CVE Feed

Heap buffer overflow in WebML in Google Chrome prior to 148.0.7778.168 allowed a remote attacker to execute arbitrary code inside a sandbox via a crafted HTML page. (Chromium security severity: Critical)

Fleet is open source device management software. Prior to version 4.80.1, a vulnerability in Fleet's IP extraction logic allows unauthenticated attackers to bypass API rate limiting by spoofing client IP headers. This may allow brute-force login attempts or other abuse against Fleet instances exposed to the public internet. Fleet extracted client IP addresses from request headers (`True-Client-IP`, `X-Real-IP`, `X-Forwarded-For`) without validating that those headers originate from a trusted proxy. The extracted IP is used as the key for rate limiting and IP ban decisions. As a result, an attacker could rotate the value of these headers on each request, causing Fleet to treat each attempt as coming from a different client. This effectively bypasses per-IP rate limits on sensitive endpoints such as the login API, enabling unrestricted brute-force or credential stuffing attacks. This issue primarily affects Fleet instances that are directly exposed to the internet without a reverse proxy that overwrites forwarded-IP headers. Instances behind a properly configured proxy or WAF are less affected. Version 4.80.1 contains a patch. If an immediate upgrade is not possible, administrators should ensure Fleet is deployed behind a reverse proxy (e.g., nginx, Cloudflare, AWS ALB) that overwrites `X-Forwarded-For` with the true client IP, and apply rate limiting at the proxy or WAF layer.

libsixel is a SIXEL encoder/decoder implementation derived from kmiya's sixel. From to 1.8.7-r1, a wrong NULL check after an allocation call in sixel_decode_raw and sixel_decode causes a NULL pointer dereference whenever the allocation fails. The check tests the address of the output parameter (always non-NULL) instead of the value the malloc returned. On allocation failure, the function continues and writes through a NULL pointer, crashing the process. This is a denial of service against any caller of these public APIs that hits a low-memory condition. This vulnerability is fixed in 1.8.7-r2.

libsixel is a SIXEL encoder/decoder implementation derived from kmiya's sixel. From to 1.8.7-r1, a signed integer overflow in the SIXEL parser's image-buffer doubling loop can lead to an out-of-bounds heap write in sixel_decode_raw_impl. context->pos_x grows by repeat_count on every sixel character with no upper bound check. Once pos_x approaches INT_MAX, the expression "pos_x + repeat_count" used to size the image buffer overflows signed int. Depending on how the overflow wraps, the resize check that should reject oversized buffers can be bypassed, after which a subsequent write computes a large attacker-influenced offset into image->data and writes past the allocation. Reachable from any caller that decodes attacker-supplied SIXEL data, including img2sixel. This vulnerability is fixed in 1.8.7-r2.

libsixel is a SIXEL encoder/decoder implementation derived from kmiya's sixel. From to 1.8.7-r1, signed integer overflow in sixel_encode_highcolor's allocation size calculation can lead to a heap buffer overflow. The public sixel_encode entry point validates only that width and height are greater than zero, with no upper bound. width and height are multiplied as plain int when computing the allocation size for paletted_pixels and normalized_pixels. Any caller that asks libsixel to encode a pixel buffer with width times height greater than INT_MAX (about 2.15 billion) will hit a wrapped allocation size; under the right wrap, the malloc succeeds with a buffer much smaller than the encoder expects, and the encoder writes past the end of the heap allocation. This vulnerability is fixed in 1.8.7-r2.

OpenImageIO is a toolset for reading, writing, and manipulating image files of any image file format relevant to VFX / animation. Prior to 3.0.18.0 and 3.1.13.0, the bounds check in TGAInput::decode_pixel computes k + palbytespp as unsigned 32-bit arithmetic. When k = 0xFFFFFFFC and palbytespp = 4, the addition wraps to 0, which compares less than palette_alloc_size and passes the check. The subsequent palette access uses the unwrapped k (0xFFFFFFFC) as the index, reading ~4 GB past the start of the palette buffer — SEGV. This vulnerability is fixed in 3.0.18.0 and 3.1.13.0.

OpenImageIO is a toolset for reading, writing, and manipulating image files of any image file format relevant to VFX / animation. Prior to 3.0.18.0 and 3.1.13.0, a signed 32-bit integer overflow in the loop index expression i * 4 inside SwapRGBABytes() causes the function to compute a large negative pointer offset when processing kABGR DPX images with large dimensions. The immediate crash is an out-of-bounds read (the memcpy at line 45 reads from &input[i * 4] first), but the subsequent write operations at lines 46–49 target the same wrapped offset — making this a combined OOB read+write primitive. This vulnerability is fixed in 3.0.18.0 and 3.1.13.0.

OpenImageIO is a toolset for reading, writing, and manipulating image files of any image file format relevant to VFX / animation. Prior to 3.0.18.0 and 3.1.13.0, a signed 32-bit integer overflow in the pixel-loop index expression i * 3 inside ConvertCbYCrYToRGB() causes the function to compute a large negative pointer offset into the output buffer, producing an out-of-bounds write that crashes the process. This vulnerability is fixed in 3.0.18.0 and 3.1.13.0.

OpenImageIO is a toolset for reading, writing, and manipulating image files of any image file format relevant to VFX / animation. Prior to 3.0.18.0 and 3.1.13.0, a signed integer overflow in QueryRGBBufferSizeInternal() in DPXColorConverter.cpp leads to a heap-based out-of-bounds write when processing crafted DPX image files. The function computes buffer sizes using 32-bit signed integer arithmetic with negative multipliers (e.g., pixels * -3 * bytes for kCbYCr descriptors and pixels * -4 * bytes for kABGR descriptors), where a negative result is used as an in-band signal that no separate buffer is needed. When the pixel count is sufficiently large, the multiplication overflows INT_MIN and wraps to a small positive value. The caller in dpxinput.cpp interprets this positive value as a required buffer size, allocates an undersized heap buffer via m_decodebuf.resize(), and then writes the full image data into it via fread, resulting in a heap buffer overflow. An attacker can exploit this by crafting a DPX file that triggers the overflow, causing a denial of service (crash) or potentially arbitrary code execution through heap corruption in any application that reads pixel data using OpenImageIO. This vulnerability is fixed in 3.0.18.0 and 3.1.13.0.

OpenImageIO is a toolset for reading, writing, and manipulating image files of any image file format relevant to VFX / animation. Prior to 3.0.18.0 and 3.1.13.0, a heap-based buffer overflow in the HEIF decoder of OpenImageIO allows out-of-bounds writes via crafted images due to a subimage metadata mismatch, leading to memory corruption and potential code execution. This vulnerability is fixed in 3.0.18.0 and 3.1.13.0.

OpenImageIO is a toolset for reading, writing, and manipulating image files of any image file format relevant to VFX / animation. Prior to 3.0.18.0 and 3.1.13.0, jpeg2000input.cpp:395 computes buffer size as const int bufsize = w * h * ch * buffer_bpp using signed 32-bit arithmetic. When the product exceeds INT_MAX, the result wraps to 0 or a small value. m_buf.resize() allocates an undersized buffer, and subsequent pixel write loops cause heap overflow. Conditional on USE_OPENJPH build flag. This vulnerability is fixed in 3.0.18.0 and 3.1.13.0.

OpenImageIO is a toolset for reading, writing, and manipulating image files of any image file format relevant to VFX / animation. Prior to 3.0.18.0 and 3.1.13.0, softimageinput.cpp:469 (mixed RLE) and :345 (pure RLE) do not clamp the run length to remaining scanline width before writing pixels. The raw packet path (line 403) correctly clamps with std::min, but RLE paths skip this check. A crafted .pic file causes heap overflow up to 65535 bytes. This vulnerability is fixed in 3.0.18.0 and 3.1.13.0.

OpenImageIO is a toolset for reading, writing, and manipulating image files of any image file format relevant to VFX / animation. Prior to 3.0.18.0 and 3.1.13.0, sgiinput.cpp:265,274 use OIIO_DASSERT for bounds checking in the RLE decode loop. In release builds, OIIO_DASSERT compiles to ((void)sizeof(x)) (dassert.h:210), making all bounds checks no-ops. A crafted .sgi file with RLE count exceeding scanline width causes heap buffer overflow and crash. This vulnerability is fixed in 3.0.18.0 and 3.1.13.0.

Timing limitations of the HRNG in RS9116 when power save mode is enabled results in predictable values

Fleet is open source device management software. Prior to version 4.81.0, a vulnerability in Fleet's software installer pipeline could allow a crafted software package to execute arbitrary commands as root (macOS/Linux) or SYSTEM (Windows) on managed endpoints when an uninstall is triggered. When a software package (.pkg, .deb, .rpm, .exe, or .msi) is uploaded to Fleet, metadata is extracted from the package binary and used to generate uninstall scripts. In affected versions, this metadata is not properly sanitized before being included in the generated scripts. A specially crafted package containing malicious values in its metadata fields could result in unintended command execution when the uninstall script runs on managed endpoints. Version 4.81.0 contains a patch. If an immediate upgrade is not possible, administrators should avoid uploading software packages obtained from untrusted or unverified sources. Additionally, administrators can manually inspect and edit auto-generated uninstall scripts before deployment.

Fleet is open source device management software. Prior to version 4.81.0, Fleet contained a denial-of-service (DoS) issue in the gRPC Launcher `PublishLogs` endpoint. In affected versions, certain unexpected input values were not handled gracefully, which could cause the Fleet server process to terminate while processing an authenticated request from an enrolled Launcher host. An authenticated attacker with access to any enrolled Launcher node key could cause an immediate and complete denial of service by sending a single gRPC request to the `PublishLogs` endpoint. This vulnerability impacts availability only. There is no exposure of sensitive data, no authentication bypass, no privilege escalation, and no integrity impact. Version 4.81.0 contains a patch. If upgrading immediately is not possible, the following mitigations can reduce exposure. Restrict network access to the Fleet gRPC endpoint where feasible (for example, limiting inbound access to known host IP ranges); deploy Fleet behind infrastructure that terminates or filters gRPC traffic if Launcher log ingestion is not required; and/or monitor for repeated Fleet process crashes or unexpected restarts indicating potential exploitation.

Fleet is open source device management software. Prior to version 4.82.0, a vulnerability in Fleet's Windows MDM enrollment flow allows authentication tokens from any Azure AD tenant to be accepted. Because Fleet validates JWT signatures using Microsoft's multi-tenant JWKS endpoint but does not enforce the `aud` (audience) or `iss` (issuer) claims, any Microsoft-signed Azure AD access token containing the expected scopes can be used to authenticate to Fleet's MDM endpoints. If Windows MDM is enabled, an attacker with access to any Azure AD tenant can obtain a valid Microsoft-signed token and use it to enroll unauthorized devices and interact with Fleet's MDM management APIs. During device management, Fleet may expose sensitive enrollment secrets embedded in MDM command payloads, enabling further unauthorized access. Version 4.82.0 contains a patch. If an immediate upgrade is not possible, affected Fleet users should temporarily disable Windows MDM.

Fleet is open source device management software. Prior to version 4.80.1, Fleet trusted client-supplied IP address headers when determining the source IP for incoming requests. This allowed authenticated and unauthenticated clients to spoof their apparent IP address and bypass per-IP rate limiting controls. Fleet determines a client’s public IP address using HTTP headers such as X-Forwarded-For, X-Real-IP, and/or True-Client-IP. These headers were trusted without validation. An attacker could supply arbitrary values in these headers, causing Fleet to treat each request as originating from a different IP address. This could allow an attacker to bypass per-IP rate limits and increase the effectiveness of brute-force or password-spraying attempts against authentication endpoints. This issue does not allow authentication bypass, privilege escalation, data exposure, or remote code execution on its own. Version 4.80.1 contains a patch. As a workaround, run Fleet behind a trusted reverse proxy or load balancer that overwrites client IP headers.

Crabbox prior to v0.12.0 contains an authentication bypass vulnerability that allows non-admin shared-token callers to impersonate other owners or organizations by spoofing identity headers. Attackers can inject malicious X-Crabbox-Owner and X-Crabbox-Org headers in requests authenticated with a shared token to bypass authorization checks and access owner/org-scoped lease operations belonging to victim accounts.

SiYuan is an open-source personal knowledge management system. Prior to 3.7.0, SiYuan's Bazaar (community marketplace) renders the name and version fields of a package's plugin.json (and the equivalent theme.json / template.json / widget.json / icon.json) into the Settings → Marketplace UI without HTML escaping. The kernel-side helper sanitizePackageDisplayStrings in kernel/bazaar/package.go HTML-escapes only Author, DisplayName, and Description — Name and Version flow through to the renderer raw. The frontend at app/src/config/bazaar.ts substitutes them into HTML template strings via ${item.preferredName} / ${data.name} / v${data.version} and assigns the result to innerHTML. As a consequence, malicious HTML in either field is parsed and executed when a user opens the marketplace tab. This vulnerability is fixed in 3.7.0.

SiYuan is an open-source personal knowledge management system. Prior to 3.7.0, SiYuan publish-mode Reader can mutate Conf and SQL index via 8 ungated APIs. POST /api/graph/getGraph, POST /api/graph/getLocalGraph, POST /api/sync/setSyncInterval, POST /api/storage/updateRecentDocViewTime, POST /api/storage/updateRecentDocCloseTime, POST /api/storage/updateRecentDocOpenTime, POST /api/storage/batchUpdateRecentDocCloseTime, and POST /api/search/updateEmbedBlock are registered with model.CheckAuth only, omitting both model.CheckAdminRole and model.CheckReadonly. Each of them writes server-side state, including atomic rewrites of <workspace>/conf/conf.json via model.Conf.Save(). Any caller whose JWT passes CheckAuth, including a publish-service RoleReader (the role assigned to anonymous publish visitors) and a RoleEditor against a workspace where Editor.ReadOnly = true, can hit them This vulnerability is fixed in 3.7.0.

SiYuan is an open-source personal knowledge management system. Prior to 3.7.0, broken access control in the searchAsset, searchTag, searchWidget, and searchTemplate publish-mode Readers can enumerate metadata from documents that are invisible to the publish service. This vulnerability is fixed in 3.7.0.

SiYuan is an open-source personal knowledge management system. Prior to 3.7.0, POST /api/tag/getTag is registered with model.CheckAuth only, omitting both model.CheckAdminRole and model.CheckReadonly, despite the handler performing a configuration write that is normally guarded by both. Any authenticated user — including publish-service RoleReader accounts and RoleEditor accounts on a read-only workspace — can call this endpoint with a sort argument to mutate model.Conf.Tag.Sort and trigger model.Conf.Save(), which atomically rewrites the entire workspace conf.json. This vulnerability is fixed in 3.7.0.

SiYuan is an open-source personal knowledge management system. Prior to 3.7.0, the kernel stores Attribute View (AV / database) names without any HTML escape, then a render template uses raw strings.ReplaceAll(tpl, "${avName}", nodeAvName) to embed the name in HTML before pushing to all clients via WebSocket. Three independent client paths (render.ts:120 → outerHTML, Title.ts:401 → innerHTML, transaction.ts:559 → innerHTML) consume the value without escaping. Because the main BrowserWindow runs nodeIntegration:true, contextIsolation:false, webSecurity:false (app/electron/main.js:407-411), HTML injection in the renderer becomes Node.js code execution. This vulnerability is fixed in 3.7.0.

Live Helper Chat is an open-source application that enables live support websites. In 4.84v, the Live Helper Chat REST API chat update endpoint allows a REST user with lhchat/use to update a chat in a department they cannot read. The endpoint accepts arbitrary chat object fields, so the user can change the chat hash and status and then access or tamper with the chat through visitor/widget paths. The same write primitive can set operation_admin, which is later emitted as operator-side JavaScript.

Gradient is a nix-based continuous integration system. In 1.1.0, when GRADIENT_DISCOVERABLE=true (the default, and the NixOS module default), anyone who can reach /proto can register as a worker without any credentials by sending a fresh, never-registered worker UUID. The resulting session has PeerAuth::Open, i.e. it sees jobs from every organisation, and can immediately NarPush/NarUploaded arbitrary store paths into nar_storage and the cached_path table. This vulnerability is fixed in 1.1.1.

Nuxt OG Image generates OG Images with Vue templates in Nuxt. The isBlockedUrl() denylist introduced in nuxt-og-image@6.2.5 to remediate GHSA-pqhr-mp3f-hrpp (Dmitry Prokhorov / Positive Technologies, March 2026) is incomplete. It has an incomplete IPv6 prefix list and is missing redirect re-validation. This vulnerability is fixed in 6.4.9.

SiYuan is an open-source personal knowledge management system. Prior to 3.7.0, he tooltip mouseover handler in app/src/block/popover.ts reads aria-label via getAttribute and passes it through decodeURIComponent before assigning to messageElement.innerHTML in app/src/dialog/tooltip.ts:41. The encoder used at the producer side, escapeAriaLabel in app/src/util/escape.ts:19-25, only handles HTML special characters (", ', <, literal &lt;) — it leaves %XX URL-escapes untouched. So a doc title containing %3Cimg src=x onerror=...%3E round-trips through escapeAriaLabel and the HTML attribute layer unmodified. Then decodeURIComponent on the consumer side converts %3C to a literal < character (a real <, NOT a character reference). When that string is assigned to innerHTML, the HTML5 tokenizer enters TagOpenState on the literal <, parses the <img> element, and the onerror handler fires. Because the renderer runs with nodeIntegration: true, contextIsolation: false, webSecurity: false (app/electron/main.js:407-411), require('child_process') is reachable from the injected handler, escalating to arbitrary code execution.This vulnerability is fixed in 3.7.0.

SiYuan is an open-source personal knowledge management system. From 2.1.12 to before 3.7.0. SiYuan's Bazaar marketplace renders package author metadata from the public bazaar stage feed into HTML without escaping. In the desktop app this becomes stored XSS, and because SiYuan's Electron windows are created with nodeIntegration: true and contextIsolation: false, a successful payload can call Node.js APIs and execute code on the host. This vulnerability is fixed in 3.7.0.

Note Mark is an open-source note-taking application. Prior to 0.19.4, no minimum length or entropy is enforced on the JWT_SECRET configuration value. The application accepts any base64-decodable secret regardless of size, including secrets as short as 1 byte. This vulnerability is fixed in 0.19.4.

Note Mark is an open-source note-taking application. From 0.13.0 to before 0.19.4, the Note Mark application allows authenticated users to upload assets to notes via POST /api/notes/{noteID}/assets, where the asset filename is provided through the X-Name HTTP request header. This value is stored directly in the database without any sanitization or validation - no path separator filtering, no directory traversal sequence rejection, and no use of filepath.Base() to strip directory components. The unsanitized name is persisted as-is in the note_assets table (Name column, varchar(80)). When an administrator subsequently runs the data export CLI commands (note-mark migrate export-v1 or note-mark migrate export), the stored asset name is passed directly into filepath.Join() and path.Join() calls as part of the output file path argument to os.Create(). Since Go's filepath.Join() resolves ../ sequences during path normalization, an attacker-controlled asset name containing directory traversal sequences causes the export process to write files to arbitrary locations on the filesystem, completely outside the intended export directory. This vulnerability is fixed in 0.19.4.

Foscam VD1 Video Doorbell before V5.3.13_1072 is vulnerable to Cleartext Transmission of Sensitive Information. The device transmits sensitive Session Description Protocol (SDP), including ICE credentials and candidates, in cleartext over network interfaces. An attacker with network visibility can intercept these credentials to hijack media streams or authenticate to Foscam's TURN/relay infrastructure to forward arbitrary traffic at the vendor's expense.

Strapi is an open source headless content management system. Strapi versions starting in 4.0.0 and prior to 5.37.0 did not sufficiently sanitize query parameters when filtering content via relational fields. An unauthenticated attacker could use the `where` query parameter on any publicly-accessible content-type with an `updatedBy` (or other admin-relation) field to perform a boolean-oracle attack against private fields on the joined `admin_users` table, including the `resetPasswordToken` field. Extracting an admin reset token via this oracle made full administrative account takeover possible without authentication. When a filter such as `where[updatedBy][resetPasswordToken][$startsWith]=a` was applied to a public Content API endpoint, the underlying query generation performed a `LEFT JOIN` against the `admin_users` table and emitted a `WHERE` clause referencing the joined column. The query parameter sanitization layer did not block operator chains that traversed into relational target schemas the caller had no read permission on, allowing the response count to be used as a one-bit oracle on any admin-table field. The patch in version 5.37.0 introduces explicit query-parameter sanitization at the controller and service boundary via three new primitives: `strictParam`, `addQueryParams`, and `addBodyParams`. Operator chains that traverse into restricted relational targets are now rejected before reaching the database.

Due to improper input handling under certain conditions, SAP NetWeaver Application Server ABAP allows an attacker to inject custom Cascading Style Sheets (CSS) data into a web page served by the application. When a user accesses or clicks the affected page, the injected CSS is executed. As a result, the issue has a low impact on confidentiality, while integrity and availability are not impacted.

Fleet is open source device management software. Prior to version 4.81.0, a vulnerability in Fleet’s Windows MDM management endpoint could allow requests to be processed without proper client certificate validation. In certain circumstances, this could allow an attacker to impersonate an enrolled Windows device and retrieve sensitive configuration data. Fleet’s Windows MDM management endpoint relies on mutual TLS (mTLS) client certificates to authenticate enrolled devices. In affected versions, requests that did not present a client certificate could be incorrectly treated as trusted. As a result, an attacker with prior knowledge of a valid enrolled device identifier could potentially impersonate that device and receive configuration payloads intended for it. These payloads may contain sensitive information such as Wi-Fi or VPN configuration data, certificates, or other secrets delivered through MDM profiles. This issue does not allow enrollment of new devices, administrative access to Fleet, or compromise of the Fleet control plane. Impact is limited to the targeted Windows device. Version 4.81.0 contains a patch. If an immediate upgrade is not possible, affected Fleet users should temporarily disable Windows MDM.

Strapi is an open source headless content management system. In Strapi versions prior to 5.33.3, the Upload plugin's Content API endpoints did not enforce the administrator-configured MIME type restrictions (`plugin.upload.security.allowedTypes` and `deniedTypes`). The same restrictions were correctly enforced on the Admin Panel upload path. The upload plugin's `enforceUploadSecurity` security check was invoked in the admin upload controller but was missing from the Content API controller. The Content API handlers `uploadFiles` and `replaceFile` (and the `upload` wrapper that dispatches to them) called the underlying upload service directly, bypassing both the magic-byte MIME detection and the configured allow/deny lists. An authenticated user with the Content API upload permission could therefore upload file types the administrator had explicitly disallowed, including HTML and SVG content. In deployments serving uploaded files from the same origin as the admin panel (default), an attacker could upload an HTML or SVG file that, when opened directly by an admin, executed JavaScript in the admin origin, enabling admin-session hijack and authenticated administrative actions against the admin API. The patch in version 5.33.3 introduces a shared `prepareUploadRequest` helper that wraps `enforceUploadSecurity` and is called from both the Content API and admin upload controllers, ensuring identical security policy enforcement on every upload entry point.

Strapi is an open source headless content management system. In Strapi versions prior to 5.33.3, changing or resetting a user's password did not invalidate the user's existing refresh-token sessions by default. The refresh-token invalidation step in the users-permissions and admin authentication controllers was conditional on a caller-supplied `deviceId`. When a password change or reset request did not include a `deviceId`, no refresh tokens were revoked, leaving every prior session active. An attacker who had previously obtained a refresh token could continue minting new access tokens after the legitimate user reset their password, allowing persistent unauthorized access for the lifetime of the refresh token (up to 30 days by default). Rotating credentials no longer terminated an active attacker session, defeating password reset as a containment measure. The patch in version 5.33.3 invalidates all refresh tokens associated with the user on every password change and password reset, regardless of whether a `deviceId` is supplied. A new device-scoped session is then issued to the caller as part of the response.

Strapi is an open source headless content management system. In versions on the 4.x branch prior to 4.26.1 and on the 5.x branch prior to 5.33.2, a database-query injection vulnerability existed in the Strapi Content-Type Builder write API. An authenticated administrator could inject arbitrary database statements through the `column.defaultTo` attribute when creating or modifying a content type. Setting `defaultTo` as a tuple `[value, { isRaw: true }]` caused the value to be passed directly into Knex's `db.connection.raw()` during schema migration without sanitization, allowing arbitrary statement execution at the database layer. Depending on the database engine, this enabled arbitrary file read via database utility functions, denial of service via forced server crash on schema-migration error, and on engines that permit external program execution, remote code execution against the database server. The patch in versions 4.26.1 and 5.33.2 addresses this by restricting all Content-Type Builder write APIs to development mode only. Production deployments running v5.33.2 or later return 404 for requests against `/content-type-builder/content-types` and related endpoints, removing the network-reachable attack surface entirely.

Strapi is an open source headless content management system. In Strapi versions prior to 5.45.0, the rate-limit middleware in the users-permissions plugin derived its rate-limit key in part from `ctx.request.body.email`, including on routes whose body schema does not contain an `email` field (`/auth/local`, `/auth/reset-password`, `/auth/change-password`). An unauthenticated attacker could include an arbitrary `email` value in the request body to obtain a fresh rate-limit key per request, effectively bypassing per-IP throttling on those routes and enabling high-volume credential brute-force, password-reset code brute-force, and credential-stuffing attempts. The rate-limit key was constructed as `${userIdentifier}:${requestPath}:${ctx.request.ip}`, where `userIdentifier = ctx.request.body.email`. On routes that legitimately use email as their identifier (e.g. `/auth/forgot-password`, `/auth/local/register`), this scoping is correct. On routes that use a different identifier (`identifier` for login, `code` for password reset, `currentPassword` for password change), the email field was not part of the route contract, but the middleware still incorporated it into the key, allowing a caller to rotate the value and obtain a unique key on every request. The patch in version 5.45.0 maintains an allow-list of routes that legitimately key on the email field and excludes that key component on every other route the middleware is mounted on. OAuth callback paths (`/connect/*`) are treated identifier-less. On routes outside the allow-list, the middleware now falls back to a fixed identifier-less key, ensuring per-IP throttling remains effective even when the request body is attacker-controlled.