Curated security, hosting and development news with vulnerability warnings from fixed trusted sources.
Articles are collected periodically from a fixed source list. D4rC Tools stores the headline, source summary and source link only; linked articles are not fetched on page view.
An unreleased Anthropic AI model found security flaws in Chrome, Safari, Android, Windows, and Linux and chained them into working proof-of-concept exploits, the company said in its latest report. The exploits were built in the lab after the model was prompted, not seen in any real attack, and Anthropic reported every flaw to the vendor first, so all of them are already patched. One proof of
The U.S. Department of Justice (DoJ) on Tuesday announced the seizure of a domain associated with a bulletproof hosting provider (BPH) named Virtualine. The move was in response to the service being used to commit and facilitate prolific criminal activity, including "the advertisement of sexually explicit images of minors, the intentional disclosure of nonconsensual intimate visual depictions
Japan's GMO Internet Group is offering up to $1 million in total for tips that help catch the attackers who breached its Onamae.com Desktop Cloud service, after refusing to pay an extortion demand. The reward program opened on October 28 and pays only if a tip helps lead to an arrest or an international wanted notice. The service hosts Windows desktops used for FX trading,
Tailscale has released version 1.104.4 to fix flaws in its built-in SSH server that could let an attacker reach services and files on other servers that only the root user should be able to touch. The flaws are in how Tailscale SSH forwards network connections, a job the Tailscale background process carries out with root privileges. To use them, an attacker must
The critical Oracle PeopleSoft bug that ShinyHunters has exploited since May is not one flaw but three chained together, the researchers who reported it said this week. Oracle patched all three under a single identifier, CVE-2026-35273, in June. Its root is a server-side request forgery (SSRF) that reaches an internal management interface because PeopleSoft trusts its own loopback address
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Tuesday added two vulnerabilities impacting Adobe Commerce and Magento, and Google's Brotli decoder library to its Known Exploited Vulnerabilities (KEV) catalog, citing evidence of active exploitation. The vulnerabilities are listed below - CVE-2026-111497 (CVSS score: 9.2) - An unchecked return value vulnerability in
### Summary On MLflow tracking servers running without authentication, which is the default for mlflow server, a remote attacker can read arbitrary files from the server’s filesystem by creating a model version tagged as a prompt whose source is a schemeless local path. The fix for CVE-2026-2614 in 3.10.0 added a prompt-specific branch to the CreateModelVersion handler that rejects file:// URIs and absolute paths but runs path validation only when the source carries a URL scheme. A schemeless value such as ../../etc, a plain relative name, ., a percent-encoded path, or a Windows drive-letter path is therefore stored unchanged as the model version’s artifact location, and the model-versions/get-artifact endpoint later serves files from it with the privileges of the server process…
### Summary mlflow.crewai.autolog() leaks CrewAI LLM provider API keys into MLflow trace telemetry. When a CrewAI Agent is instrumented with MLflow autologging, MLflow serializes the agent's __dict__ and converts its fields to strings. The llm field contains the CrewAI LLM object, whose string representation includes its api_key in plaintext. As a result, an application that calls mlflow.crewai.autolog() can unintentionally send its LLM provider API key to the MLflow trace backend or configured OTLP destination. The application does not need to explicitly log the credential or place it in a span attribute. The credential is extracted automatically by MLflow's instrumentation. This was verified on **MLflow 3.16.0** with **CrewAI 1.15.18** using MLflow's own OTLP exporter. The f…
## Summary MLflow 3.12.0 introduced registration and deserialization support for third-party scorers in OSS MLflow. During scorer deserialization, the server accepts a request-controlled Python module path beneath one of four allowed package prefixes, imports that module, obtains a request-controlled class, and instantiates it synchronously. An HTTP caller can combine this behavior with MLflow's built-in experiment, run, and artifact-upload endpoints: 1. Create an experiment whose local artifact location points to a writable allowed scorer package. 2. Create a run, causing MLflow to allocate //artifacts/. 3. Upload a Python module into that run artifact directory. 4. Submit a serialized third-party scorer whose module path resolves to the uploaded file. 5. Trigger server-side import and…
### Summary A regular expression used to obfuscate URL query strings can consume excessive CPU when processing certain attacker-controlled input. This can stall request processing and cause denial of service in applications using affected HTTP tracing integrations. ### Details The default query-string obfuscation pattern can exhibit excessive backtracking. The pattern is applied synchronously when recording HTTP URL metadata, allowing query-string processing to occupy application resources for an extended period. As a workaround, set DD_TRACE_OBFUSCATION_QUERY_STRING_REGEXP to the fixed regex below, then restart the application. ``` DD_TRACE_OBFUSCATION_QUERY_STRING_REGEXP='(?:(?:"|%22)?)(?:(?:old[-_]?|new[-_]?)?p(?:ass)?w(?:or)?d(?:1|2)?|pass(?:[-_]?phrase)?|secret|(?:api[-_]?|private[-_…
### Impact Stored XSS vulnerability in the media plugin. Attackers can inject malicious scripts via elements crafted through the data-mce-object attribute, which are created and executed before sanitization. Impacts users of TinyMCE with the media plugin enabled. ### Patches This vulnerability has been patched in TinyMCE 8.9.3 and TinyMCE 7.9.4 by ensuring that, when using the media plugin, any content created through the data-mce-object attribute is properly sanitized first. ### Workarounds No official workaround available. ### Fix To avoid this vulnerability: Upgrade to TinyMCE 8.9.3 or higher. Upgrade to TinyMCE 7.9.4 or higher. ### Acknowledgements Tiny thanks [Fariskhi Vidyan](https://github.com/farisv) and David Vieira Kurz (HiSolutions AG) for their help identifying this vulnerabil…
### Summary In affected versions, the SDK's server-side bearer authentication had no way to check that a token was issued for your server (its audience). Where one authorization server issues tokens for several services, a valid token for another service was also accepted by your MCP server. ### Am I affected? Yes, if your HTTP server uses the SDK's bearer authentication (auth=AuthSettings(...) with token_verifier=), its authorization server also issues tokens for other services, and your TokenVerifier does not check the token's audience (aud) itself. Affected versions: - mcp 1.10.0 through 1.29.1 - mcp 2.0.0 through 2.1.1 ### Fix Upgrade to: - 1.x: mcp 1.30.0 or later - 2.x: mcp 2.2.0 or later **Upgrading alone changes nothing.** The patched versions add a `validate_token…
## Summary JaegerPropagator.extract() in opentelemetry-propagator-jaeger did not enforce limits when extracting inbound uberctx-* baggage headers. Attacker-controlled baggage could cause unnecessary CPU work and short-lived heap allocations while processing more data than the propagator should accept. ## Impact Services that enable the optional Jaeger propagator and accept untrusted inbound propagation headers may experience increased per-request resource usage. This can contribute to denial-of-service risk when transport-level limits are absent or configured above the baggage limits. The practical impact for most Python web deployments is limited. The package must be installed separately, and the Jaeger propagator is not enabled by the default OpenTelemetry Python configuration. Co…
### Summary In affected versions, the SDK's Streamable HTTP and SSE client transports followed HTTP redirects to any host. If the configured MCP endpoint answered with a redirect to a different origin: - custom headers set on the client (for example an X-API-Key credential) were sent to that origin, and - on a 307 or 308 redirect the request body was re-sent there too. The SDK's OAuth client providers followed redirects on their own discovery, registration and token requests in the same way. A lone Authorization bearer header was not exposed; the HTTP library already drops it on cross-origin redirects. Anyone able to make an endpoint you trust redirect elsewhere (a stale redirect to a lapsed domain, a misrouted proxy rule, redirect logic an attacker can influence) could capture th…
### Summary In affected versions, when a connected MCP server declared an outputSchema for a tool, the SDK's client validated each call_tool result with jsonschema's default resolver, which fetches any $ref it cannot resolve locally by opening the URL. A server could therefore put an http://, https:// or file:// reference in a tool's output schema and make the client: - request that URL, or read that local file, from the client's own network position on every call of the tool (the content was used only as a schema and never sent back to the server), and - stall: the fetch ran synchronously on the event loop with no timeout, so a target that accepted the connection and then went quiet froze every session in the process for as long as it held the connection. ### Am I affected?…
### Impact A Cross-Site Scripting (XSS) vulnerability has been discovered in the CKEditor 5 engine. This vulnerability could be triggered by inserting specially crafted markup under a specific General HTML Support configuration that allows inserting objects, leading to unauthorized JavaScript code execution in a browser context isolated from the origin of the application embedding the editor. This vulnerability affects only installations where General HTML Support is enabled and configured to allow the affected HTML content. ### Patches The problem has been recognized and patched. The fix will be available in version 48.5.1 (and above) and 47.7.4 (LTS). ### For more information Email us at [[email protected]](mailto:[email protected]) if you have any questions or comments abou…
### Impact A Cross-Site Scripting (XSS) vulnerability has been discovered in the CKEditor 5 engine. This vulnerability is caused by a prototype pollution issue in the es-toolkit library used in the CKEditor 5 codebase. This vulnerability could lead to unauthorized JavaScript code execution when the editor processes incoming style attribute values. The underlying issue has been patched by the library maintainers, and the fix has been incorporated into CKEditor 5. ### Patches The problem has been recognized and patched. The fix will be available in version 48.5.1 (and above) and 47.7.4 (LTS). ### For more information Email us at [[email protected]](mailto:[email protected]) if you have any questions or comments about this advisory.
### Impact When legacy mode is enabled and session.implicit_hosting is enabled, the legacy WebRTC data channel incorrectly authorizes input events due to a logic error in LegacyIsHost(). As a result, an authenticated room member may be able to send mouse and keyboard input through the legacy WebRTC control channel even when controls are locked or the member would otherwise not be authorized through the modern control API. The issue affects only the legacy WebRTC input path. The modern HTTP/WebSocket control path correctly enforces authorization checks and is not affected. To exploit this issue, an attacker must: * Be authenticated as a room member. * Successfully establish a WebRTC connection. * Access a deployment using the legacy WebRTC input handler. No unauthenticated access or pr…
### Summary Atlantis passes the Terraform workspace value from a pull request's atlantis.yaml into a shell command without quoting or escaping it. A contributor whose pull request is eligible for an Atlantis plan can include shell metacharacters in the workspace name and execute arbitrary commands as the Atlantis service account. This does not require custom workflows, allow_custom_workflows, or permission to run terraform apply. The executed command inherits the Atlantis process environment, which commonly contains cloud-provider, Terraform, and VCS credentials. ### Details Verified against current upstream/main commit 92ce5c51d1a050357e44df3535483425e59fa984 on August 13, 2026. The vulnerable data flow is: 1. The PR-controlled project workspace is validated in [`server/core/…
## Summary File read paths resolve symbolic links with filepath.EvalSymlinks and then operate on the resolved target without re-checking that the target still lies inside the user's source/scope. A symlink located inside a user's accessible directory that points outside the scope (for example to /etc/passwd, another tenant's directory, SSH keys, or container secrets) is followed transparently, and its contents are served through the normal download, preview, media, and OnlyOffice endpoints. Any authenticated user with download access can read files anywhere the server process can, whenever such a symlink is reachable in their scope. The same symlink-following read path is used by the public share endpoints, which are authorized only by a share hash and not by any account, so when a pu…
### Impact When an originless Agent composes the redirect and cache interceptors as compose(redirect(), cache()) with a shared cache (type: 'shared'), the cache builds its key from the original request URL before the redirect handler changes the origin. The response produced after a redirect is then stored under the original key, even when the redirect crossed to a different origin. An attacker who controls the target of a redirect from a trusted URL can have their response cached under that trusted URL. A later caller requesting the trusted URL receives the attacker's body from the shared cache without either origin being contacted, allowing persistent cache poisoning. The same key confusion also discloses private data: when an authenticated request returns a private, no-store re…
### Impact undici's Agent passes its constructor options through a JSON-based deep clone (JSON.parse(JSON.stringify(...))) before forwarding them to each per-origin Pool. JSON cannot represent functions, so a caller-supplied tls option containing a checkServerIdentity callback is silently dropped before it reaches the TLS layer. As a result, a TLS peer whose certificate a custom checkServerIdentity was written to reject, but which passes Node's default hostname and chain checks, is silently accepted when the request is made through Agent. Because Agent is the default dispatcher behind request, fetch, and setGlobalDispatcher, this is the path most applications use. EnvHttpProxyAgent is also affected on its direct and NO_PROXY routes, which delegate to an internal…
## Impact undici crashes the client process when a TLS 1.2 server renegotiates after the connection is established and presents a different, CA-trusted certificate whose hostname does not match the requested host. Node.js re-runs hostname verification on the renegotiated handshake and destroys the socket with ERR_TLS_CERT_ALTNAME_INVALID. undici's established socket error handler asserted that this error code could never occur, so the peer-controlled error became an uncaught AssertionError that terminates the Node.js process. Both the HTTP/1.1 and HTTP/2 paths are affected. A malicious or compromised HTTPS endpoint reached through undici or the global fetch() can crash the client with a single server-initiated renegotiation, taking down all in-flight requests and background work in…
### Impact undici's WebSocket client (including Node.js's bundled globalThis.WebSocket) can be driven to exhaust process memory by a malicious or compromised WebSocket server. For every PING frame received, the receiver automatically writes a PONG response to the socket and ignores the false return value from socket.write(), which is a backpressure signal, so a server that streams small PING frames while never reading the client's PONGs makes the outbound write queue and heap grow without bound until the Node.js process runs out of memory and terminates. The attack is remote and unauthenticated, needs only a normal WebSocket connection, and is asymmetric: small inbound PINGs produce a much larger retained outbound queue. Availability is the only impact. undici 5.x, 6.x, 7.x, and 8.x…
## Impact undici's decompress interceptor forwards decompressed response data to the downstream handler from zlib stream event callbacks without catching exceptions those callbacks may throw. undici's engine converts a synchronous throw from a downstream handler into a per-request failure, but that protection is not in place when the throw happens asynchronously from a zlib callback. As a result, a downstream handler that throws by design, for example the built-in dump interceptor which throws when a response exceeds its maxSize, escapes as an uncaught exception and terminates the Node.js process. A malicious or compromised HTTP server can crash any client that composes the decompress interceptor with dump (or another response handler that throws), by returning a small compresse…
### Summary In affected versions, MCP servers reachable over HTTP put no limit on the size of a POST body or on the number of messages in a JSON-RPC batch. Large requests could exhaust the server's memory, and one long batch could block it from serving anyone else. No credentials were needed, unless your application authenticates requests before they reach the SDK. ### Am I affected? Yes, if clients you do not fully trust can reach your server over Streamable HTTP. A body parser such as express.json() does not protect you, because it does not limit batch length. Servers that only use stdio are not affected. Affected versions: - @modelcontextprotocol/sdk 1.10.0 through 1.30.0 - @modelcontextprotocol/server 2.0.0 - @modelcontextprotocol/node 2.0.0 - @modelcontextprotocol/hono 2.0.…
### Description Yaml\Parser::lexInlineStructure() recursed through lexInlineMapping() and lexInlineSequence() for every { and [ of an inline collection, without entering the nesting level that the rest of the parser bounds with maxNestingLevel. A document made of deeply nested inline collections, such as [[[[...]]]], exhausted the stack while it was being lexed: depending on the PHP version and its stack settings, the process crashed or the request ended with a fatal error. Any code parsing untrusted YAML is affected. That includes #[MapRequestPayload] when the Yaml component is installed and the endpoint does not restrict acceptFormat: a single unauthenticated request with a YAML content type is enough. ### Resolution The inline lexer now enters and leaves the same nest…
### Description HeaderUtils::split() parsed header values with a pattern whose token alternative also matched whitespace, followed by a `(?
### Summary A crafted RSA key inside a JWK can cause operations that parse JWKs to consume excess CPU. ### Details go-jose does not check for bounds on RSA modulus and primes before validating them. Values are passed directly to crypto/rsa.PrivateKey.Validate(). On Go 1.24+, this takes cubic time complexity with respect to the RSA modulus and primes. This allows an attacker to supply a crafted JWK to consume excessive CPU. https://github.com/go-jose/go-jose/blob/5b371aecfcf7aabe06068d67da4344591ecddc76/jwk.go#L655-L700 ### Proof of concept No special configuration of go-jose is required to reproduce this vulnerability. ```go package main import ( "encoding/base64" "encoding/json" "fmt" "math/big" "time" "github.com/go-jose/go-jose/v4" ) func main() { // The RSA parameters can be complet…
### Impact CTFer.io's CTFd Terraform Provider stores ctfd_team.password as non-sensitive values (CWE-522). A local or remote attacker, who can have access to the Terraform state, can read team passwords without the need for the state password for decryption. ### Patches Patched in v2.8.2. ### Workarounds No workaround exist, yet CTFer.io recommends you keep your state in safe environments with a limited set of users who can access it as a good practice to avoid exploitation of this vulnerability.
Comments
Comments
Cybersecurity researchers have disclosed details of three cyber attacks targeting organizations in Japan in which artificial intelligence (AI) was a key element, either in the driving seat or to assist a human operator. TrendAI, which analyzed three exposed server environments used by threat actors to conduct the malicious cyber operations, described them as distinct from previously published
## Summary Hydra's legacy target blacklist omits PyYAML's unsafe loading surfaces. If an application passes attacker-controlled targets and arguments to instantiate(), a target such as yaml.unsafe_load can execute Python code through a config-controlled YAML stream containing Python object-construction tags. ## Impact and prerequisites The attacker must control the relevant target and arguments that the application instantiates. The configured callable executes with the application's process privileges. This is a gap in the best-effort legacy blacklist. It is not a bypass of Hydra 1.4's execution whitelist supplied by trusted Python code. One realistic exposure is a model downloaded from a third-party repository, such as Hugging Face, that bundles attacker-controlled configuration. When a…
## Summary qs.stringify(obj, { arrayFormat: 'comma' }) throws an uncaught TypeError: Cannot convert object to primitive value when an array in obj contains an element that carries an own, non-callable toString property (for example { a: [ { toString: 'x' } ] }). A remote, unauthenticated attacker who can influence an object that the application later re-serializes with the comma array format triggers the throw. In a service that lacks a synchronous error boundary around the serialization call, the uncaught exception crashes the worker process, producing a denial of service. ## Root Cause When arrayFormat: 'comma' is selected, the comma branch collapses the whole array into a single value via Array.prototype.join, or, under encodeValuesOnly: true, maps each element through…
### Impact undici's experimental WebSocket client (enabled by default via allowH2: true) crashes the Node.js process when it connects to a malicious server over HTTP/2. If the server answers the WebSocket upgrade with a response that fails handshake validation, undici aborts the request, but the abort is a no-op because the request is already marked completed, so the underlying ClientHttp2Stream is left open with no error listener. When the server then destroys the connection, the orphaned stream emits an unhandled error and the process exits. A remote, unauthenticated server can therefore deny service to any client that opens a WebSocket to it over HTTP/2. undici 8.11.0 and 8.11.1 are affected; the regression was introduced in 8.11.0. ### Patches Upgrade to 8.11.2. ### Workaround…
Reporter: Cavan Loughran, Celvex Group Inc. Summary Authlib's symmetric (oct) key handling accepts an empty key (empty bytes b'' / empty string '') as a valid HMAC signing and verification key. An empty HMAC key produces a deterministic digest that any party can reproduce, so an attacker who knows (or can cause) the application to verify a token with an empty key can forge arbitrary tokens that pass HS256/HS384/HS512 signature verification. This is the Authlib member of the cross-language empty-HMAC-key family already assigned CVE-2026-44351 (fast-jwt, npm) and CVE-2026-45363 (ruby-jwt, RubyGems). This is DISTINCT from Authlib's existing recent JWT-verification advisories. It is not closed by CVE-2026-27962 / GHSA-wvwj-cvrp-7pv5 (JWS JWK header injection, the key=None case), nor by CVE-20…
In rustls 0.23.44 and 0.23.45, application data that a client buffers before the handshake completes is sent to the server even when the server rejects Encrypted Client Hello (ECH). When the server rejects ECH, the client completes the handshake with the outer ClientHello and authenticates the server against the ECH config's public_name. At the end of the handshake, start_traffic() encrypts and queues any plaintext buffered by ClientConnection::writer(). It does this *before* checking for EchStatus::Rejected and queueing the ech_required alert. The application then receives Error::PeerIncompatible(ServerRejectedEncryptedClientHello), as expected. However, the client is still expected to call write_tls() to send the fatal alert, and that call also sends the encrypted applicat…
## Summary The client records the server-selected key-exchange group per server name when it processes the ServerHello, before the server is authenticated. A network attacker can answer one connection attempt for a target name with a ServerHello that selects X25519 and close the connection. Later ClientHellos for that name carry one X25519 share while supported_groups still lists X25519MLKEM768 first. If the server accepts that share instead of sending a HelloRetryRequest, the handshake completes with classical key exchange only and no error. An attacker that answers one connection attempt for the target name with a ServerHello selecting X25519 changes every later ClientHello of that client for that name, for as long as the entry stays in the store. No certificate or completed handshake i…
### Summary Formatting a client session value with Debug exposes its TLS 1.2 master secret or TLS 1.3 resumption PSK. A custom session store that logs these values can disclose keys to readers of the logs. Ordinary connection and built-in cache diagnostics remain redacted, and the session-store interface already warns that its values are sensitive. ### Details Derived formatters traverse the session's secret-bearing fields. This conflicts with the project's explicit [key-material redaction policy](https://github.com/rustls/rustls/blob/99f2358cae2954837dbb866faf6727de75489ab9/SECURITY.md#boundary-public-api). ### Proof of concept Custom ClientSessionStore implementations could leak secret data by using the provided Tls12ClientSessionValue or Tls13ClientSessionValue in a way that trig…
### Summary The built-in (but disabled by default) Brotli certificate decoder bounds the output buffer but accepts an input-selected history window. A header of five bytes can specify a 1 GiB window allocation. A peer can abort a process whose allocator cannot satisfy that request. A client must enable Brotli and connect to the peer; a server must enable Brotli and request a client certificate. Neither direction requires a trusted attacker certificate. ### Details rustls calls BrotliDecompress which enables Brotli's large-window extension. rustls applies a limit to certificates, but this applies to the decompressed result, and not to any internal allocations the process requires. ### Impact Client applications that enable the brotli feature will be vulnerable to malicious servers. Ser…
### Summary Authlib's RFC 9101 (JAR) extension passes the attacker-controlled request_uri authorization parameter straight to the get_request_object() fetch hook with no validation, at a pre-authentication endpoint. Authlib already defines the pre-registration allowlist for this (request_uris client metadata and require_request_uri_registration server metadata) but the extension never consults it, and both docstrings instruct integrators to implement the hook as an unrestricted requests.get(request_uri). Deployments following the documented pattern get an unauthenticated SSRF primitive. ### Details Verified against authlib 1.7.2 and current main. All line numbers below are from authlib/oauth2/rfc9101/authorization_server.py. The extension registers itself on the `before_get_…
### Summary OpenIDMixin.parse_id_token() reads the nonce_supported claim from the ID token it is validating. When that claim is boolean false, the expected nonce is set to None and nonce validation is skipped for that token. An ID token that carries this claim is therefore accepted in a session whose nonce does not match. A captured, validly signed ID token belonging to a victim can be replayed into an attacker-controlled session, authenticating the attacker as the victim at the relying party. The code is identical in authlib/integrations/base_client/sync_openid.py:82-83 and authlib/integrations/base_client/async_openid.py:84-85, so all framework integrations are affected. Fixed versions: none; reproduced on 1.8.0 (commit c529a61a). Preconditions: the ID token must carry a val…
## Description ### Summary Authlib's default RFC 7523 JWT bearer grant leaves get_audiences() empty, so verify_claims() only checks that an aud claim *exists* instead of requiring it to match the current authorization server's token endpoint. An attacker holding a legitimately issued JWT assertion for *any* audience can present it to a different authorization server sharing the same signing key and obtain an access token. To be completely transparent, AI tools helped surface it. However, we have independently reproduced and carefully validated the issue. We follow responsible disclosure practices and are sharing this privately. ### Affected Entrypoints - authlib.oauth2.rfc7523.JWTBearerGrant — default get_audiences() returns empty list - authlib.jose.jwt.decode() with `claims_…
## Report metadata | Field | Value | | --- | --- | | Package | authlib (PyPI, pkg:pypi/authlib) | | Repository | https://github.com/authlib/authlib | | Component | authlib.oidc.core.grants.implicit.OpenIDImplicitGrant.validate_authorization_request | | Vulnerability class | Authorization bypass: a client's scope restriction is bypassed for OpenID Connect ID token issuance | | Verified against | c529a61af40e3efe05286140e00fbdced07d7987 (v1.8.0). Earlier releases were not checked. | | API surface | Public API (OpenIDImplicitGrant, registered with server.register_grant(...)) | | Affected in default configuration | Yes. No non-default option is needed once OpenIDImplicitGrant is registered. Client metadata decides exposure: Authlib's standard ClientMixin checks response type a…
## Summary OctKey.import_key() carries the only guard that stops an asymmetric public key from being used as an HMAC secret in authlib.jose. That guard is raw_key.startswith(POSSIBLE_UNSAFE_KEYS) — anchored at byte 0 — while the loader that consumes the identical bytes on the asymmetric path (cryptography.hazmat.primitives.serialization.load_pem_public_key / load_pem_x509_certificate) skips anything before the -----BEGIN line. A PEM public key or X.509 certificate carrying **one leading byte** — a newline, a space, four spaces of indentation, a UTF-8 BOM, or a comment line — is therefore not recognised, falls through, and becomes the HMAC secret. An attacker holding only the issuer's public key then mints an HS256 token with arbitrary claims that jwt.decode() accepts as au…
## Summary PhpSpreadsheet's Worksheet\Drawing::setPath() validates an external-image URL against an operator-supplied isWhitelisted callback **once**, on the initial URL only, then fetches it with file_get_contents() using a stream context that does **not** set follow_location => 0. A whitelisted host that returns an HTTP 302 Location: redirect to an internal target is therefore followed by PHP's HTTP wrapper (default follow_location=1, max_redirects=20) with **no re-validation of the redirect hop** — defeating the whitelist and yielding a blind SSRF from the server. This is an unfixed variant of CVE-2026-59931 (GHSA-6hq5-7373-42rg, the WEBSERVICE redirect-whitelist-bypass): the follow_location => 0 fix (commit 7ef7b25e) was applied ONLY to Calculation/Web/Service.php;…
## Summary When PhpSpreadsheet loads an OpenDocument Spreadsheet (.ods), Ods::processTableRow() reads the table:number-rows-repeated attribute of a `` with an unchecked (int) cast, then materialises one cell per repeated row. This value is **completely unbounded**, so a **single ~1.4 KB file** can instruct the reader to create up to **millions of cells**, exhausting memory (fatal OOM). No authentication, no special configuration, and the default read filter are required - any application that loads a user-supplied .ods is affected. --- ## Affected versions The number-rows-repeated expansion has existed in the ODS reader since its inception (pre-dating the current release branches), so all maintained lines are affected. | Branch | Affected | |--------|----------| | 1.x | ≤ 1.30…
### Summary PhpSpreadsheet's XmlScanner::scan() uses a regex pattern that only accounts for NULL byte (0x00) interleaving when detecting `open('poc.xlsx', ZipArchive::CREATE | ZipArchive::OVERWRITE); $xlsx->addFromString('[Content_Types].xml', '' .'' .'' .'' .'' .'' .''); $xlsx->addFromString('_rels/.rels', '' .'' .''); $xlsx->addFromString('xl/workbook.xml', '' .''); $xlsx->addFromString('xl/_rels/workbook.xml.rels', '' .'' .'' .''); $xlsx->addFromString('xl/worksheets/sheet1.xml', '' .'0'); $xlsx->addFromString('xl/sharedStrings.xml', $shared_strings); $xlsx->close(); echo "XLSX size: " . filesize('poc.xlsx') . " bytes\n"; // Load with PhpSpreadsheet $mem_before = memory_get_usage(true); $reader = new \PhpOffice\PhpSpreadsheet\Reader\Xlsx(); $reader->setReadDataOnly(true); $spreadshee…
### Summary The SpreadsheetML XML reader (Reader\Xml) scans file content **twice** through XmlScanner::scan(), and each scan runs an unentity() callback that calls html_entity_decode() **after** the pattern check. By double-encoding the `' . "\n" . '' . "\n" . '' . "\n" . '' . "\n" . ' ' . "\n" . ' ' . $refs . '' . "\n" . ' ' . "\n" . ''; file_put_contents('poc.xml', $xml); echo "File size: " . filesize('poc.xml') . " bytes\n"; $mem_before = memory_get_usage(true); $reader = new \PhpOffice\PhpSpreadsheet\Reader\Xml(); echo "canRead: " . ($reader->canRead('poc.xml') ? 'true' : 'false') . "\n"; $spreadsheet = $reader->load('poc.xml'); $peak = memory_get_peak_usage(true); echo "Peak memory: "…
### Summary A crafted ~27 KB two-block zstd frame makes DecodeAll and the streaming decoder panic with an out-of-range slice index instead of returning a decode error. The same bytes decode fine with the reference zstd. Any program that decompresses untrusted zstd input can be crashed by it. Reproduced on v1.20.0 and v1.20.1. Released in https://github.com/klauspost/compress/releases/tag/v1.20.2 ### Details The panic is in the FSE decoder that huff0.ReadTable uses to read a block's Huffman-weight table. Three conditions have to hold at once, and the crafted frame arranges all three. **1. A count that doesn't fit is accepted.** readNCount permits a tableLog up to 15 ([tablelogAbsoluteMax](https://github.com/klauspost/compress/blob/v1.20.0/fse/decompress.go#L9), checked at [`fse/decom…
### Summary authlib.common.security.is_secure_transport() is used throughout Authlib to enforce the OAuth 2.0 / OIDC requirement that endpoint and redirect URIs use TLS, with an RFC 8252 §7.3 exemption for loopback http:// URIs. The check is implemented as a textual str.startswith prefix match, so a URL such as http://localhost:[email protected]/cb — whose actual host is evil.example and whose scheme is plain http — is accepted as "secure". An attacker who can supply a URL to any of the sinks that rely on this check (OIDC RP-Initiated Logout dynamic client registration, RFC 8414 authorization-server metadata validation, error-redirect rendering) can smuggle a cleartext non-loopback URL past the TLS-required guard. This is a defense-in-depth / spec-conformance failure; severity i…
### Impact A maliciously crafted DNS message may cause dns.name.from_wire_parser() to use excessive CPU for DNS name decompression. Applications that process untrusted DNS wire form messages with dnspython may be degraded or DoS'd by the extra CPU use required. ### Patches This problem is fixed in dnspython 2.9.0. ### Workarounds There are no known workarounds. ### Credit Evgenios Gkritsis, Department of Informatics, University of Piraeus & Athena Research Center, Greece Constantinos Patsakis, Department of Informatics, University of Piraeus & Athena Research Center, Greece George Stergiopoulos, Department of Informatics, Athens University of Economics and Business, Greece
### Summary A regular expression used to obfuscate URL query strings can consume excessive CPU when processing certain attacker-controlled input. This can stall request processing and cause denial of service in applications using affected HTTP tracing integrations. ### Details The default query-string obfuscation pattern can exhibit excessive backtracking. The pattern is applied synchronously when recording HTTP URL metadata, allowing query-string processing to occupy application resources for an extended period. As a workaround, set both DD_TRACE_OBFUSCATION_QUERY_STRING_REGEXP and DD_TRACE_OBFUSCATION_QUERY_STRING_PATTERN to the fixed Python-compatible regex below, then restart the application. Setting both variables covers all affected ddtrace versions, including releases before 1.19.0…
### Summary A flaw in dd-trace-py’s IAST instrumentation can cause the Python application process to terminate unexpectedly when processing certain attacker-controlled input. Only applications with IAST enabled are affected. IAST is disabled by default. ### Details The affected IAST string-processing instrumentation does not properly handle certain internal errors, allowing an error to terminate the application process. Exploitation requires IAST to be enabled and application code to perform an affected string-processing operation on attacker-controlled input. If you are unable to upgrade to a patched version of dd-trace-py, you can mitigate this issue by either disabling IAST (setting DD_IAST_ENABLED=false) or by replacing replace % string interpolations with str.format or f-strings for…
### Impact The C extension's BSON decoder scans a Regex element's pattern and flags strings with strlen before checking the document's remaining length, so a crafted truncated Regex element can read past the end of the caller's buffer. bson.decode and bson.decode_all accept mmap and array buffers, which carry no NUL sentinel; the read can cross into unmapped memory and kill the process with an uncatchable SIGSEGV. Any application decoding untrusted BSON from such buffers is affected — impact is availability only, since the out-of-bounds bytes are never returned as content. NUL-terminated buffers (bytes, bytearray) raise InvalidBSON at worst; the driver's own transport and pure-Python installs are not affected. All releases through 4.18.2 are affected when the C extension is loaded. ###…
## Summary An untrusted multimodal request can set mm_processor_kwargs.code_revision. When vLLM serves a model whose Hugging Face processor uses dynamic auto_map code and the operator enabled --trust-remote-code, this request value reaches AutoProcessor.from_pretrained and selects the Python revision imported by Transformers. This lets a client replace the operator-reviewed processor module with code from another revision of the same model repository. The operator-selected model/data revision remains pinned; only the executed processor code changes. Importing a Python module is arbitrary code execution in the vLLM API/renderer process. The PoC below demonstrates this with an import-time filesystem write. It is fully offline and does not require a GPU, model weights, or changes t…
## Summary The FileSessionManager defaults to storing session data in /tmp/strands/sessions/ (a world-writable directory) using predictable file paths and a write pattern (open(path + ".tmp", "w") → os.replace()) that follows symbolic links. A local attacker can exploit this to overwrite arbitrary files or inject malicious session data for persistent prompt injection. ## Vulnerability Details **Component:** strands-py/src/strands/session/file_session_manager.py **Root Cause:** 1. Default storage directory is tempfile.gettempdir() + "/strands/sessions/" (line 58) — resolves to /tmp/strands/sessions/ on Linux/macOS 2. The _write_file() method (lines 116-123) writes to a predictable .tmp path without symlink protection: ```python tmp = f"{path}.tmp" with open(tmp, "w", enco…
There is a local confidentiality issue in the default FileSessionManager storage behavior. By default it writes session data under /tmp/strands/sessions and relies on the process umask for directory and file permissions. On systems using the common umask 022, the session directories are 0755 and JSON files are 0644, so other local users can read persisted agent session data. This matters on shared developer machines, CI hosts, notebooks, jump boxes, and multi-user GPU servers. Session storage is intended to persist conversation/history/state, so it can contain prompts, model output, tool inputs/outputs, intermediate state, and secrets that pass through the agent. Relevant code: - strands-py/src/strands/session/file_session_manager.py:58: default storage path is `os.path.join(t…
### Impact Snappy.uncompressCharArray, uncompressShortArray, uncompressIntArray, uncompressLongArray, uncompressFloatArray and uncompressDoubleArray sized the result array by dividing the declared uncompressed byte length by the element size, but native code wrote the full byte length. If the declared length is not a multiple of the element size, up to 7 bytes are written past the end of the Java array, corrupting the heap. ### Patches Fixed in 1.1.10.9 ([#739](https://github.com/xerial/snappy-java/pull/739)). These methods now reject uncompressed lengths that are not a multiple of the element size with SnappyIOException. ### Workarounds Check that Snappy.uncompressedLength(input) is a multiple of the element size before calling these methods.
### Impact Snappy.arrayCopy(Object src, int offset, int byteLength, Object dest, int destOffset) called a native memcpy without validating the ranges. A negative or oversized byteLength or offset read or wrote outside the arrays and crashed the JVM. SnappyOutputStream.rawWrite and SnappyInputStream.rawRead, and therefore the typed-array write/read methods, pass caller-supplied ranges to this method. ### Patches Fixed in 1.1.10.9 (#741). Snappy.arrayCopy checks both ranges before copying. The stream read/write methods validate offsets and lengths, and reject typed-array ranges whose byte size overflows int. ### Workarounds Validate offsets and lengths before calling these methods.
### Impact Several decompression paths in snappy-java allocate an output buffer sized by the uncompressed length declared in the first bytes of the compressed data, before decompressing or validating it. A few bytes of crafted input can therefore force an allocation of up to 2 GB, causing OutOfMemoryError (denial of service), or a NegativeArraySizeException for declared lengths of 2^31 or more. Affected entry points include Snappy.uncompress(byte[]), Snappy.uncompressString, the Snappy.uncompress*Array methods, SnappyInputStream, and SnappyFramedInputStream. The maxChunkSize limit added for CVE-2023-43642 bounds only the compressed chunk length, so it does not prevent this. ### Patches Fixed in 1.1.10.9 (#739). Snappy.uncompressedLength() now rejects negative declared le…
**Description** def _escape(v: str) -> str: return str(v).replace("'", "\\'").replace("\\", "\\\\") The escaping order is backwards, exactly like the classic HTML double-escape bug: step 1 inserts a literal backslash in front of every ', then step 2 re-matches that just-inserted backslash and doubles it. For an input ending in a single quote, the two steps combined produce \\' (two backslashes followed by the quote) in the output — which, under the standard backslash-escape convention used by Cypher/openCypher single-quoted string literals, is parsed as one escaped backslash followed by an unescaped, string-terminating quote. The quote is not neutralized; the string breaks open. This escaped value is interpolated directly into Cypher query strings built by string formatting (not parameter…
## Summary An authenticated LiteLLM proxy user could read arbitrary local files from the proxy runtime by sending a crafted vertex_ai_credentials value to /utils/transform_request. The request-body safety blocklist rejected vertex_credentials but missed its alias vertex_ai_credentials, which let the value reach the Vertex AI credential loader and trigger a file read that is sent to an attacker-controlled URL. ## Impact A low-privilege authenticated user (only internal_user_viewer, obtainable through SSO login on many deployments) could read any file readable by the LiteLLM process. Reading files such as /proc/self/environ can disclose LITELLM_MASTER_KEY and provider API keys, so the impact can escalate to full proxy-admin control. ## Remediation Upgrade to 1.95.0 or later.…
## Am I affected? You are affected if you use @better-auth/drizzle-adapter from 1.6.17 through versions before 1.7.7 (including the 1.7.0 prereleases from beta.6 through rc.6) or @better-auth/kysely-adapter 1.6.17 through 1.6.31, and keep rate-limit records in a PostgreSQL database. All of these must be true: - Rate limiting is enabled. - Rate-limit records are stored in the database. - The database is PostgreSQL, used through one of the adapter versions above. You are not affected if you store rate-limit records in memory (the default) or in secondary storage. ## Summary Database-backed rate limiting on PostgreSQL can let a burst of concurrent requests go over the configured maximum. When many requests arrive at the same time, several of them can pass the limit check together. This w…
## Am I affected? You may be affected if you use better-auth from 1.5.0-beta.12 through versions before 1.7.7 with the OAuth Proxy plugin. You are affected only if **all** of these are also true: - On a participating environment, the OAuth Proxy plugin has no dedicated secret, or its secret equals that environment's global Better Auth secret. - Sign-in state is stored in a cookie (account.storeStateStrategy: "cookie"). With a database or secondary storage, this is only the case if you chose it explicitly. - Implicit account linking is on (the default). You are not affected by this path if every participating environment uses a proxy secret distinct from its global Better Auth secret, if sign-in state is stored in the database, or if you use a version before 1.5.0-beta.12. ## Sum…
## Am I affected? You may be affected if you use better-auth from 1.4.0-beta.18 through versions before 1.7.7 with the Magic Link plugin and social or Generic OAuth sign-in. You are affected if **all** of these are true: - Magic Link is enabled with default storeToken: "plain", or with storeToken: "hashed" and global verification.storeIdentifier: "hashed" (including a hashed default). - At least one built-in social provider or Generic OAuth provider is enabled. - OAuth state uses account.storeStateStrategy: "database", the default when a database or secondary storage is configured. Cookie-backed state blocks this path, subject to the OAuth Proxy condition below. With custom hashers, confirm the flows cannot resolve to the same stored identifier. ## Summary Better Auth accept…
### Summary The /sso/debug/callback endpoint reflected identity-provider-controlled fields into an inline `` block via json.dumps, which escapes JSON string syntax but not HTML context characters such as ``. An IdP-returned string field containing `` (for example a user-editable display name) closed the script block and executed attacker JavaScript in the proxy's origin. ### Impact Reflected XSS on the LiteLLM proxy origin. The route required no authentication, so an attacker with an account on the configured IdP whose claims pass such strings through verbatim could complete the SSO flow themselves and deliver the resulting callback URL to a logged-in proxy admin. Scripts executing in the proxy origin can read the dashboard session cookie and act with the victim's privileges, up to fu…
> **A note on the referenced files.** This report was submitted through GitHub's private > vulnerability reporting API, which carries no attachments. Every transcript this report leans > on is reproduced **inline above**, verbatim. The file names it mentions (poc.sh, poc.log, > per-case transcripts, evidence.md, negative-controls.md, blob-sweep tables) are the local > evidence bundle those inline excerpts were extracted from, with a checksum manifest. I am happy > to send the bundle by any channel you prefer, or to open a private fork — just say which. The > PoC scripts are self-contained and reproduce the result from a clean environment. --- **Affected:** ` "The audit also found that **reset's positional commit could carry the option before its argument > separator**. […] **Check…
## Summary GitPython's public Submodule.move() API verifies that its destination is lexically inside the parent repository, but it does not reject symbolic links in the destination's intermediate components before filesystem operations. On systems that check out Git symlinks, an untrusted repository can therefore redirect a fixed, application-selected move destination outside the parent worktree. The reporter reproduced this on the latest release, GitPython 3.1.62, and current main. In the strict scenario, the repository commits vendor -> ../outside, while the consuming application independently passes the fixed literal vendor/module. Submodule.move() follows vendor, moves the initialized submodule checkout to outside/module, rewrites .gitmodules and the index to the symli…
## Summary Remote.pull() in GitPython forwards its positional refspec argument directly to the git pull child process while its unsafe-option guard inspects only keyword arguments. An application that passes attacker-influenced data as the refspec (for example a user-supplied branch or refspec parameter in a web or CI service) can supply an option-shaped value such as --upload-pack=. The value bypasses the unsafe_git_pull_options check, is placed after the -- separator on the git pull command line, and is executed as the upload-pack program for local-path or SSH transports. This yields arbitrary command execution with the privileges of the calling process using the documented default arguments (allow_unsafe_options=False). ## Root Cause Remote.pull() is documented to r…
### Summary GitPython has spent 2022-2026 rolling out two reusable guards method-by-method: Git.check_unsafe_protocols(url) (raises UnsafeProtocolError for :: remote-helper URLs - CVE-2022-24439) and Git.check_unsafe_options(...) (raises UnsafeOptionError for denylisted option spellings - CVE-2023-40267, CVE-2026-42215, CVE-2026-67323). Three wrapper methods that build a git argv from caller-controlled values still have **neither**: 1. **Git.ls_remote** (git/cmd.py:1128-1143) - checks options (:1139-1141) but forwards the URL positional unchecked; no allow_unsafe_protocols parameter exists. The same URL through clone_from raises UnsafeProtocolError. 2. **Repo.merge_base** (git/repo/base.py:895-936) - forwards **kwargs/*rev to self.git.merge_base with no guard;…
### Summary Repo._clone()'s handling of separate_git_dir never disables Git.polish_url()'s environment-variable expansion, unlike every other caller of that helper that handles attacker-influenced data. Since separate_git_dir can be derived from an attacker-controlled .gitmodules submodule name, a completely ordinary clone + submodule_update(init=True, recursive=True) on a malicious repository expands an environment variable of the attacker's choosing and exposes its value both as a persistent on-disk directory name and, whenever DEBUG-level logging is enabled (a common CI/verbose-logging configuration), in a Popen(...) log line that is not redacted the way credential-bearing arguments are. ### Root cause / Details git/repo/base.py's _clone() has: ```python sep_dir = k…
### Summary This is a residual of GHSA-g5vv-9gxw-82hx / CVE-2026-87819. The 3.1.60 fix made Actor._from_string regex-free, but parse_actor_and_date still runs _re_actor_epoch on the same author/committer/tagger line one step earlier. That pattern backtracks quadratically when the trailing epoch or timezone is missing or malformed, so the same class of slow parse is still reachable in 3.1.62. ### Details git/objects/util.py:322: ``python _re_actor_epoch = re.compile(r"^.+? (.*) (\d+) ([+-]\d+).*$") `` .+? and (.*) can both absorb the same run of spaces, and the required (\d+) ([+-]\d+) tail is only checked after them. On a long line with no valid tail, the search retries every split point. It is called from parse_actor_and_date (git/objects/util.py:335), which runs on…
## Summary Nodemailer's SMTP client applies four regular expressions to a peer's EHLO response to detect the supported SASL mechanisms. Each pattern uses an alternation with two overlapping whitespace-matching sub-patterns, which produces quadratic (O(n^2)) backtracking whenever the mechanism name is absent from the line. A malicious or MITM SMTP server that answers EHLO with 250-AUTH followed by a long whitespace-only run and a non-matching token blocks the Node.js event loop for tens of seconds to hours during a pre-authentication handshake. No authentication to the target SMTP server is required, and any application that connects to an attacker-influenced host is affected. ## Root Cause Each of the four AUTH mechanism regexes in SMTPConnection._actionEHLO contains the sub-pattern `…
## Summary Nodemailer's address parser has three published, patched fixes for RFC 5322 comment ( ... ) handling inside an addr-spec: GHSA-cc9r-2j5m-2m83 (comment glued two domain atoms into one), GHSA-g57g-f23g-4646 (the same glue for a quoted local part), and the related ReDoS fixes GHSA-prgh-xp8r-p3m5 / GHSA-v53p-9fqp-m79j. All three fixes live in _handleAddress()'s joins / _recoverAddrSpec() logic in src/addressparser/index.ts, and that logic is exercised whenever the local text-accumulation path in the tokenizer runs. That logic is **never reached at all** for the single most common way a recipient is written: a display name followed by an angle-bracket address, Name . When the input inside the angle brackets itself contains an RFC 5322 comment, e.g. `` Legit User `` the…
## Summary The SMTP client response parser reassembles multiline server replies by appending each continuation line onto a single response-queue entry and re-testing whether that entry is still a partial (continuation) reply on every line. The partial-response test runs str.slice(str.lastIndexOf('\n') + 1) on the growing entry, which forces a full flatten of the entire accumulated string on each line. A malicious or compromised SMTP server (or an on-path attacker on a plaintext or pre-STARTTLS hop) that streams continuation lines and never sends a completion line drives O(n^2) CPU on the single-threaded Node.js event loop before any authentication, freezing the whole process for an attacker-controlled duration. ## Root Cause The invariant that a server cannot cause superlinear client-si…
### Summary An authenticated user with internal_user privileges can escalate to proxy_admin and achieve remote code execution. The proxy uses one encryption key for two purposes: sealing secrets at rest and minting session tokens. An attacker exploits this by requesting a new API key with a crafted metadata field containing a forged admin credential as the "secret" value. The proxy encrypts and returns this payload. When this value is then presented as a bearer token, the proxy decrypts it, trusts the forged admin identity, and grants full administrative access, including the ability to execute arbitrary commands via the MCP stdio endpoint. ### Affected versions Versions later than 1.91.0 are exploitable in the default configuration. Versions 1.87.0 through 1.90.x are exploitable only…
### Impact The msgpack-python C extension (msgpack._cmsgpack) leaks a decoded map value when deserialization rejects a key with strict_map_key=True (the default). Repeated malformed inputs can cause unbounded process memory growth and denial of service in applications that deserialize attacker-controlled MessagePack data. Both unpackb() and the streaming Unpacker API are affected. The pure-Python fallback does not exhibit this leak in the reported tests. The report measured approximately 4 MiB retained per failed parse and 100 MiB after 25 failed parses on Windows with Python 3.14 and msgpack 1.2.2. These are reporter-provided results. ### Additional impact: stale parser-stack references The reference-ownership audit performed while fixing this leak found an additional memory-safe…
## Summary WekaClassifier in NLTK passes a caller-controlled model file path to the weka Java VM without any nltk.pathsec validation. WekaClassifier.train puts the caller's model_filename verbatim into the weka argv ["-d", model_filename, ...] (weka *writes* the trained model there), and WekaClassifier._classify_many puts self._model into ["-l", self._model, ...] (weka *reads* it). Under nltk.pathsec.ENFORCE=True (the default on current HEAD), NLTK's stated invariant — established by advisory CVE-2026-81726 / GHSA-8mgp-746c-j5xp — is that caller-controlled model-artifact paths must be bounded to the allowed NLTK data roots before any native tool opens them. WekaClassifier is the one java()-based tool wrapper that skips this entirely, giving an untrusted workflow arbi…
Authlib (prior to 1.9.0) contains a signature verification bypass vulnerability. The JsonWebSignature.deserialize_json() method accepts a JSON Serialization JWS object and returns the payload as successfully verified without checking for a signature and without requiring a cryptographic key. ## Root cause authlib/jose/rfc7515/jws.py, read on the installed 1.8.0 artifact: ``python is_valid = True # set before anything is checked for header_obj in obj["signatures"]: # empty array means the body never runs jws_header, valid = self._validate_json_jws(...) if not valid: is_valid = False ... if is_valid: return rv # returns the payload as verified `` is_valid is initialised to True and only ever set to False from inside the loop. A JWS JSON Serialization object whose signatures arra…
## Am I affected? You may be affected if you use better-auth from 1.3.8-beta.3 up to, but not including, 1.7.0-rc.3, and you enable the deviceAuthorization() plugin. This range includes all 1.6 releases. You are affected only if all of these are also true: - An attacker can start a device request: validateClient is not configured, or the attacker can use a valid public client. - Before allowing approval, your approval page does not show, from server-checked data, which client is asking and the exact permissions it requests. - Users may approve a code or complete verification link they did not start and cannot compare with a device they have. - Your application accepts the session token that the device flow returns, for example through the bearer plugin. You are not affected if y…
### Impact Since 2006 / pylint 0.15.2, pylint caches its stats as pickles in PYLINT_HOME and reload them on the next run with pickle.load. Someone without access to the configuration or linted code, but with access to the cache directory (predictable PYLINT_HOME on a multi-user host) can write a crafted pickle that will runs arbitrary code when pylint access its stat cache. ### Patches From pylint 4.0.9, the result cache is now stored as JSON instead of pickle, preventing code-execution. ### Workarounds Until upgrading do not point PYLINT_HOME to an untrusted, shared, or group-writable directory. The default value, ~/.cache/pylint, is writable only by the user running pylint. ### Credits - Reported privately via Tidelift coordinated disclosure (Dan Lenz).
## Impact Satori does not properly escape certain values before including them in generated SVG output. This can allow crafted values to be interpreted as SVG markup. The impact depends on how the generated SVG is consumed. ## Workaround No complete workaround exists besides upgrading. Applications that cannot immediately upgrade should not render attacker-controlled content with Satori.
### Summary When the OffloadingConnector uses TieringOffloadingSpec with a P2P secondary tier, an API client can submit many unique non-listening peer addresses through kv_transfer_params.prefill. vLLM keeps every never-established P2P session and its ZeroMQ sockets after the request finishes. On vLLM 0.26.0, 340 such peers filled the default ZeroMQ context socket quota; creating the next peer raised an uncaught ZMQError, terminated EngineCore, and shut down the API server. The path was reproduced with a normal API key and no access to the internal P2P listener. ### Details #### Root cause In vLLM 0.26.0, P2PSecondaryTierManager.on_new_request() reads the following request-controlled fields: ```json { "kv_transfer_params": { "prefill": { "kv_request_id": "id", "remote_host": "…
The public /api/remove endpoint uses attacker-controlled extras.sam_model in both SAM filenames and GitHub download URLs. Path traversal plus a URL fragment lets an unauthenticated attacker place malicious ONNX models beside a target file. The models read the file as ONNX external data and encode its bytes into the returned PNG. This issue affects SAM's sam_model in 2.0.81 and returns complete file contents. *Screenshot 1 — First unauthenticated request primes the path and returns HTTP 500.* ### Technical details rembg/commands/s_command.py:188-223,384-397 accepts public extras JSON and passes it to new_session(). rembg/sessions/sam.py:313-355 inserts sam_model into encoder/decoder filenames and fixed GitHub URLs, then invokes pooch.retrieve() without a pinned hash. Requ…
### Summary twisted.names.dns.DNSDatagramProtocol, the UDP transport used by twisted's DNS client (twisted.names.client.Resolver, createResolver, and anything built on it, including MX lookups in twisted.mail), accepts a DNS response for a pending query solely on the basis of the 16-bit transaction ID matching. It never checks that the response actually originated from the server the query was sent to. Any host that can deliver a UDP packet to the resolver's listening port with the correct transaction ID can inject a forged answer, independent of the real server's address. This is a materially weaker defense than standard DNS resolver practice (transaction ID plus source-address validation, typically also source-port rotation), and enables off-path DNS response spoofing / cache po…
### Impact An input validation issue allowed members to delete comment likes / dislikes that they were not authorized to delete. ### Vulnerable versions This vulnerability is present in Ghost from v5.9.0 up to v6.44.1. ### Patches v6.44.1 contains a fix for this issue. ### How to update For self-hosters using Docker, find [Docker's official Ghost image here](https://hub.docker.com/_/ghost). Updating a Docker-based Ghost instance [is documented here](https://docs.ghost.org/install/docker#updating-ghost). If your Ghost is a Ghost-CLI install see our documentation on [updating it to the latest version here](https://docs.ghost.org/update). ### References Ghost thanks [Adel Assakaf](https://github.com/assakafpix) and [Arthur Gurunlian](https://github.com/Buggs777) as well as Geoffrey Diederich…
## Summary kernel/mcp/tools/file.go's resolvePath() includes an explicit comment stating its sensitive-path blocklist is meant to align with the HTTP file API's established blacklist (refuseToAccess() in kernel/api/file.go), following the remediation for GHSA-cvhv-7xhj-xjp8 ("Unauthenticated Administrator takeover exfiltrated via the MCP"), which specifically recommended restricting this tool's scope. In practice, the MCP tool implements only 1 of the 4 static blocklist entries the HTTP API enforces. ## Comparison kernel/api/file.go refuseToAccess() blocks: 1. conf/conf.json (API token, cookie signing key, accessAuthCode) 2. data/snippets/conf.json 3. the entire data/templates directory 4. data/.siyuan/publishAccess.json 5. (dynamic, per-request) any path the current p…
### Summary fpdf2 expands SVG `` references by deep-copying the complete referenced graphics tree. A small SVG can use nested references to make the number of in-memory objects grow exponentially, causing excessive CPU and memory consumption during FPDF.image() processing. This was reproduced with an 895-byte SVG on fpdf2 2.8.7. Under a 128 MB process limit, the input kept the process busy for about 14 seconds before raising MemoryError. ### Details The issue is in fpdf2's own SVG reference handling. It does not rely on XML entities, external URLs, local files, or vulnerable behavior in another dependency. When fpdf2 resolves an SVG reference, [SVGObject.build_xref()](https://github.com/py-pdf/fpdf2/blob/a3307f96a0d165831029b3fb644aee5a7ca03328/fpdf/svg.py#L1584) passes the referenc…
## Summary fpdf2 implicitly dereferences image resource references supplied to rendering APIs. If an application passes untrusted HTML, SVG, template data, or image paths into those APIs, fpdf2 may issue outbound HTTP(S) requests or attempt to open local filesystem paths. This behavior is relevant in applications that render untrusted or user-provided input without sanitizing or restricting embedded resource references. ## Description In affected versions, fpdf2 uses a shared internal image-loading mechanism for multiple features. This loader: * fetches http:// and https:// URLs over the network * opens other string values as local filesystem paths During document processing, resource references embedded in input content are automatically dereferenced using this mechanism. This…
# CompactHashMap/CompactHashSet/MapMakerInternalMap native deserialization triggers uncontrolled memory allocation ## Summary CompactHashMap, CompactHashSet, and MapMakerInternalMap implement readObject(ObjectInputStream) by reading an attacker-controlled elementCount or initialCapacity from the stream. On the very first insertion during or after deserialization (for the CompactHash* collections, inside allocArrays()) or immediately during deserialization (for MapMakerInternalMap), the implementation eagerly allocates arrays sized to that full, unvalidated value (up to CompactHashing.MAX_SIZE = 2^30-1, ~1.07 billion, for example) before the rest of the stream's actual content has been consumed. A 72-byte malicious serialized stream containing a single tiny element can…
The GoAhead embedded web server loaded a certificate revocation list into its default TLS backend but never turned revocation checking on, so a revoked client certificate could still pass login on affected setups. The flaw is in builds that use the OpenSSL backend with mutual-TLS, where the server checks a client certificate and vice versa, and a revocation list is configured to block revoked
Apple on Monday backported fixes for a high-severity security flaw that was exploited in real-world cyber attacks to older devices. The vulnerability, tracked as CVE-2026-86950 (CVSS score: 8.8), is an out-of-bounds write affecting CoreGraphics that could result in arbitrary code execution when processing a maliciously crafted file. Patches for the flaw were released last month by Apple in
### Impact Agno versions prior to 2.3.24 contain an arbitrary code execution vulnerability in the model execution component. The field_type parameter on a FunctionCall input field was passed to Python's built-in eval(), allowing an attacker who could influence that value to execute arbitrary Python code in the Agno process. This is a code injection flaw (CWE-95) with full impact to confidentiality, integrity, and availability. CVSS 3.1 score: 9.3 (Critical). Any deployment that exposes Agno to user-influenced FunctionCall inputs (for example via an API, web UI, or untrusted tool/model output) is affected. ### Patches Fixed in **Agno 2.3.24**. The unsafe eval() call was replaced with a fixed type_mapping dictionary lookup, and field_type is now validated as either a known strin…
### Summary A call to the token endpoint with a grant type of "urn:ietf:params:oauth:grant-type:device_code" and a bogus device_code is able to get a bearer token as long as they know a valid client ID enabled for the device workflow. ### Details [Section 3.4 of the RFC 8628 spec](https://datatracker.ietf.org/doc/html/rfc8628#section-3.4) says that the token endpoint should take in device_code as a required parameter.DeviceCodeGrant.validate_token_request does not perform any validation of device_code. DeviceCodeGrant.validate_token_request should be validating that device_code was passed in as a parameter and that it is valid, similar to how AuthorizationCodeGrant.validate_token_request does the checks [here](https://github.com/oauthlib/oauthlib/blob/a015b446622f2b9255175…
### Summary An unauthenticated remote attacker can crash a running MONAILabel server or silently corrupt its medical AI inference results by uploading a single crafted DICOM file (~132 KB) via the REST API. The vulnerability stems from pydicom's lack of validation for Decimal String (DS) and Integer String (IS) DICOM Value Representations, which allows NaN and infinity values to pass through MONAI's PydicomReader and propagate through the entire spatial transform and ML inference pipeline. No authentication is required for the default MONAILabel REST API. The attacker needs only network access to the server (default port 8000). **Severity:** CVSS 3.1: 8.7 HIGH **CVSS Vector:** CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:N/I:H/A:L **CWE:** CWE-20 (Improper Input Validation) → CWE-682 (Incorrect Cal…
dcmjs, a JavaScript library for DICOM manipulation used in web-based medical imaging viewers, contains an input validation bypass in its handling of DS (Decimal String) and IS (Integer String) value representations. The DecimalString and IntegerString classes in ValueRepresentation.js use a regex sanitizer followed by JavaScript's Number() constructor. While the regex strips direct NaN/Infinity strings, it preserves scientific notation (including the 'e' character), enabling: 1. Infinity injection via scientific notation overflow (1e999999) 2. NaN injection via sign-only remnants from stripped inf strings ("-inf" -> "-") 3. IS accepting scientific notation in violation of PS3.5 AFFECTED PRODUCT ================ Product: dcmjs (JavaScript DICOM manipulation library) Repository: https://git…
# Security Vulnerability Report: Wails v2 — Origin Validator Bypass via Wildcard Pattern | | | |---|---| | **Product** | wailsapp/wails (v2) | | **Affected File** | v2/internal/frontend/originvalidator/originValidator.go | | **Affected Versions** | All versions containing originValidator (latest as of 2026-03-26) | | **CVSSv4.0** | **7.6 HIGH** — CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:P/VC:H/VI:H/VA:L/SC:L/SI:L/SA:N | | **CWE** | CWE-625: Permissive Regular Expression | | **Platforms** | Windows / Linux / macOS (all desktop targets) | --- ## Summary Wails v2 provides BindingsAllowedOrigins to restrict which web origins can invoke Go backend methods via IPC. The wildcard pattern matching in wildcardPatternToRegex converts * to .* without anchoring subcomponents, allowing suffix-matche…
### Summary A SQL Injection vulnerability exists in the Third Party (Company) update functionality of Dolibarr ERP/CRM. The localtax1_value parameter (lt1) is concatenated directly into an SQL UPDATE statement in societe/class/societe.class.php at line 1672 without any escaping, type casting, or parameterized query binding. An authenticated user with standard "Third Party: Create/Edit" permission can extract any data from the database, including password hashes, API keys, and all business data. Tested on Dolibarr v23.0.0 (latest stable), default installation via official Docker image, MariaDB 10.11. No custom modules, no modified configuration. ### Details **Vulnerable code** (societe/class/societe.class.php, line 1672): ```php $sql .= ",localtax1_value =".$this->localtax1_value;…
### Impact An unauthenticated user could gain access to a server to execute arbitrary code on this server. ### Patches This is fixed in the version 2.2.13. Users should upgrade to this version as soon as possible.
We are launching a PHP Capture the Flag challenge integrated with our Bug Bounty program to test our WAF protections. Security researchers are invited to bypass our defenses and capture the flag to help make the Internet safer.
An examination of global Internet disruptions observed by Cloudflare Radar during the third quarter of 2026. This review analyzes how power failures, natural disasters, and government shutdowns impacted network connectivity and traffic.
Microsoft has fixed a flaw in its Authenticator app for Android that let a malicious app with no permissions pose as a trusted Microsoft app, such as Teams, and collect sign-in tokens for a victim's work account. Microsoft tracks the flaw as CVE-2026-80097 and fixed it in Authenticator for Android version 6.2606.4246, so anyone using the app should update. In the researchers' tests,
An attacker who gained access to Keystone's website obtained the key that its hardware wallets use to prove they are genuine, the company said in a statement on October 26. Someone with that key could build counterfeit devices that pass the online check owners use to verify a Keystone. A buyer who trusted a counterfeit could have their crypto stolen, but the incident did not involve
Fake login pages are nothing new, but generative AI is making them easier to build and more convincing. Attackers can recreate familiar login screens, impersonate trusted brands and tailor phishing attempts to specific employees or organizations. Some fake login pages go beyond stealing passwords. They can also capture MFA codes or hijack authenticated sessions, allowing attackers to gain
The North Korea-linked Lazarus Group has been attributed to a cyber espionage campaign dubbed Operation FlashHole targeting at least three organizations in South Korea spanning the construction, software, and wire-communication equipment manufacturing sectors. According to Russian cybersecurity company Kaspersky, the campaign has been active since at least July 2025. It's assessed to pick up
A criminal using the Amadey malware can no longer access or steal data from 180,000 infected computers around the world, including some in Australia, the Australian Signals Directorate (ASD) says. The agency's operators gained access to the criminal's operation's backend and cut it off from those computers. "The next steps that we took were to disconnect that Amadey controller from all those
OpenAI's compaction feature, which shrinks a long conversation to save space, could be turned into a way to copy a model's hidden reasoning out in plain text, security firm Xint has shown. OpenAI's reasoning models work through a problem before answering, and the service normally returns only the raw reasoning as an encrypted item the user cannot read. The method got past
The Computer Emergency Response Team of Ukraine (CERT-UA) has issued an alert warning of unauthorized remote access to video surveillance systems manufactured by Dahua. "Cases of compromising IP cameras and DVRs with access to the video stream, changing the device configuration and creating third-party accounts have been recorded," the State Special Communications Service of Ukraine (SSSCIP)
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Monday confirmed active exploitation of the recently patched PHP2Shell flaws by adding them to its Known Exploited Vulnerabilities (KEV) catalog. The PHP vulnerabilities, both rated 9.2 on the CVSS scoring system, facilitate unauthenticated remote code execution on affected deployments - CVE-2026-109618 - An out-of-memory
Adobe Commerce and Magento contain an unrestricted upload of file with dangerous type vulnerability that could result in arbitrary code execution in the context of the current user, potentially gaining elevated access or control over the victim's account or session.
Required action: Apply mitigations in accordance with vendor instructions, ensuring compliance with CISA’s BOD 26-04 Prioritizing Security Updates Based on Risk (see URL in Notes) guidance and CISA’s “Forensics Triage Requirements” (see URL in Notes). Follow applicable BOD 26-04 guidance for cloud services or discontinue use of the product if mitigations are unavailable. Stakeholders are responsible for evaluating each asset's internet exposure and ensuring adherence to BOD 26-04 patching guidelines.
CISA due date: 2026-11-10
Google Brotli contains an unchecked return value vulnerability that allows a remote unauthenticated attacker to corrupt adjacent decoder state memory, disclose sensitive memory contents, cause a denial of service, or potentially execute arbitrary code using a crafted Brotli-compressed stream.
Required action: Apply mitigations in accordance with vendor instructions, ensuring compliance with CISA’s BOD 26-04 Prioritizing Security Updates Based on Risk (see URL in Notes) guidance and CISA’s “Forensics Triage Requirements” (see URL in Notes). Follow applicable BOD 26-04 guidance for cloud services or discontinue use of the product if mitigations are unavailable. Stakeholders are responsible for evaluating each asset's internet exposure and ensuring adherence to BOD 26-04 patching guidelines.
CISA due date: 2026-11-10
Comments
A configuration mismatch in GLM 5.3 on Workers AI triggered an SGLang bug that exposed KV cache pages across requests. Here is how we addressed the issue.
Comments
An unknown threat actor has infected 953 online stores, most of them running Magento, as part of a long-running campaign active since July 2025. According to Dutch e-commerce security company Sansec, the activity is designed to add a small piece of code to the store's pages that, when visited by unsuspecting users, serves malicious JavaScript loader code through Magento CMS blocks. "When you
PHP has rebuilt its official Windows packages for PHP 8.2 through 8.5 to fix CVE-2026-111497, a flaw in the Brotli compression library that the packages include. The flaw can let an attacker corrupt memory and potentially run code. PHP says the flaw can be reached through the cURL extension when the extension decompresses HTTP responses automatically. cURL leaves that off by default, but
Cybersecurity researchers are calling attention to what they say is a "recurring social engineering pattern" across hundreds of cases in which threat actors use Microsoft Teams' external chat creation feature to conduct impersonation scams or to deliver malware. "A consistent pattern of attacker behavior has emerged across these incidents: the use of the same throwaway tenants, the same 'IT
This2Shell, That2Shell, Everything2Shell. That's pretty much how the last week went. Critical bugs that let attackers run code remotely are showing up everywhere. Hackers are scanning for targets, some flaws are already being used in attacks, and security teams have more urgent patches than they can keep up with. And that's not even the full story. A leaked key led to a software supply-chain
Two critical flaws in Wasmtime, a WebAssembly runtime that can be built into other software, could allow a malicious WebAssembly module to escape the sandbox intended to contain it, the project's maintainers say. An attacker would first need a vulnerable Wasmtime build to compile a module they control, and one of the flaws affects only ARM64 systems. Fixed versions were released on
Every SOC struggles with a pile of alerts. Very few of those alerts explain how an attack started or how it progressed. Network data, like that from a network detection and response (NDR) platform, connects individual alerts into one timeline, and it feeds the defensive systems that can tie all events together into one coordinated view. SOC teams often first see the shape of an intrusion
A rentable Android spyware kit called uCreator takes over Telegram and WeChat accounts by hiding its code inside working copies of the two apps, researcher Ovi found. It only reaches people who install one of these tampered copies, spread outside the official app stores, or who run WeChat inside the kit makers' own container app. The kit's Telegram part matches Baohuo, a backdoor that Doctor
Cybersecurity researchers have disclosed a Microsoft 365 campaign that has been observed leveraging an adversary-in-the-middle (AitM) relay to capture sign-in sessions and then replay the stolen tokens to break into email accounts at 79 organizations. Huntress, which observed the activity between July 14 and September 22, 2026, said it detected 497 successful Microsoft 365 sign-in events across
An automated campaign is using stolen GitHub tokens to plant workflows that print a repository's secrets into its GitHub Actions logs, which the attackers then download before deleting the runs. The attacks have succeeded in private repositories and GitHub Enterprise environments, according to an analysis by cloud security company Wiz. A GitHub safeguard that pauses
A malicious instruction file sitting in a developer's Downloads folder tried to make their AI coding agent install a known-malicious npm package, security firm Forever Security said in a report published last week. The company caught the attempt on a real customer's machine and stopped the agent before any code ran. It could not determine how the file got into the folder. The file
PHP contains an out-of-memory corruption vulnerability that allows for unauthenticated remote code execution.
Required action: Apply mitigations in accordance with vendor instructions, ensuring compliance with CISA’s BOD 26-04 Prioritizing Security Updates Based on Risk (see URL in Notes) guidance and CISA’s “Forensics Triage Requirements” (see URL in Notes). Follow applicable BOD 26-04 guidance for cloud services or discontinue use of the product if mitigations are unavailable. Stakeholders are responsible for evaluating each asset's internet exposure and ensuring adherence to BOD 26-04 patching guidelines.
CISA due date: 2026-10-29
PHP contains a stale internal request-state vulnerability that allows for unauthenticated remote code execution.
Required action: Apply mitigations in accordance with vendor instructions, ensuring compliance with CISA’s BOD 26-04 Prioritizing Security Updates Based on Risk (see URL in Notes) guidance and CISA’s “Forensics Triage Requirements” (see URL in Notes). Follow applicable BOD 26-04 guidance for cloud services or discontinue use of the product if mitigations are unavailable. Stakeholders are responsible for evaluating each asset's internet exposure and ensuring adherence to BOD 26-04 patching guidelines.
CISA due date: 2026-10-29
A single hacked device on a Tailscale network can be used to gain root access to other servers it can reach, according to a security researcher who published the findings this week. The attack requires two things: Tailscale SSH set up to allow port forwarding, and an attacker who already controls one device that the network's access rules permit to connect to the target. The findings were
In an update, Palo Alto Networks Unit 42 said it's retracting its earlier assessment that the post-compromise activity targeting the European government agency involved the exploitation of the PHP2Shell flaws. "What researchers saw was an attack against a European government agency executed in the context of a vulnerable PHP application," a spokesperson for the company told The Hacker News via
OpenAI has announced plans to conduct frontier model training with no live internet access, as it resumes reinforcement learning (RL) training and internal inference of its most advanced AI models. "Based on the current state of alignment research, monitoring systems, and the security of internet infrastructure, we believe the responsible approach is to conduct frontier model training with no
Comments
A memory-corruption flaw in Google's Brotli decoder sits in many programs that unpack Brotli-compressed data, among them Bun, Microsoft's .NET, the Python Brotli and brotlicffi packages, and Envoy proxies set up to decompress Brotli. Tracked as CVE-2026-111497, the flaw allows crafted compressed data to overwrite past the end of a buffer and corrupt a program's memory,
LiteLLM on Friday revealed that an Agent Platform master key accidentally published to a public GitHub repository in late May 2026 ultimately allowed the threat actor to push a malicious version of the popular Python package. "On May 27, 2026, a configuration file was committed to the public LiteLLM-Labs/lite-harness repository which inadvertently contained the master key to a deployment of
Google has updated its privacy policy to spell out that using personalized artificial intelligence (AI) features in Search or Gemini could allow the company to serve targeted ads to users based on their activity on other connected Google services. "You may choose to allow certain Google services, including services like Search or Gemini Apps, to access data from other Google services you use,
Between November 2025 and March 2026, the Dutch security firm Eye Security pushed more than 2 million updates to computers it did not own to switch off a background updater that it said could have allowed an outsider to take over the machine. The updater belongs to Time4Popcorn, an abandoned copy of Popcorn Time, a free app that streamed pirated films and TV over BitTorrent. Popcorn Time was
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Friday added two vulnerabilities impacting Collabora Online Built-in CODE Server and Arista VeloCloud Orchestrator (VCO) to its Known Exploited Vulnerabilities (KEV), citing reports of active exploitation in the wild. The critical-severity vulnerabilities are listed below - CVE-2026-110447 (CVSS score: 9.2) - An operating
ReadWriteLock and AsyncReadWriteLock (the SQLite-backed reader/writer lock) remove their private .filelock-* link directories with shutil.rmtree (a recursive delete) on a remembered path. They do not check that the path still names the directory they created. This happens on the hard-link route (a hard link is a second name for the same file). That route is every connection on macOS since 4.1.1 and any POSIX process without a usable /proc/self/fd (a Linux path that names the process's own open files), such as a chroot or container that does not mount /proc. Another local user who can rename entries in the directory that holds the lock database can rename one of the directories of the user running the lock (below: the lock user) onto the link directory's name, or onto a forged…
On Windows, ReadWriteLock and AsyncReadWriteLock (the SQLite-backed reader/writer lock) connect SQLite to the lock database by its path, in every directory. SQLite derives three side files from that path: -journal, the rollback journal (pages saved to undo an interrupted write); -wal, the write-ahead log (changes kept outside the main file); and -shm, the shared-memory index of the WAL. SQLite acts on whichever of these it finds beside the database before the lock runs any statement. Another local account that can create files in the database's directory, but not delete files of other accounts there, places such a side file beside the database and denies everyone DELETE on it, which a file's creator may do. C:\Windows\Temp (the temporary directory of services, granting `BUIL…
On Windows the soft locks (SoftFileLock, MarkerSoftFileLock, SoftFileLease, their async forms, and a FileLock that fell back to a soft lock) read their marker by path. The marker is the lock file whose presence says a process holds the lock. Each acquire() attempt that finds a marker opens it with os.open to read the holder's process id, and the lock renames and deletes a marker it judges stale, by path. On Windows each of these calls waits for an oplock. An oplock (opportunistic lock) is a request by a process that holds a handle on a file to be told before another process opens, renames or deletes that file; the other process waits until the holder answers the break or closes its handle. Another local account that can open the marker for reading can take a Level 1 oplock on…
## Summary The app-data seeder can leave an environment linked to temporary package files that virtualenv deletes when creation finishes. The environment-variable and configuration-file forms of symlink_app_data can enable this state even though the command-line flag is disabled for temporary app-data. The resulting environment depends on a cache pathname that its owner no longer controls. Anthropic's OSS Scanner reported this as ANT-2026-ZOE2GEPC in virtualenv-20261015-005, received on 15 October 2026. This draft records source review and functional remediation checks. The reporter's cross-account code-execution claim has not been independently reproduced. ## Affected area Integrity of app-data references after temporary cache cleanup. ## Attacker-controlled input Symlink seeding m…
## Summary The pip seeder acquires bundled wheels with download=False and lets pip access the index when --download is selected. Index candidates on that path do not pass through virtualenv's wheel digest verification. Anthropic's OSS Scanner reported this as ANT-2026-74ALJBUF in virtualenv-20261015-005, received on 15 October 2026. ## Affected area Wheel selection by the pip seeder when downloads are enabled. ## Attacker-controlled input The pip seeder and download option must be selected. The finding concerns the default-index digest-verification policy. Explicit custom-index settings already opt out of comparison against PyPI metadata; this patch retains that policy. ## Affected code The relevant 21.14.8 source is in [pip seeder](https://github.com/pypa/virtualenv/blob/a795ab116356…
## Summary Wheel downloads go directly into the reusable wheelhouse, while the verification target is selected from pip's textual output. A verification failure does not remove the downloaded file from that cache. Anthropic's OSS Scanner reported this as ANT-2026-ST00DRO3 in virtualenv-20261015-005, received on 15 October 2026. ## Affected area Download selection and promotion into the reusable wheel cache. ## Attacker-controlled input A download must fail verification, or the archive selected from pip output must differ from the file later reused. Default-index digest verification must be enabled. Existing custom-index exceptions are outside the promised PyPI comparison. ## Affected code The relevant 21.14.8 source is in [wheel download and digest check](https://github.com/pypa/virtu…
## Summary Caller-supplied prompt text can become executable syntax in the xonsh activation script. An attacker must influence the value passed to virtualenv, and a user must subsequently activate the environment. ## Affected area Xonsh reads a UTF-8 activation script using a legacy multibyte encoding such as GBK. Printable non-ASCII bytes can change how the reader recognizes an adjacent quote. The report concerns POSIX activation. It does not establish exposure through unrelated activators. ## Attacker-controlled input The attacker controls prompt or destination data passed as a virtualenv argument. The user later sources the generated activation script. ## Affected code The relevant 21.14.8 source is in [xonsh quote()](https://github.com/pypa/virtualenv/blob/a795ab116356afde11039cf2763d…
## Summary The app-data copy seeder uses shutil.copy, which preserves cached file modes independently of the current umask. A cache populated under permissive permissions can therefore produce files with wider access in later environments. Anthropic's OSS Scanner reported this as ANT-2026-487MH6OW in virtualenv-20261015-005, received on 15 October 2026. ## Affected area Permissions of files copied from app-data into a new environment. ## Attacker-controlled input The cache must contain files created with broader permissions than the current environment creator intends. Another account must also be able to traverse the new environment and reach those files. A private parent directory prevents that access. No control of a private live cache is assumed. ## Affected code The relevant 21.1…
## Summary Caller-supplied prompt text can become executable syntax in the bash activation prompt. An attacker must influence the value passed to virtualenv, and a user must subsequently activate the environment and draw the prompt. ## Affected area Bash runs with compatibility level 4.2 and draws the environment prompt. Its replacement-string behavior differs from newer compatibility modes, leaving prompt text subject to a second interpretation. The report concerns POSIX activation. It does not establish exposure through unrelated activators. ## Attacker-controlled input The attacker controls prompt or destination data passed as a virtualenv argument. The user later sources the generated activation script. ## Affected code The relevant 21.14.8 source is in [bash prompt assembly](https://…
## Summary The venv creator reads the standard library's generated pyvenv.cfg with splitlines() and merges all resulting keys over virtualenv's configuration. Line-boundary characters from destination data can therefore become configuration structure before output sanitization. Anthropic's OSS Scanner reported this as ANT-2026-6ZQHJ6VI in virtualenv-20261015-005, received on 15 October 2026. ## Affected area Configuration readback by the stdlib venv creator. ## Attacker-controlled input An untrusted caller must influence the destination pathname passed to the venv creator. That pathname must contain a character that Python treats as a line boundary. The builtin creator follows a different configuration-writing path. ## Affected code The relevant 21.14.8 source is in [venv readback](ht…
## Summary The app-data seeder joins paths from existing distribution metadata to site-packages and deletes the resulting paths without checking containment within the environment. Anthropic's OSS Scanner reported this as ANT-2026-SXOEHWP7 in virtualenv-20261015-005, received on 15 October 2026. ## Affected area Removal of a previously installed seed package. ## Attacker-controlled input An environment being reused must contain distribution metadata influenced by another account. The invoking user must have deletion rights over the paths named by that metadata. Installing a new environment without existing package metadata does not enter this cleanup path. ## Affected code The relevant 21.14.8 source is in [seed uninstall](https://github.com/pypa/virtualenv/blob/a795ab116356afde11039c…
### Impact Origin validation error in Microsoft Authentication Library (MSAL) for Node.js allows an unauthorized attacker to disclose information over a network. This affects the auth_code flow. ### Patches Version 7.1.0 and higher are patched. ### Workarounds Ensure you pass in only the auth_code to to acquireTokenByCode
### Impact undici's fetch() crashes the Node.js process when a malicious server returns a response with more than five content codings and closes the connection before sending the body. undici rejects such responses (a content-encoding limit), but it created the internal response stream before that check and returned without attaching an error listener or aborting the request. The early close then surfaces a content-length mismatch on that orphaned internal stream, which Node.js treats as an unhandled error and terminates the process. A remote, unauthenticated server can trigger this with a single response, even when the application correctly catches the fetch() rejection. undici 6.x, 7.x, and 8.x are affected. ### Patches Upgrade to 6.29.1, 7.30.1, or 8.12.2. undici now valid…