero findings for payloads that reach code execution or arbitrary file write. Because fickling.load() chains check_safety() into pickle.loads() as a security gate, a victim who runs unauthenticated, attacker-supplied pickle or model content through the tool (the required user interaction, UI:R) can end up executing that content with the privileges of the analysis process. The advisory demonstrates three concrete primitives - dataclasses._create_fn (Python 3.7-3.12), mailcap.findmatch (through Python 3.12) and readline.add_history()/write_history_file() on POSIX - while noting the underlying module-origin trust root cause is broader and remains unfixed; no public exploit code is referenced in the available intelligence and there is no KEV entry, but the tool's whole purpose is vetting untrusted pickles, so the severity is real for anyone relying on it as a gate.
Arbitrary plugin installation in Formidable Forms (free edition) through 6.35 lets authenticated low-privilege users install attacker-supplied plugins on the WordPress server, yielding code execution under the web server account. Exploitation requires an account with subscriber-level access or above that holds the delegable frm_view_forms capability, and it is only possible on free-only sites where Formidable Pro is not installed; an attacker combines a permission callback that trusts only a shared token (without checking install_plugins or activate_plugins) with weak URL prefix validation in validate_url_is_in_s3_bucket() reached through the 'file_url' parameter. No public exploit or CISA KEV entry is identified at time of analysis, so the risk is concentrated on free-only sites that permit self-registration or otherwise hand low-trust users subscriber access.
Authenticated users of TAKO AI Agent for Okta can execute arbitrary Python and OS commands on the server because the application's script validator can be bypassed with dynamic getattr call chains; all versions prior to 3.0.1-beta are affected, and the default Docker deployment runs the process as root. The public advisory includes a working proof-of-concept that returns uid=0(root) output in that default container, enabling disclosure of Okta API tokens and the JWT key, SQLite database access, and backend pivoting. No CISA KEV entry is recorded at time of analysis; the vendor-released fix is 3.0.1-beta.
Privilege escalation in the WebToffee 'Export and Import Users and Customers' plugin for WordPress (all versions up to and including 2.7.5) lets any authenticated user with Subscriber-level access or above plant a crafted value in the Biographical Info field that is later reinterpreted as a second CSV row during import, creating a brand-new account whose login, password, e-mail and role are fully attacker-chosen - including Administrator. Exploitation is conditional rather than automatic: the payload only detonates when a privileged operator runs the plugin's own export and then re-imports the generated file, so sites that never round-trip exported user data back through the importer are not exposed. No public exploit code or confirmed in-the-wild activity was identified at the time of analysis, but the 8.8 severity and the low Subscriber-level prerequisite make this a this-week remediation priority for any site that uses the plugin's export/import cycle.
Time-based blind SQL injection in Apache Fineract through 1.15.0 lets an authenticated user who already holds read access to centers or to a specific datatable inject arbitrary SQL through the orderBy, sortOrder and order query parameters of the center read endpoint (GET /api/v1/centers) and the datatable read endpoints (GET /api/v1/datatables/{datatable}/{apptableId} and /api/v1/datatables/{datatable}/{apptableId}/{datatableId}). The bug is CWE-89: caller-supplied ordering values are concatenated straight into the SQL ORDER BY clause and the pre-existing regex validation (ColumnValidator) can be bypassed; the project's own fix replaces that regex with an allowlist check against the datatable's real column headers. Exploitation is not anonymous - it requires a valid account with the relevant read permission - and because the injection is blind and time-based, bulk data extraction is slow; EPSS is 0.28% (20th percentile), there is no CISA KEV listing and no public exploit code identified at time of analysis, and Apache recommends upgrading to 1.16.1.
Privilege escalation in IXP Manager before 7.3.0 lets an authenticated, non-administrative account raise its own privileges to full administrator status over the network without user interaction. The flaw scores CVSS 3.1 base 8.8 (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H), meaning the attacker must already hold a low-privilege login but then gains high confidentiality, integrity and availability impact across the exchange's management plane. Exposure scales with the number of low-trust accounts an installation issues - deployments granting member or customer self-service logins carry the most risk, while single-operator instances with no untrusted accounts are largely insulated. No public exploit code has been identified at time of analysis and the issue is not in CISA KEV; the vendor fixed it in v7.3.0 and states that restricting administrative functions reduces but does not eliminate exposure.
Broken access control in the Cloud Foundry Cloud Controller API leaves the /internal/v3/staging/:guid/droplet_completed and /internal/v3/staging/:guid/build_completed callback endpoints reachable on the public API listener, because the shipped Nginx routing configuration carries deny rules for /internal/v4/ and /internal/v5/ but omits the equivalent rule for /internal/v3/. An attacker who holds a valid build or droplet GUID - and a SpaceAuditor can enumerate those GUIDs without any write permission - can post a forged staging callback, transition a droplet to STAGED, and persist arbitrary process and sidecar metadata with no associated package or actor, escalating past the intended role boundary. No public exploit code was identified at time of analysis, and the vendor has published fixed versions; triage guidance places remediation in this week's work rather than an emergency response.
Heap use-after-free in Ladybird Browser's LibJS FinalizationRegistry allows a malicious webpage to crash the WebContent process and, in production builds, potentially corrupt memory. The bug is reachable by unauthenticated remote attackers per the assessed CVSS vector (AV:N/PR:N) but requires the victim to load attacker-controlled JavaScript (UI:R); the advisory includes a public reproduction PoC that triggers the UAF on ASAN builds, while CISA KEV does not list active exploitation. Full type-confusion/RCE exploitation is not demonstrated in the provided data because it additionally depends on a same-size heap spray and an ASLR-defeating information leak.
Heap use-after-free in Google Chrome's Input component allows a remote attacker to execute arbitrary code inside the sandbox if a victim opens a crafted HTML page in any Chrome build prior to 156.0.8078.79. User interaction is mandatory (CVSS UI:R) and no authentication or non-default configuration is needed, so the realistic delivery path is a malicious or compromised site, injected page content, or malvertising; because the flaw yields renderer-level execution described as 'inside the sandbox', host impact is bounded absent a separate sandbox escape. Exploitation is not confirmed in the wild (SSVC exploitation: none, EPSS 0.24%, 15th percentile), but the vendor-rated Chromium severity is High with a total technical impact score, and a fixed build is already published.
Use-after-free in Google Chrome releases prior to 156.0.8078.79 can let an unauthenticated remote attacker execute arbitrary code inside the renderer sandbox if a victim opens a crafted HTML page; because the CVSS scope is unchanged, host compromise would require chaining a separate sandbox escape. No public exploit has been identified at time of analysis, EPSS is low at 0.24% (15th percentile), and CISA SSVC records no known exploitation and no automatable delivery. A vendor patch is available in Chrome 156.0.8078.79, and the issue is assessed as serious but not an emergency under the this_week priority.
Arbitrary code execution inside the renderer sandbox is possible in Google Chrome builds prior to 156.0.8078.79, where improper input validation in the V8 engine lets a crafted HTML page drive the engine into executing attacker-influenced code. The attack is network-delivered and needs no credentials, but requires the victim to open attacker-controlled web content (CVSS 8.8, AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H), and scope is unchanged - the payload runs inside the sandbox rather than on the underlying host. No exploitation has been observed per SSVC (Exploitation: none), no public exploit was identified at time of analysis, EPSS is 0.24% (15th percentile), and the vendor has shipped a fixed build (156.0.8078.79).
Improper input validation in Chrome's CacheStorage component allows a remote, unauthenticated attacker to execute arbitrary code inside the browser's renderer sandbox by convincing a user to open a crafted HTML page in a Chrome build older than 156.0.8078.79 (CVSS 3.1: 8.8, AV:N/AC:L/PR:N/UI:R, High Chromium severity). The user-interaction requirement (UI:R) is the main practical barrier, and any code gained this way is confined to the sandboxed renderer, so reaching the host would need a second, separate sandbox-escape flaw. No public exploit code was identified at time of analysis, there is no CISA KEV listing, EPSS is low at 0.21% (12th percentile), and CISA SSVC rates exploitation as none and automatable as no, which supports treating this as a real but moderate-urgency patch-this-week item rather than an emergency.
Use-after-free in Skia, Chrome's graphics rendering library, lets a remote, unauthenticated attacker execute arbitrary code inside the renderer sandbox when a victim loads a crafted HTML page in Google Chrome builds prior to 156.0.8078.79. The attack is a drive-by or malicious-link scenario requiring only that the target browses to attacker-controlled content (UI:R), with no credentials or non-default configuration needed; code execution remains confined to the renderer sandbox, so host compromise would require chaining a separate sandbox escape. No public exploit has been identified at time of analysis, CISA KEV does not list the issue, and EPSS is low at 0.20% (10th percentile), though the vendor rates the flaw High severity and SSVC assigns total technical impact, yielding an overall priority of remediating this week rather than an emergency response.
A use-after-free flaw in Chrome's Blink rendering engine affects versions prior to 156.0.8078.79 and lets a remote attacker run arbitrary code inside the browser sandbox when a victim opens a crafted HTML page. Exploitation requires the user to visit or be navigated to attacker-controlled content (UI:R) but no credentials or privileges (PR:N), and the resulting code execution is confined to the sandbox, so reaching the host operating system would require chaining a separate sandbox-escape bug. No public exploit has been identified at time of analysis, EPSS is low at 0.20% (10th percentile), and the issue is not in CISA KEV; the vendor-rated Chromium severity is High and the priority is to patch this week.
Integer overflow in Google Chrome's Editing component (CWE-190) lets a remote, unauthenticated attacker who convinces a victim to load a crafted HTML page potentially execute arbitrary code, though the vendor states the resulting execution is confined to the Chromium sandbox. All Chrome builds prior to 156.0.8078.79 are affected; the flaw carries a CVSS score of 8.8 (AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H) and a Chromium security severity of High. No public exploit code has been identified at time of analysis and EPSS is low at 0.20% (10th percentile), while SSVC rates technical impact as total but exploitation as none and automatable as no, placing it in a 'this_week' remediation queue rather than an emergency.
Remote code execution inside the Chrome renderer sandbox affects Google Chrome builds prior to 156.0.8078.79, where memory corruption in the DOM lets a crafted HTML page execute arbitrary code. Attackers must convince a victim to load attacker-controlled HTML (UI:R) and need no credentials, but the impact is confined to the sandboxed renderer unless a separate sandbox-escape vulnerability is chained. No public exploit identified at time of analysis; the issue is not in CISA KEV, and an EPSS score of 0.20% (10th percentile) indicates low predicted exploitation.
Arbitrary code execution confined to the Chrome renderer sandbox is possible in Google Chrome builds prior to 156.0.8078.79, where a use-after-free in the Quota component is reachable when a victim renders a crafted HTML page. Exploitation is unauthenticated (PR:N) but requires user interaction (UI:R) - the target must visit or load the attacker-controlled page - and the description scopes the resulting code execution to 'inside the sandbox', so impact on the underlying operating system would require chaining a separate sandbox-escape flaw not covered by this CVE. No public exploit or PoC is identified in the available sources, CISA SSVC records Exploitation: none and Automatable: no, EPSS is low at 0.20% (10th percentile), and a vendor patch exists, which places this at a this_week remediation cadence rather than an emergency response.
Memory corruption in the Quota component of Google Chrome allows a remote attacker to potentially execute arbitrary code inside the browser sandbox by convincing a user to load a crafted HTML page; all builds prior to 156.0.8078.79 are affected. Exploitation needs the victim to visit or render attacker-controlled content (UI:R) but no authentication or privileges (PR:N), and the resulting code execution is confined to the renderer sandbox, so host compromise would require a separate sandbox escape. No public exploit has been identified at time of analysis, SSVC records Exploitation: none and Automatable: no, and EPSS is low at 0.20% (10th percentile), which supports a this_week rather than emergency response cadence; the vendor has released a fixed build.
Memory corruption in the Dawn (WebGPU) component of Google Chrome affects builds prior to 156.0.8078.79 and lets a remote, unauthenticated attacker potentially execute arbitrary code inside the browser sandbox when a victim loads a crafted HTML page (CVSS 3.1: 8.8, AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H). The user-interaction requirement and the sandbox scope limit real-world impact: exploitation grants code execution within the renderer sandbox, not host compromise, and would require a separate sandbox-escape bug to break out. No public exploit has been identified at time of analysis, EPSS is low at 0.20% (10th percentile), and SSVC records exploitation as none; a vendor patch is available, and the issue is prioritized for remediation this week.
Use-after-free in Google Chrome's Frames component before 156.0.8078.79 allows a remote, unauthenticated attacker to execute arbitrary code inside the renderer sandbox when a victim loads a crafted HTML page. Exploitation requires user interaction, and the resulting code execution is confined to the sandbox; full host compromise would require a separate sandbox escape that this CVE does not describe. The vendor has released a fixed build, while the supplied data shows no public exploit identified at time of analysis and an EPSS score of 0.20% (10th percentile).
Memory corruption in the FileSystem component of Google Chrome prior to 156.0.8078.79 allows remote attackers to potentially execute arbitrary code inside the browser's renderer sandbox when a victim loads a crafted HTML page. Affected users are those running any Chrome build below 156.0.8078.79; no authentication is required but successful exploitation depends on user interaction (visiting an attacker-controlled or compromised page). No public exploit code has been identified at time of analysis, the issue is not listed in CISA KEV, and EPSS is low (0.20%, 10th percentile), although the vendor rates the Chromium security severity as High and SSVC assigns total technical impact.
Memory corruption in Google Chrome's network component affects versions prior to 156.0.8078.79 and lets a remote attacker who convinces a user to load a crafted HTML page potentially execute arbitrary code inside the browser sandbox. The CVSS vector (AV:N/AC:L/PR:N/UI:R/S:U) confirms no privileges are required but user interaction is mandatory, and scope is unchanged, so meaningful host compromise would require chaining a separate sandbox escape. No public exploit has been identified at time of analysis: SSVC records Exploitation as none and Automatable as no, EPSS is low at 0.20% (10th percentile), and the issue is not in CISA KEV despite a total technical impact rating.
Use-after-free in Google Chrome's FileSystem component allows remote attackers to execute arbitrary code inside the renderer sandbox when a user loads a crafted HTML page, affecting all Chrome builds prior to 156.0.8078.79. Exploitation requires user interaction (UI:R) and no authentication (PR:N), and code execution is confined to the sandbox, so host compromise would need a separate sandbox escape. No public exploit code has been identified at time of analysis and Chrome is not in CISA KEV; EPSS is 0.20% (10th percentile), so risk is serious but not emergency-grade, with patching targeted for this week.
Memory corruption in the Autofill component of Google Chrome builds prior to 156.0.8078.79 lets a remote attacker execute arbitrary code inside the renderer sandbox when a victim renders a crafted HTML page. The flaw is unauthenticated per the CVSS vector (PR:N) but requires user interaction (UI:R), and because scope is unchanged (S:U), code execution is confined to the sandbox - escaping to the host requires a separate sandbox-escape vulnerability. No public exploit code and no active exploitation are confirmed (CISA KEV not listed, SSVC exploitation: none), EPSS is low at 0.20% (10th percentile), and a vendor patch is available.
Stored cross-site scripting in the WP Visitor Statistics (Real Time Traffic) WordPress plugin affects every release before 8.8 and allows an unauthenticated visitor to plant script payloads through the plugin's normal tracking path, which are then rendered unescaped on the administrative statistics dashboard. Because the payload fires only when a logged-in administrator opens the statistics pages, successful compromise depends on admin interaction and is delayed rather than immediate; the assessed impact is limited-scope script execution in the admin's browser (assessed CVSS 6.1-equivalent: AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:N) rather than the higher 8.8 severity some feeds report. EPSS is low at 0.17% (7th percentile), CISA SSVC records no known exploitation and non-automatable attack logic, and no public exploit was identified at time of analysis, though a vendor fix is available and the vuln.today priority places remediation in this week's patch window.
Proof verification bypass in OpenVM's generated Solidity OpenVmHalo2Verifier allows remote, unauthenticated attackers to get malformed proof data treated as valid, so integrating applications can accept arbitrary app commitments and public values without the Halo2 proof suffix, accumulator, app commitments, or public values being checked. The issue affects OpenVM SDK users who deployed the generated wrapper from 1.x before 1.7.0 and from affected 2.0/2.1 prereleases, and is fixed in 1.7.0, 2.0.0-rc.3 and 2.1.0-preview. No public exploit is identified at time of analysis and no CISA KEV entry is present in the supplied data; the independent assessment rates integrity impact High with no confidentiality or availability impact.
PHP object injection in MasterStudy LMS through version 3.7.55 allows an authenticated attacker with Subscriber-level or higher access to reach an unsafe deserialization path, carrying a CVSS 8.8 rating for high confidentiality, integrity, and availability impact if a usable PHP gadget chain is present. The main prerequisite is a valid low-privilege WordPress account-open self-registration on course sites is the primary enabling condition-and no public exploit was identified at time of analysis. The vuln.today decision label is this_week, matching the assessment that this is a credible but not emergency-grade priority, with the vulnerable endpoint and triggering parameter unstated in the available sources.
Subscriber-level users of the WP BASE Booking WordPress plugin through version 6.5.2 can escalate their privileges because the plugin performs an incorrect privilege assignment (CWE-266), with successful exploitation giving high confidentiality and integrity impact. The attack requires an authenticated subscriber account but no user interaction and is network-reachable; sites that allow open registration make that prerequisite much easier to satisfy. Patchstack reported the issue, the vuln.today priority is this_week, and no public exploit was identified at time of analysis; this is serious but conditionally gated rather than an emergency.
PHP object injection in the WordPress plugin GS Testimonial Slider (versions up to and including 3.3.10) allows an authenticated user with Contributor-level privileges to submit crafted serialized input that the plugin deserializes, potentially leading to high-impact confidentiality, integrity and availability effects (CVSS 8.8, AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H). Because the vector requires PR:L, anonymous exploitation is not possible; the risk is concentrated on sites that grant Contributor accounts freely or with loose vetting. Actual code execution depends on a usable POP gadget chain being present in the plugin, WordPress core, or other installed plugins/themes, and the exact injectable parameter or sink is not described in the available sources. No public exploit code has been identified at time of analysis; the issue is a prioritization candidate for this week rather than an emergency, and no fixed version is listed in the available data.
PHP object injection in the WordPress plugin Enable Media Replace through version 4.2.2 gives authenticated Subscriber-level users a path to high-impact compromise when a usable POP gadget chain is present. The CVSS 3.1 vector is AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, so exposure is limited to sites where an attacker can obtain a low-privilege account, such as sites with user self-registration enabled; the exact vulnerable feature and trigger are not described in the available sources. No public exploit identified at time of analysis and no CISA KEV status is supplied, and the this_week priority label makes this a genuine but conditional priority rather than an emergency.
Privilege escalation in the Ak Framework WordPress plugin, bundled with the Newsy - Viral News & Magazine theme, through version 2.9.8 allows attackers to reach the ajax_refresh_nonce and ak_admin_ajax handlers and pass attacker-controlled option names and values into update_option(), including users_can_register and default_role. Setting those options to enable open registration with the administrator role gives an attacker a path to create an administrator account. The advisory describes the handlers as reachable without authentication, while the supplied CVSS vector specifies PR:L (authenticated), so the authentication prerequisite should be confirmed against the vendor advisory before scoping exposure; exploitation is limited to sites running this theme/plugin, and no public exploit code or CISA KEV entry appears in the supplied intelligence.
Cross-site request forgery in Meow Apps AI Engine, a WordPress plugin, through version 3.8.8 allows unauthenticated remote attackers to induce a logged-in privileged user into executing a state-changing action by visiting attacker-controlled content (CVSS 8.8). The plugin does not verify request authenticity for that action (CWE-352), and successful exploitation requires the victim to be authenticated with sufficient privileges and to interact (UI:R); the exact endpoint and resulting state change are not described in the available sources. No public exploit was identified at time of analysis, and the available data does not list this CVE in CISA KEV.
OS command injection in Dell OpenManage Enterprise affects all versions prior to 4.8.0: a remote attacker who already holds valid low-privileged credentials on the management console can pass crafted input that reaches an operating system command without proper neutralization, resulting in arbitrary command execution with high impact to confidentiality, integrity and availability (CVSS 8.8). The authentication requirement is the main limiting factor - unauthenticated exploitation is not indicated - but no user interaction is needed and attack complexity is low, so any environment where low-privilege accounts exist or can be obtained carries realistic risk. No public exploit code or confirmed active exploitation is identified at time of analysis; Dell's DSA-2026-467 advisory provides the fix in 4.8.0 and this is assessed as a near-term 'this week' priority rather than an emergency.
OS command injection in Dell OpenManage Enterprise versions prior to 4.8.0 lets a remote, low-privileged authenticated attacker execute arbitrary operating-system commands, with high confidentiality, integrity, and availability impact. The CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H) confirms network access, low privileges, and no user interaction, with the affected appliance required to be network-reachable. A vendor patch to 4.8.0 is available, no public exploit was identified at time of analysis, and the this_week priority reflects a credible, actionable issue rather than an inflated high-CVSS artifact.
Unsafe Java deserialization in the OpenSearch SQL plugin allows a remote user holding only basic read/search permissions to run arbitrary code on the node by sending a crafted `cursor` parameter to the `plugins/sql` endpoint. The flaw affects the 2.x line from 2.8.0.0 up to (but not including) 2.19.6.0 and the 3.x line from 3.0.0 up to 3.7.0.0, and it is fixed in releases 2.19.6.0 and 3.7.0.0; the plugin ships enabled by default in OpenSearch distributions, so no unusual configuration is needed to reach the vulnerable path. No public exploit code was identified at time of analysis, but successful code execution additionally depends on a usable gadget chain being present on the server classpath, and the vendor priority places remediation in this week's workload.
MyTube deployments running in multi-user mode (loginEnabled=true) expose all locally stored and cloud-proxied media - videos, thumbnails, small thumbnails, avatars, subtitles and cloud signed URLs - to unauthenticated clients because the static and cloud route handlers are mounted before the authentication middleware. An attacker with network reach to the instance can fetch `/videos/*`, `/images/*`, `/subtitles/*`, `/avatars/*`, `/api/cloud/thumbnail-cache/*` and `/cloud/videos|images/:filename` without any session cookie, token or user interaction, and can combine this with the separate hidden-video path leak to download media the operator intended to hide. The default single-user configuration (loginEnabled=false) has no authentication boundary and is not affected; the flaw is fixed in version 1.10.3, a public PoC reference (curl-based request examples) is documented in the advisory, and there is no indication of confirmed active exploitation (no CISA KEV listing).
Command-line argument injection into the yt-dlp invocation used by the Dream Displays Minecraft mod/plugin (versions 1.5.0 through 1.8.1) lets player-supplied media/watch-party URLs carry option-like tokens that yt-dlp parses as its own command-line flags, altering the downloader's behaviour. The impact is scored as integrity-only (CVSS 4.0 8.7, and an assessed 3.1 vector of AV:N/AC:L/PR:N/UI:N with C:N/I:H/A:N), and exploitation is confined to servers actually running a vulnerable release with the display/watch-party feature in use; no public exploit code was identified at time of analysis and the issue does not appear in CISA KEV. The vendor fixed the flaw in version 1.8.2 by validating URLs against a media URL policy and inserting an end-of-options marker before passing them to yt-dlp.
Proof soundness in OpenVM before 1.7.0, and in affected 2.0 and 2.1 prereleases, can be undermined by a malicious prover who fabricates Merkle expansion rows at or below the leaf layer, allowing the committed memory root to be changed without a matching memory operation. The flaw affects proofs that use persistent memory across continuation segment boundaries; deployments not using persistent memory or running fixed releases 1.7.0, 2.0.0-rc.3, or 2.1.0-preview are not exploitable. Exploitation requires control of proof generation, no public exploit was identified at time of analysis, and the available priority signals point to a genuine but scoped issue rather than mass exploitation.
Authentication bypass in Delinea Secret Server before 12.3.1 lets a user skip the primary login step and authenticate by completing only the second-factor login step when two-factor authentication is configured. The vendor rates it 8.7 High; exploitation requires the attacker to be able to satisfy the second factor and is assessed as high complexity, so it is not a trivial unauthenticated bypass. No public exploit code or active-exploitation confirmation was identified in the available sources at time of analysis.
Unauthenticated GraphQL queries against Coop deployments earlier than 1.0.0 expose organization records and reviewer identity data, including org IDs, names, email addresses, website URLs, on-call alert addresses, and reviewers' IDs, names, roles, permissions, and creation dates. The flaw is missing authentication on the Query.allOrgs field reachable at /api/v1/graphql, rated CVSS 8.7 (CWE-306); any publicly accessible instance is affected, while authenticated private network placement prevents outside querying. The vendor advisory includes a public proof-of-concept GraphQL query, and the issue is fixed in Coop 1.0.0 via PR #699 / commit b29242f.
Delinea Secret Server versions earlier than 12.3.1 can return cloudadmin permissions to a user who submits an unsigned JWT under conditions not enumerated in the available sources, creating an authorization-bypass and privilege-escalation risk in a privileged access management product. The issue is network-reachable but rated high complexity and requires an existing privileged/authenticated context (CVSS PR:H), and it is remediated in 12.3.1. No public exploit code or CISA KEV active-exploitation confirmation was identified in the supplied data.
Heap memory corruption inside AWS Nitro Enclaves built from the kmstool-enclave sample application: an application running on the parent EC2 instance can send a crafted ordered sequence of SetClient and Decrypt control messages over vsock and trigger a use-after-free and double-free in the enclave's heap (CWE-416/CWE-415), potentially achieving arbitrary code execution within the enclave. Only enclave images that include kmstool-enclave or code derived from it are affected, in aws-nitro-enclaves-sdk-c versions up to and including v0.4.5; the SDK library itself is not affected and AWS states that no cross-instance, cross-account or cross-tenant access is possible. The issue is fixed in v0.4.6, no workarounds exist, and no public exploit code was identified at time of analysis.
OSAC operator's console-proxy can be induced to execute an attacker-supplied credential plugin inside its pod, yielding code execution as the osac-console-proxy service account in affected revisions such as 8220502. The issue is high severity (CVSS 8.6) and affects deployments in remote or auto VM cluster mode, with auto the default; deployments using local mode exclusively are unaffected because they do not read kubeconfig Secrets. Exploitation requires a principal able to create or modify a Secret labeled osac.openshift.io/remote-cluster-kubeconfig: "true" in a namespace where the proxy's RoleBinding grants Secret read access, so practical risk depends on who holds that Secret-write privilege. No KEV or EPSS data was supplied, and no working public exploit is confirmed in the provided sources; a unit-test fixture in the upstream patch demonstrates an exec stanza. The exec/auth-provider path is fixed upstream in commit 7e4222b (not yet released), while the separate clusters[].server SSRF path is not addressed by that patch.
Stored cross-site scripting in the Shopper headless e-commerce admin panel (2.0.0 up to but not including 2.10.1, plus 3.0.0-beta.1 and beta.2) lets authenticated back-office staff who hold only a content-upload permission such as edit_products or add_brands plant HTML or SVG media that executes in an administrator's session when that administrator opens the file. Because the media is served inline from the same origin at /storage/ with no Content-Disposition or Content-Security-Policy, and because the CreateTeamMember store() flow trusts a client-supplied role_id, the injected script can mint a persistent attacker-controlled administrator account with access to customer data, orders, settings and user/role management. Exploitation requires two parties (a content editor who can upload plus an administrator who opens the file), since direct account creation by the low-privileged uploader is blocked by the access_setting check; the flaw is fixed in 2.10.1 and 3.0.0-beta.3, and the upstream GHSA advisory documents a proof-of-concept chain, but no CISA KEV listing appears in the provided data.
Path traversal in LM Studio's model downloader before 0.4.22+1 allows a crafted model repository to write startup-loaded configuration files outside the models directory, resulting in execution of an attacker-supplied Python payload on every launch. Exploitation is not autonomous: the victim must be tricked into downloading the malicious repository, so the CVSS 4.0 score of 8.7 reflects high confidentiality, integrity, and availability impact gated by user interaction. Vendor-released patch: 0.4.22+1; no public exploit identified at time of analysis.
Unauthenticated denial of service in Axigen Mail Server lets a single HTTP request with an extremely deeply nested SOAP/XML or JSON body sent to a WebMail, WebAdmin, or WebMail Proxy listener exhaust the stack and crash SMTP, IMAP, POP3, and web services. The affected branches are releases before 10.5.69, 10.6.x before 10.6.40, and 10.7.x before 10.7.2; the CVSS 4.0 vector has PR:N and availability-only impact, and the provided intelligence reports no CISA KEV entry or public exploit code. Our assessment rates this this_week: a real prioritize-this-week problem for exposed Axigen deployments, but not a top-tier emergency, because exploitation depends on network reachability to the relevant web listeners.