GitHub Advisory Database의 정보
5페이지
10월 20일 (화)
Mercury: Unauthenticated Remote Code Execution via /api/kernels
影響: pip:mercury <= 3.2.5 → 3.2.6 Mercury runs a JupyterLab server with anonymous access by default (mercury_app/__main__.py sets --IdentityProvider.token='' when no --token/MERCURY_TOKEN is given; when tokenless it also sets --ServerApp.disable_check_xsrf=True). It relies on a blocklist (mercury_app/block_handler.py BLOCKED_PATTERNS: /lab, /api/terminals, /api/shutdown, /files) to keep anonymous users off dangerous Jupyter endpoints. The vulnerability /api/kernels is NOT in the blocklist (and cannot be — Mercury needs kernels…
LoopX serve-status permissive CORS policy exposes control-plane state and private Markdown files
影響: pip:loopx >= 0.0.1, < 0.4.5 → 0.4.5 loopx serve-status sends Access-Control-Allow-Origin: on every response, including its unauthenticated GET routes. Because the endpoint is unauthenticated and bound to a fixed, well-known loopback port, any website the operator visits can read LoopX control-plane state and the contents of Markdown files inside every registered goal repo, using a plain fetch() with no user interaction. The write endpoints are not affected — they are correctly Origin-checked. This report is strictly about…
10월 17일 (토)
Arcane: Delegated users:update permission allows admin account takeover via password reset (no privilege-hierarchy check)
影響: go:github.com/getarcaneapp/arcane/backend/v2 < 2.6.0 → 2.6.0 A user holding the delegated users:update permission can reset the global administrator's password and take over the admin account. UpdateUser sets the target user's PasswordHash from the request body for ANY target with no comparison of caller-vs-target privilege. Because role assignment and role CRUD are deliberately gated by RequireGlobalAdmin, users:update is intended to sit BELOW admin (profile management); this finding defeats that separation. Root Cause UpdateUser (handler backen…
Arcane: Environment access token bypasses the admin-required authorization gate on `/api/debug/pprof`
影響: go:github.com/getarcaneapp/arcane/backend/v2 < 2.11.0 → 2.11.0 Arcane v2.10.2 contains an authorization inconsistency in AuthMiddleware.managerAuth. When a request supplies a valid environment access token through the X-Arcane-Agent-Token header, managerAuth resolves the associated environment, installs an environment-scoped permission set, and immediately calls next(c). This occurs before the middleware evaluates the AdminRequired option and verifies that the caller is a global administrator. As a result, an environment-scoped access token can acc…
LMCache: Remote code execution via unsafe deserialization in multiprocess mode transport
影響: pip:lmcache >= 0.3.9, < 0.5.6 → 0.5.6 LMCache is vulnerable to Unauthenticated Remote Code Execution via Pickle Deserialization on the Multiprocess ZMQ Transport Details LMCache multiprocess mode, also called distributed mode, opens an unauthenticated ZeroMQ ROUTER so worker processes can register and share KV cache blocks. The intended clients are sibling LMCache processes. There is no CURVE, ZAP, password, or message authentication on that socket. LMCache used only inside a vLLM process does not open this port. The transp…
kin-openapi openapi3filter: nil-pointer panic in ValidateResponse on a response header declared with content
影響: go:github.com/getkin/kin-openapi <= 0.145.0 → 0.146.0 ValidateResponse panics on a nil pointer when a spec declares a response header with content instead of schema. doc.Validate() accepts that document, so the crash comes from a spec the library just called valid. Same class as GHSA-jpcw-4wr7-c3vq, which covered the request-parameter case in v0.144.0; this is the response-header path. Affects github.com/getkin/kin-openapi 0 { Two things I checked rather than assumed. The response does not need to carry the header; declaring it is enough,…
Wasmtime: WASI hard links bypass wasmtime-wasi's FilePerms for destination
影響: rust:wasmtime-wasi < 24.0.11 → 24.0.11, rust:wasmtime-wasi >= 25.0.0, < 36.0.12 → 36.0.12, rust:wasmtime-wasi >= 37.0.0, < 45.0.3 → 45.0.3, rust:wasmtime-wasi >= 46.0.0, < 46.0.1 → 46.0.1 Wasmtime's WASI filesystem implementation lets a guest with a read-only source file capability overwrite that file by hard-linking it into another preopen that allows file writes. An attacker who can run a WASI guest can bypass WasiCtxBuilder per-preopen FilePerms and modify host files that were exposed as FilePerms::READ. This attack is possible through the wasip1, wasip2, and wasip3 interfaces to the WASI filesystem. This attack is also possible through an indirection of renaming a ha…
lopdf has a stack overflow via deeply nested PDF objects
影響: rust:lopdf < 0.42.0 → 0.42.0 lopdf::Document::load_mem (and the other load entry points) parses nested PDF arrays and dictionaries with unbounded recursion. A small crafted PDF whose Catalog contains a deeply nested array (/X [[[ … ]]], on the order of 10,000 levels) exhausts the call stack and aborts the process with SIGABRT. Because this is a stack-overflow abort rather than a panic!, it cannot be caught with catch_unwind: any service that parses untrusted PDF input with lopdf can be crashed by a ~21 KB file, resulting i…
JavaCV: Integer Overflow in IPCameraFrameGrabber Causes Infinite-Loop DoS
影響: maven:org.bytedeco:javacv < 1.5.14 → 1.5.14 IPCameraFrameGrabber parses the Content-Length value from the multipart MJPEG HTTP stream returned by the configured IP camera without any bounds checking. A malicious or compromised camera endpoint - or any network attacker able to tamper with the plaintext HTTP response (these streams are conventionally accessed over http://, as documented in the class itself) - can supply a Content-Length value in the range 1073741825-2147483647. This drives the internal buffer-growth routine ensureB…
Claude Agent SDK: Argument Injection via resume Option Allows Arbitrary Command Execution
影響: pip:claude-agent-sdk < 0.2.121 → 0.2.121 The Claude Agent SDK for Python passed the resume option to the Claude CLI without validation, allowing a value beginning with - to be interpreted as a separate CLI flag. An attacker who controlled the resume value could inject an --mcp-config flag defining a stdio MCP server, resulting in arbitrary command execution at CLI startup that bypassed the SDK's permission controls. Exploitation required an application to pass untrusted input as the session ID to resume. Users are advised to update to…
CoreDNS `k8s_external` headless AXFR can emit an empty transfer batch that panics the `transfer` plugin
影響: go:github.com/coredns/coredns >= 1.9.4, <= 1.14.4 → 1.14.5 A network DNS client allowed to request AXFR for a CoreDNS zone can trigger a panic when CoreDNS is configured with k8s_external headless-service zone transfers and Kubernetes contains a headless service endpoint with no declared ports. The built-in k8s_external transferer can emit an empty []dns.RR batch, and the shared transfer plugin indexes records[0] without checking that the batch is non-empty. In normal CoreDNS mode the panic is recovered and returned as SERVFAIL, but repeated tr…
ckan-mcp-server: DataStore Table UI resource does not validate postMessage origin (in) and posts to targetOrigin "*" (out)
影響: npm:@aborruso/ckan-mcp-server < 0.4.112 → 0.4.112 The MCP Apps UI resource (an interactive DataStore table rendered in the host's webview/iframe) communicates with its parent via postMessage. Its inbound message listener performs no event.origin validation, and its outbound messages are sent with target origin ''. Any window/frame able to reach the document can inject data into the table and interfere with the MCP UI handshake, and the UI's own JSON-RPC messages are broadcast without an origin restriction. Affected code src/resources/d…
swoosh has email-derived URL path injection in Microsoft Graph endpoint
影響: erlang:swoosh >= 1.12.0, < 1.26.3 → 1.26.3 Swoosh.Adapters.MsGraph builds its Microsoft Graph API request URL by interpolating the sender's email address directly into the URL path (/users/{from}/sendMail) without percent-encoding or validation. In applications that derive the from address from untrusted or user-influenced input (a relay, a contact form, or a "send as" feature), an attacker can inject URL-special characters into the local part of the address to escape the intended path segment and rewrite the request's path and…
ClickMigrate has Path Traversal in "revision -m" argument
影響: pip:ClickMigrate < 1.0.3 → 1.0.3 A path traversal vulnerability in the revision command allowed a crafted migration name supplied via the -m argument to escape the intended migrations directory. Under specific conditions, an attacker could cause the tool to create a .sql file in an arbitrary location writable by the user executing the command. This issue is only exploitable if an attacker can influence the migration name passed to the revision command, such as through an insecure automation or CI/CD pipeline that forwar…
gomarkdown/markdown IsSafeURL panics on short URLs
影響: go:github.com/gomarkdown/markdown < 0.0.0-20260724235538-61af6e1e6971 → 0.0.0-20260724235538-61af6e1e6971 IsSafeURL slices url[:nPath] before checking that len(url) >= nPath, so a URL shorter than a path prefix (0, 1 or 2 bytes) triggers a slice-bounds panic, and this is reached end to end when the html.Safelink hardening flag is enabled and the input contains an empty or very short link destination, giving an unrecoverable crash DoS. Confirmed at HEAD (e08cff8; present since 2022). The defect parser/parser.go, IsSafeURL: go func IsSafeURL(url []byte) bool { nLink := len(url) for _, path :=…
10월 16일 (금)
TanStack Router: Unauthenticated reflected XSS in TanStack Start server-function responses
影響: npm:@tanstack/start-server-core >= 1.143.12, < 1.169.39 → 1.169.39, npm:@tanstack/react-start >= 1.143.12, < 1.168.60 → 1.168.60, npm:@tanstack/solid-start >= 1.143.12, < 1.168.57 → 1.168.57, npm:@tanstack/vue-start >= 1.143.12, < 1.168.56 → 1.168.56 An unauthenticated attacker can craft a URL to a server function in an affected TanStack Start app. If someone opens the link, the app can return attacker-controlled HTML from its own origin and run the attacker's JavaScript there. The script can access same-origin data available to that visitor and make requests with their authority. Cause The server-function transport can pass request payload fields into internal middleware state instead of limiting them to public input fields. A malfo…
Pipecat: Public exposure of Pipecat’s development runner can allow unauthenticated bot session creation
影響: pip:pipecat-ai >= 0.0.98, <= 1.5.0 → 1.6.0 Pipecat’s built-in development runner exposes endpoints that create bot sessions and, for the Daily transport, paid Daily rooms. These development endpoints intentionally do not authenticate callers. The runner binds to localhost by default and is designed exclusively for local development. It is not built or supported as a production server. Exploitation therefore requires an operator to deliberately make the development runner reachable from an untrusted network, such as by setting --host 0.0…
Podman: The `podman run` command can be instructed to disable almost all sandboxing by image annotation
影響: go:github.com/containers/podman >= 4.4.0, < 5.8.8 → 5.8.8, go:go.podman.io/podman/v6 >= 6.0.0, < 6.1.3 → 6.1.3 Podman v4.4.0 added a feature where container checkpoints (from the podman container checkpoint command) could be distributed via OCI registry, and pulled from that registry and run via the podman run command. This allowed easy transfer of checkpoints between systems. However, checkpoints allow complete control of how the container is created - including full control of the container sandbox. When podman run detects a checkpoint, it will silently ignore all user input regarding how the c…
JupyterLab: Command link trust prompt in JupyterLab can trust a document other than the one clicked
影響: pip:jupyterlab >= 4.6.0, <= 4.6.3 → 4.6.4, pip:jupyterlab >= 4.5.7, <= 4.5.10 → 4.5.11, pip:notebook >= 7.5.6, <= 7.6.2 → 7.6.3, pip:jupyterlite-core >= 0.7.5, <= 0.8.3 → 0.8.4 A Markdown file or a notebook output can carry a button that runs a JupyterLab command. Since 4.5.7 the first click on such a button opens a prompt naming the command, with Cancel, Run Once and Trust. Choosing Trust runs the command and marks the document trusted, so its later buttons run without a prompt. Which document gets marked is decided when the user answers, not when they click. Both trust commands read whichever document is current at that moment: markdownviewer:trust uses…
SFTPGo source-side file-pattern filter bypass during copy operations
影響: go:github.com/drakkan/sftpgo/v2 >= 2.5.0, <= 2.7.5 → 2.7.6 File pattern filters (denied_patterns / allowed_patterns) were evaluated on the destination of a copy only. A user holding the copy permission could copy a file denied by the filters to a permitted name and then read it. A recursive directory copy reproduced the contents of a restricted directory under a path where the filters do not deny them, and the hide policy was affected in the same way: a file reported as missing by stat and by directory listings could still be copied. Details An…
Claude Code: Plugin Hash Pin Bypass Allows Bypass of Hash Pinning
影響: npm:@anthropic-ai/claude-code < 2.1.179 → 2.1.179 The plugin install flow in Claude Code and Claude Desktop checked out a marketplace's pinned commit SHA using git checkout without verifying that the resulting HEAD matched the pinned SHA. An attacker controlling a plugin's upstream repository could create a branch named identically to the pinned 40-character SHA, causing the checkout to silently resolve to the attacker's branch tip instead of the reviewed commit, resulting in attacker-controlled code being installed bypassing the pinned hash.…
containerd: Unpack DoS via repeated opaque whiteouts
影響: go:github.com/containerd/containerd/v2 < 2.0.12 → 2.0.12, go:github.com/containerd/containerd < 1.7.35 → 1.7.35, go:github.com/containerd/containerd/v2 >= 2.2.0, < 2.2.8 → 2.2.8, go:github.com/containerd/containerd/v2 >= 2.3.0, < 2.3.5 → 2.3.5 A vulnerability exists in containerd's layer unpack handler where the processing of certain image layers causes superlinear time complexity. During the unpack operation, specific file configurations within the layer can trigger repeated directory traversals. Consequently, pulling or importing a maliciously crafted image can result in increased unpack times and a moderate increase in CPU usage. This delays container startup, even if the compressed image layers are small. Additionally, in…
Distribution: Client `manifests.Get()` does not verify returned content against requested digest
影響: go:github.com/docker/distribution <= 2.8.3, go:github.com/distribution/distribution/v3 < 3.1.2 → 3.1.2 registry/client.(manifests).Get(ctx, dgst, …) fetches a manifest by digest but never hashes the returned body and compares it against the requested digest. A malicious or compromised registry, or any MITM on the registry connection, can serve arbitrary manifest bytes in response to a digest-pinned request and the client returns the substituted content transparently to the caller. Any application relying on digest pinning for supply-chain integrity — image policy controllers, mirrors, pu…
Qwik: Crashing any Qwik Server via Resource thenable
影響: npm:@builder.io/qwik < 1.20.0 → 1.20.0, npm:@qwik.dev/core < 2.0.0-beta37 → 2.0.0-beta37 A Qwik server in the default configuration can be crashed with one unauthenticated HTTP request. During qwik-json body parsing, ResourceSerializer.$fill$ does Promise.resolve(resource._resolved). If _resolved is a plain object whose .then is a QRL for an unregistered symbol, the async QRL rejects, the rejection is orphaned, and Node terminates the process via its default unhandled-rejection policy. Impact One unauthenticated request crashes the server. Looping keeps every replica dead a…
containerd CRI plugin: Checkpoint restore bypasses destination security context
影響: go:github.com/containerd/containerd/v2 >= 2.1.0, < 2.2.7 → 2.2.7, go:github.com/containerd/containerd/v2 >= 2.3.0, < 2.3.4 → 2.3.4 A vulnerability in containerd's CRI implementation allows a container restored from an untrusted checkpoint via the CreateContainer API to bypass the destination security context and execute with elevated privileges. When restoring a container from a checkpoint archive or annotated OCI image, CRIU restores process credentials, Linux capabilities, no_new_privs, and seccomp state directly from checkpoint data rather than enforcing the destination CRI ContainerConfig. An attacker who can ru…
Backstage: Improper access control in backend action invocation
影響: npm:@backstage/backend-defaults < 0.17.8 → 0.17.8 An authenticated user could invoke actions that an operator had excluded through action configuration. Existing action permissions and downstream authorization remain enforced. Patches Patched in @backstage/backend-defaults version 0.17.8 Workarounds Limit access to action invocation endpoints to trusted service callers until upgrading.
kin-openapi: Incomplete Fix for GHSA-p6wj-qrr4-pgh5: Three Unpatched nil-Items Panic Paths in openapi3filter
影響: go:github.com/getkin/kin-openapi >= 0.136.0, < 0.149.0 → 0.149.0 The fix for GHSA-p6wj-qrr4-pgh5 addressed a nil-pointer dereference that occurs when an OpenAPI 3.1 array schema omits the items keyword (which is legal under JSON Schema 2020-12). That fix patched three decoder paths: the standalone parseArray function, the top-level array check in UrlencodedBodyDecoder, and MultipartBodyDecoder. Three additional crash paths were missed. All of them reach the unguarded urlValuesDecoder.parseArray method (openapi3filter/req_resp_decoder.go:576) or the u…