Vulnerability database
Every flaw the U.S. government’s cybersecurity agency (CISA) has confirmed attackers are using, plus the new WordPress plugin and theme bugs published each day, sorted by what they mean for a small business. We refresh it twice a day. If you see something your website or office runs, tell us.
1,839 results · page 1 of 37
Oct 30, 2026LightSAML implements SAML 2.0 data model classes, XML serialization and deserialization, XML security, certificate support, and message bindings.CVE-2026-63182 · Other software · litesaml · reported by GitHubHigh 8.1
LightSAML implements SAML 2.0 data model classes, XML serialization and deserialization, XML security, certificate support, and message bindings. Prior to 4.7.1 and 5.0.1, LightSAML and xmlseclibs disagree about which assertion an inbound SAML signature covers. `Response::deserialize` uses `manyElementsFromXml` to collect only direct-child `saml:Assertion` elements of `samlp:Response`, and `AssertionSignatureValidatorAction` validates their `ds:Signature` using an out-of-band trusted IdP credential resolved by Issuer, not a key from the document. `XMLSecurityDSig::validateReference()` removes the enveloped signature from the shared DOM, but `processRefNode()` resolves `Reference URI="#id"` document-wide through an XPath of the form `//*[@Id]` and takes `item(0)`, which need not be the assertion whose claims LightSAML consumes. An attacker who has captured one genuinely signed assertion, such as an earlier login response, can place its signature-stripped bytes earlier in `samlp:Extensions` and present a forged direct-child assertion with the same ID and copied genuine signature. Validation then covers the hidden genuine bytes while LightSAML surfaces unsigned attacker-controlled `Su
- Product
- lightsaml
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 30, 2026PDF Reader MCP is a Model Context Protocol server for reading PDF documents.CVE-2026-62264 · Other software · SylphxAI · reported by GitHubHigh 7.5
PDF Reader MCP is a Model Context Protocol server for reading PDF documents. In versions 2.2.0 and later prior to 3.0.1, when `MCP_TRANSPORT=http` enables the opt-in HTTP server, `@sylphx/pdf-reader-mcp` reads `MCP_API_KEY` and logs that `X-API-Key` authentication is enabled but never validates the `X-API-Key` header on incoming requests. The `/mcp` endpoint consequently accepts unauthenticated clients. With the default bind address of `0.0.0.0`, any client able to reach the port can invoke every PDF tool and read any PDF the server process can open on the host, despite the misleading authentication-enabled startup log. URL-based sources additionally permit limited server-side request forgery; private and loopback destinations are blocked by default unless `--allow-private-ips` is set. The default `stdio` transport is unaffected. This issue is fixed in version 3.0.1.
- Product
- pdf-reader-mcp
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 30, 2026LM Studio before 0.4.22+1 contains a path traversal vulnerability in the model downloader's resolveDownloadRequest that allows attackers to write files outside CVE-2026-119979 · Other software · Element Labs · reported by VulnCheckHigh 8.7
LM Studio before 0.4.22+1 contains a path traversal vulnerability in the model downloader's resolveDownloadRequest that allows attackers to write files outside the models directory using backslash-separated traversal filenames. Attackers can trick users into downloading a crafted model repository that plants a plugin manifest.json, install-state.json and mcp-bridge-config.json, running an attacker-supplied Python payload on every launch.
- Product
- LM Studio
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordlmstudio.aivulncheck.com
Oct 29, 2026modelscope agent-zero 1.7, 1.8, 1.9, and 1.10 is vulnerable to Directory Traversal in python/helpers/file_browser.py:FileBrowser.save_text_file.CVE-2026-51851 · Other software · reported by MITRE
modelscope agent-zero 1.7, 1.8, 1.9, and 1.10 is vulnerable to Directory Traversal in python/helpers/file_browser.py:FileBrowser.save_text_file. The save_text_file method accepts user-controlled file paths without normalization or validation, allowing path traversal attacks.
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordgist.github.comgithub.com
Oct 29, 2026Wasmtime 46.0.0 before 48.0.6 and 49.0.0 before 49.0.3 panics when the host reads a component model stream as a different Rust item type than it was created witCVE-2026-119135 · Other software · bytecodealliance · reported by VulnCheckMedium 5.9
Wasmtime 46.0.0 before 48.0.6 and 49.0.0 before 49.0.3 panics when the host reads a component model stream as a different Rust item type than it was created with, because the host-to-host read path asserts matching TypeIds instead of returning an error. Guest code can return a host-created StreamReader<U> that the embedder reads as StreamReader<T>, where both types map to the same component model type, crashing the embedding process.
- Product
- wasmtime
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordgithub.comvulncheck.com
Oct 29, 2026Wasmtime before 36.0.18, 37.0.0 before 48.0.6, and 49.0.0 before 49.0.3 caps component-model resource handle tables only per component instance and does not chaCVE-2026-119134 · Other software · bytecodealliance · reported by VulnCheckMedium 6.9
Wasmtime before 36.0.18, 37.0.0 before 48.0.6, and 49.0.0 before 49.0.3 caps component-model resource handle tables only per component instance and does not charge handle allocations to any ResourceLimiter or StoreLimits. A malicious guest can repeatedly call resource.new and instantiate sub-components to exceed the per-instance cap many times over, causing unbounded host memory growth.
- Product
- wasmtime
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordgithub.comvulncheck.com
Oct 29, 2026Crawl4AI from 0.9.0 before 0.9.5 fails to filter the CosineStrategy model_name field in untrusted Docker API request bodies, allowing authenticated clients to cCVE-2026-119120 · Other software · unclecode · reported by VulnCheckHigh 7.7
Crawl4AI from 0.9.0 before 0.9.5 fails to filter the CosineStrategy model_name field in untrusted Docker API request bodies, allowing authenticated clients to choose which HuggingFace model the server loads. Attackers can reference a malicious repository with a pickle weight file to execute Python code under PyTorch below 2.6, or exhaust disk and memory.
- Product
- crawl4ai
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordgithub.comvulncheck.com
Oct 29, 2026The WIFI4 N300 router (Model: T1 Pro) v1.0.0 by U-speed contains an insecure access control flaw that allows an attacker to gain unauthenticated root shell.CVE-2026-79589 · Other software · reported by MITRE
The WIFI4 N300 router (Model: T1 Pro) v1.0.0 by U-speed contains an insecure access control flaw that allows an attacker to gain unauthenticated root shell. The vulnerability is caused by the exposure of the UART debugging interface having a lack of authentication, authorization and access control functionality. Once the interface is soldered and connected to the attacker, they gain root privilege to the system.
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordgithub.comu-speed.com.cn
Oct 29, 2026Agno through 3.1.11 contains an unsafe reflection vulnerability in PandasTools that allows attackers to invoke arbitrary pandas functions by supplying model-conCVE-2026-119050 · Other software · agno-agi · reported by VulnCheckCritical 9.2
Agno through 3.1.11 contains an unsafe reflection vulnerability in PandasTools that allows attackers to invoke arbitrary pandas functions by supplying model-controlled function names to getattr(). Attackers can use prompt injection to make the agent call pandas.read_pickle on attacker-controlled paths for code execution, or call other readers to read local files.
- Product
- agno
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 28, 2026Ultralytics parse_model Code Injection Remote Code Execution Vulnerability.CVE-2026-118618 · Other software · Ultralytics · reported by zdi-disclosures@trendmicro.comHigh 7.8
Ultralytics parse_model Code Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Ultralytics. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The specific flaw exists within the parse_model function. The issue results from the lack of proper validation of a user-supplied string before using it to execute Python code. An attacker can leverage this vulnerability to execute code in the context of the current user. Was ZDI-CAN-30121.
- Product
- Ultralytics
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 28, 2026Ultralytics parse_model Code Injection Remote Code Execution Vulnerability.CVE-2026-118616 · Other software · Ultralytics · reported by zdi-disclosures@trendmicro.comHigh 7.8
Ultralytics parse_model Code Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Ultralytics. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The specific flaw exists within the parse_model function. The issue results from the lack of proper validation of a user-supplied string before using it to execute Python code. An attacker can leverage this vulnerability to execute code in the context of the current user. Was ZDI-CAN-30123.
- Product
- Ultralytics
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 27, 2026MLflow is an open source AI engineering platform for agents, large language models, and machine learning models.CVE-2026-116692 · Other software · mlflow · reported by GitHubCritical 9.8
MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.12.0rc0 until 3.17.0, third-party scorer deserialization in mlflow/genai/scorers/base.py accepts request-selected descendants of THIRD_PARTY_SCORER_ALLOWED_MODULES rather than requiring an exact allowed module. A caller can use MLflow artifact endpoints to place a Python module beneath a writable allowed package and submit its module and class through /ajax-api/3.0/mlflow/scorer/invoke. Scorer.model_validate_json() imports the module and constructs the class before trace processing. Authentication-disabled deployments permit this code execution without credentials, while basic-auth deployments require a principal capable of creating an experiment and run and uploading artifacts. In either configuration, the allowed scorer package must be writable by the server process. This issue is fixed in version 3.17.0.
- Product
- mlflow
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 27, 2026MLflow is an open source AI engineering platform for agents, large language models, and machine learning models.CVE-2026-116691 · Other software · mlflow · reported by GitHubMedium 6.5
MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 2.19.0 until 3.17.0, mlflow.crewai.autolog() records CrewAI Agent objects without consistently sanitizing nested credential-bearing values. mlflow/crewai/autolog.py::_get_agent_attributes() converts the Agent llm field to a string that includes the provider API key, while mlflow/crewai/autolog.py::_construct_full_inputs() expands the Agent object into mlflow.spanInputs. Applications can therefore send plaintext provider credentials to the MLflow trace backend or a configured OTLP destination, where trace readers can obtain them. This issue is fixed in version 3.17.0.
- Product
- mlflow
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 27, 2026MLflow is an open source AI engineering platform for agents, large language models, and machine learning models.CVE-2026-116690 · Other software · mlflow · reported by GitHubHigh 7.5
MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.10.0 until 3.17.0, the prompt branch in _create_model_version() accepts schemeless relative sources because path validation runs only when a source has a URL scheme, and mlflow.prompt.is_prompt is interpreted by tag presence rather than a true value. A remote caller can store a server-local path as the model version artifact location and use /api/2.0/mlflow/model-versions/get-artifact to read files accessible to the MLflow server process. Default unauthenticated servers require no privileges, while basic-auth deployments are exposed to authenticated users who can create registered models. This issue is fixed in version 3.17.0.
- Product
- mlflow
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 27, 2026The Model Context Protocol Python SDK is the official Python SDK for Model Context Protocol servers and clients.CVE-2026-116686 · Other software · modelcontextprotocol · reported by GitHubMedium 6.5
The Model Context Protocol Python SDK is the official Python SDK for Model Context Protocol servers and clients. From 1.10.0 until 1.30.0 and 2.2.0, ClientSession, ClientSessionGroup, and Client tool calls validate server-declared outputSchema documents with jsonschema's default resolver, causing an unresolved $ref value to fetch an HTTP, HTTPS, or file URL from the client's network and filesystem context. A server can trigger blind requests or local-file reads and can stall every session in the process because resolution runs synchronously on the event loop without a timeout. This issue is fixed in versions 1.30.0 and 2.2.0.
- Product
- python-sdk
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 27, 2026The Model Context Protocol Python SDK is the official Python SDK for Model Context Protocol servers and clients.CVE-2026-116685 · Other software · modelcontextprotocol · reported by GitHubMedium 5.9
The Model Context Protocol Python SDK is the official Python SDK for Model Context Protocol servers and clients. From 1.8.0 until 1.30.0 and 2.2.0, streamable_http_client, streamablehttp_client, sse_client, ClientSessionGroup, OAuth, and create_mcp_http_client client paths enable follow_redirects and follow redirects to a different origin, forwarding caller-supplied custom credential headers and, for 307 or 308 responses, the request body. An attacker who can influence a trusted endpoint's redirect destination can receive that material. This issue is fixed in versions 1.30.0 and 2.2.0.
- Product
- python-sdk
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 27, 2026The Model Context Protocol Python SDK is the official Python SDK for Model Context Protocol servers and clients.CVE-2026-116684 · Other software · modelcontextprotocol · reported by GitHubMedium 6.8
The Model Context Protocol Python SDK is the official Python SDK for Model Context Protocol servers and clients. From 1.10.0 until 1.30.0 and 2.2.0, server-side bearer authentication configured through AuthSettings with token_verifier accepts a valid token issued for another service when the authorization server serves multiple resources and the TokenVerifier does not independently validate the aud audience. This resource-confusion flaw lets a token holder authenticate to the wrong MCP server and exercise that token's resulting access. This issue is fixed in versions 1.30.0 and 2.2.0.
- Product
- python-sdk
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 27, 2026The 3D Viewer for Nextcloud is a comprehensive 3D model viewer application for Nextcloud that supports multiple 3D file formats.CVE-2026-104189 · Other software · maz1987in · reported by GitHubHigh 7.5
The 3D Viewer for Nextcloud is a comprehensive 3D model viewer application for Nextcloud that supports multiple 3D file formats. In versions 1.7.9 through 3.3.1, the app's public file endpoint performed no share authorization. Anyone who knows a share token could read the shared file's contents even when the link share was protected by a password. `PublicFileController` extended `OCP\AppFramework\Controller` and marked its routes `#[PublicPage]`. Nextcloud only enforces share passwords through `PublicShareMiddleware`, which returns early for any controller that is not an `OCP\AppFramework\PublicShareController` instance — so no password, expiry, or share-type check ran for these routes. `ShareFileService::loadLinkShare()` compounded this by treating any token that `getShareByToken()` matched as authorised. A share token is the part of the URL that people paste into chats, emails, and tickets. Password protection exists precisely for the case where holding the token should not by itself grant access, and that guarantee did not hold. Requests to `GET /ocs/v2.php/apps/threedviewer/public/file/{token}/{fileId}` returned `200` with the full file body for password-protected shares, with
- Product
- 3Dviewer-Nextcloud
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 27, 2026Use-after-free in the DOM: CSS Object Model component.CVE-2026-117941 · Computers, phones & browsers · Mozilla · reported by security@mozilla.org
Use-after-free in the DOM: CSS Object Model component. This vulnerability was fixed in Firefox 159, Firefox ESR 115.44, Firefox ESR 153.6, Thunderbird 159, and Thunderbird 153.6.
- Product
- Firefox
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordbugzilla.mozilla.orgmozilla.org
Oct 26, 2026llama.cpp before b5649 contains a heap out-of-bounds read vulnerability in llama_vocab::impl::print_info() that allows attackers to crash model loading via overCVE-2025-71439 · Other software · ggml-org · reported by VulnCheckMedium 6.8
llama.cpp before b5649 contains a heap out-of-bounds read vulnerability in llama_vocab::impl::print_info() that allows attackers to crash model loading via oversized special token IDs. Attackers can supply a crafted GGUF model with a short tokenizer.ggml.tokens array, causing llama-cli or llama-server to read past id_to_token and crash.
- Product
- llama.cpp
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026llama.cpp before b2731 contains an out-of-bounds read vulnerability in llama_model_loader that allows attackers to read memory outside the mapped model file viaCVE-2024-58399 · Other software · ggml-org · reported by VulnCheckMedium 6.9
llama.cpp before b2731 contains an out-of-bounds read vulnerability in llama_model_loader that allows attackers to read memory outside the mapped model file via crafted GGUF tensor offsets. Attackers can supply a crafted or truncated GGUF model file with arbitrary 64-bit tensor offsets to crash the process or load adjacent process memory as model weights.
- Product
- llama.cpp
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026vLLM is an inference and serving engine for large language models.CVE-2026-115918 · Other software · vllm-project · reported by GitHubMedium 6.5
vLLM is an inference and serving engine for large language models. From 0.16.0 until 0.32.0, the /v1/responses endpoint allows an authenticated caller to place an attacker-controlled template in chat_template_kwargs[chat_template] when no server template is configured. ChatParams.get_apply_chat_template_kwargs() preserves that value without enforcing trust_request_chat_template, so the request can bypass the intended trust policy and cause CPU or memory exhaustion through Jinja rendering. This issue is fixed in version 0.32.0.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026vLLM is an inference and serving engine for large language models.CVE-2026-115917 · Other software · vllm-project · reported by GitHubLow 3.4
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the OpenAI-compatible server configures Starlette CORSMiddleware with allowed_origins set to a wildcard by default. A malicious web page can use a victim's browser to reach an unauthenticated vLLM server on localhost, an internal network, or a VPN, call endpoints including GET /v1/models and POST /v1/chat/completions, consume GPU resources, and read responses. Cookie-based cross-site request forgery does not apply because credentials are disabled by default, and deployments requiring bearer authentication are exposed only when the caller already has a usable token. This issue is fixed in version 0.32.0.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026vLLM is an inference and serving engine for large language models.CVE-2026-115916 · Other software · vllm-project · reported by GitHubMedium 6.5
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, vllm/tokenizers/hf.py computes max_token_id with an off-by-one upper bound, allowing a pre-tokenized prompt ID equal to the vocabulary size to pass validation. For a tensor-parallel size of one and an exact multiple-of-64 vocabulary, VocabParallelEmbedding.forward does not mask that value before F.embedding uses it, causing a CUDA device-side assertion and terminating the engine. This issue is fixed in version 0.32.0.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026vLLM is an inference and serving engine for large language models.CVE-2026-115915 · Other software · vllm-project · reported by GitHubHigh 8.1
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, ImageMediaIO passes attacker-controlled image bytes to Pillow through Image.open(BytesIO(data)) without restricting decoder selection, and the MIME label does not constrain the selected format. Image-capable request paths through parse_image() and fetch_image() can therefore dispatch an EPS payload to host components. A host with Ghostscript affected by CVE-2024-29510 can execute commands with the vLLM process's privileges, while environments using Pillow 12.0.0 through 12.2.0 can trigger CVE-2026-59203 and exhaust media-loading workers. Deployments without the applicable vulnerable dependency are not affected by that corresponding impact. This issue is fixed in version 0.32.0.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026vLLM is an inference and serving engine for large language models.CVE-2026-115914 · Other software · vllm-project · reported by GitHubMedium 6.5
vLLM is an inference and serving engine for large language models. From 0.11.1 until 0.32.0, the DeepSeek-OCR, DeepSeek-OCR-2, and Unlimited-OCR image path allows a remote client that can reach the OpenAI-compatible API to submit a legal image whose normalized pixels sum to zero when that image is alone in its per-step encoder batch. _parse_and_validate_image_input returns no embedding, embed_multimodal propagates that value, and sanity_check_mm_encoder_outputs triggers a fatal assertion that shuts down the engine for all users. This issue is fixed in version 0.32.0.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026vLLM is an inference and serving engine for large language models.CVE-2026-115913 · Other software · vllm-project · reported by GitHubMedium 6.5
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mm_audio.py and phi4mm_utils.py uses eager attention by default, allowing a remote client to submit one legal long audio clip that causes MultiHeadedAttention.forward_attention to materialize and retain full attention matrices in self.attn. The scaled dot-product attention branch does not materialize these matrices. The resulting quadratic GPU-memory use can cause an out-of-memory condition and terminate EngineCore for all users. This issue is fixed in version 0.32.0.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026vLLM is an inference and serving engine for large language models.CVE-2026-115912 · Other software · vllm-project · reported by GitHubHigh 7.5
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, an authenticated caller can place forged shared-memory address and monotonic_id values in public multimodal features.kwargs_data. In scale-out deployments with shared-memory multimodal caching enabled, workers trust those values and pass attacker-selected bytes to pickle.loads(), allowing arbitrary Python execution even when insecure serialization is disabled. This issue is fixed in version 0.32.0.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 26, 2026vLLM is an inference and serving engine for large language models.CVE-2026-115911 · Other software · vllm-project · reported by GitHubMedium 6.5
vLLM is an inference and serving engine for large language models. From 0.19.0 until 0.32.0, authenticated callers can supply a request chat template when no server-side --chat-template is configured, or a malicious model repository can provide a template that is compiled and rendered by tokenizer.apply_chat_template() without a wall-clock timeout, output-size limit, or iteration bound. Rendering runs in a ThreadPoolExecutor with one worker by default, so a nonterminating render can block /v1/responses and other chat-template users after the request ends, while an output-expanding render can exhaust server memory. Harmony models are not affected because they do not use this Jinja path. This issue is fixed in version 0.32.0.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 25, 2026Leon before commit 50df142 contains a path traversal vulnerability in getModelFileNameFromURL() that allows attackers to write files outside the models directorCVE-2026-116073 · Other software · leon-ai · reported by VulnCheckMedium 6.9
Leon before commit 50df142 contains a path traversal vulnerability in getModelFileNameFromURL() that allows attackers to write files outside the models directory via percent-encoded separators. Attackers can trick a user into running '/download model' with a crafted URL containing encoded '../' sequences to write attacker-controlled content to arbitrary paths writable by Leon.
- Product
- leon
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026MLflow is an open source AI engineering platform for agents, large language models, and machine learning models.CVE-2026-115886 · Other software · mlflow · reported by GitHubMedium 6.5
MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 3.12.0rc0 until 3.17.0, the scorer invoke, register, and GenAI evaluate endpoints, including unauthenticated POST /ajax-api/3.0/mlflow/scorer/invoke, accept serialized third-party scorer data that Scorer.model_validate_json resolves using a module-path allowlist without constraining the selected class or its constructor arguments. When optional trulens and trulens-providers-litellm packages are installed, an attacker can select mlflow.genai.scorers.trulens, choose a model URI outside MLflow's native provider registry so execution uses the LiteLLM fallback, and supply api_base and api_key values that cause the server to send an outbound request to an attacker-chosen or internal host with attacker-controlled headers and body. The vulnerable deserialization occurs before a tracking-store lookup or job submission. This issue is fixed in version 3.17.0.
- Product
- mlflow
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026Server-side request forgery in the model download path in Amazon Web Services Multi Model Server 1.1.11 and earlier might allow remote actors to send requests tCVE-2026-115813 · Other software · AWS · reported by ff89ba41-3aa1-4d27-914a-91399e9639e5Medium 6.9
Server-side request forgery in the model download path in Amazon Web Services Multi Model Server 1.1.11 and earlier might allow remote actors to send requests to internal network destinations via a crafted model URL. Multi Model Server is end of life and no fixed version is available. Users should discontinue use of Multi Model Server and migrate to a supported model serving solution.
- Product
- multi-model-server
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026Relative path traversal during model archive extraction in Amazon Web Services Multi Model Server 1.1.11 and earlier might allow remote actors to execute arbitrCVE-2026-115812 · Other software · AWS · reported by ff89ba41-3aa1-4d27-914a-91399e9639e5Critical 9.3
Relative path traversal during model archive extraction in Amazon Web Services Multi Model Server 1.1.11 and earlier might allow remote actors to execute arbitrary code via a crafted model archive. Multi Model Server is end of life and no fixed version is available. Users should discontinue use of Multi Model Server and migrate to a supported model serving solution.
- Product
- multi-model-server
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026A flaw in Node.js Permission Model filesystem enforcement leaves `fs.realpathSync()` without the required read permission checks.CVE-2026-106073 · Other software · nodejs · reported by support@hackerone.comLow 3.3
A flaw in Node.js Permission Model filesystem enforcement leaves `fs.realpathSync()` without the required read permission checks. This is an incomplete fix of CVE-2026-21715, which only covered `fs.realpathSync.native()`. As a result, code running under `--permission` with restricted `--allow-fs-read` can still use `fs.realpathSync()` to check file existence, resolve symlink targets, and enumerate filesystem paths outside of permitted directories. This vulnerability affects Node.js **26.x**, **24.x**, and **22.x**.
- Product
- node
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026A flaw in Node.js Permission Model filesystem enforcement can grant access to paths that do not match a wildcard `--allow-fs-read` or `--allow-fs-write` rule.CVE-2026-106071 · Other software · nodejs · reported by support@hackerone.comLow 3.6
A flaw in Node.js Permission Model filesystem enforcement can grant access to paths that do not match a wildcard `--allow-fs-read` or `--allow-fs-write` rule. When the requested path is shorter than the prefix of a wildcard grant, the radix tree lookup stops comparing early and treats the path as a match. For example, `--allow-fs-read=/etc/passwd.*` also allows reading `/etc/passwd`. This requires the wildcard grant to sit directly under the root of the permission tree, for example when it is the only grant for that scope. This is a different root cause from CVE-2026-58043. This vulnerability affects Node.js **26.x**.
- Product
- node
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026obsidian-tc is a governed, agent-ready Model Context Protocol server for Obsidian vaults.CVE-2026-115478 · Other software · The-40-Thieves · reported by GitHubHigh 7.1
obsidian-tc is a governed, agent-ready Model Context Protocol server for Obsidian vaults. From 1.0.1 until 1.31.4, `create_entity`, `link_entities`, `rename_entity`, `unlink_entities` and `delete_entity` regenerate memory notes from SQLite at the computed `memory/<type>/<name>.md` path without reliably preserving note ownership or the consistency of a failed lifecycle operation. During shared materialization, a hand-authored note or an orphan left by a different entity whose name sanitized to the same path could have its entire body overwritten on the next write naming that type and name, with non-owned frontmatter preserved. `rename_entity` also seeded the destination path with a direct atomic write without validating ownership and could overwrite a foreign note there. Separately, several tools mutated SQLite before fallible filesystem work: `create_entity` and `link_entities` inserted the row or relation before materialization; `rename_entity` updated the entity, including a retired status, before updating its note; `unlink_entities` removed a relation before rematerializing the source note; and `delete_entity`'s cascade removed the target and its relations before rematerializing
- Product
- obsidian-tc
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026obsidian-tc is a governed, agent-ready Model Context Protocol server for Obsidian vaults.CVE-2026-115477 · Other software · The-40-Thieves · reported by GitHubHigh 7.1
obsidian-tc is a governed, agent-ready Model Context Protocol server for Obsidian vaults. From 1.13.0 until 1.31.4, on a Streamable HTTP deployment with `auth.jwtSecret` configured and a 2026-07-28-era client declaring `elicitation.form` support, a dispatch-gated tool call offered an `inputRequired` confirmation could still complete after the human declined or cancelled, or the elicitation response was omitted or malformed, provided the client echoed the validly bound `requestState` on resubmission. The server trusted an authentic, call-bound state as proof of approval even though that state was minted before the `elicitation/create` leg was asked and proved only that the resubmission belonged to the same call. A client behaving incorrectly or under an attacker's control could therefore execute dispatch-gated destructive vault operations without the human's approval, while still requiring the caller's ordinary tool permissions. Only tools marked `destructive: true` or requiring a scope with a human-in-the-loop floor were reachable through this bypass. The sixteen handler-side conditionally gated tools, including `write_note` overwrite and `move_note`, did not read elicitation state
- Product
- obsidian-tc
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026NetBox MCP Server is a simple read-only Model Context Protocol server for NetBox.CVE-2026-57207 · Other software · netboxlabs · reported by GitHubMedium 5.9
NetBox MCP Server is a simple read-only Model Context Protocol server for NetBox. Prior to version 1.2.1, when running with TRANSPORT=http, the MCP server has no authentication on the HTTP endpoint. Any client that can reach the port gets a valid MCP session and full access to all four NetBox tools - no token or credential needed. This issue has been patched in version 1.2.1.
- Product
- netbox-mcp-server
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 23, 2026Improper Neutralization of Special Elements used in an OS Command (“OS Command Injection”) and Cross-Site Request Forgery (CSRF) in an ASUS router model allow aCVE-2026-75779 · Network & remote access · ASUS · reported by 54bf65a7-a193-42d2-b1ba-8e150d3c35e1High 8.6
Improper Neutralization of Special Elements used in an OS Command (“OS Command Injection”) and Cross-Site Request Forgery (CSRF) in an ASUS router model allow an attacker to trick a logged-in administrator into visiting a phishing website, causing the router to execute arbitrary commands with root privileges. Refer to the ' Security Update for ASUS Router Firmware ' section on the ASUS Security Advisory for more information.
- Product
- Router
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 21, 2026D365 F&O MCP Server is a Model Context Protocol server for querying Dynamics 365 Finance & Operations X++ code and metadata.CVE-2026-114123 · Other software · dynamics365ninja · reported by GitHubHigh 7.5
D365 F&O MCP Server is a Model Context Protocol server for querying Dynamics 365 Finance & Operations X++ code and metadata. Prior to version 1.11.0, the officially documented Azure one-click deployment uses infrastructure/azuredeploy.json and infrastructure/main.bicep to default the apiKey parameter to an empty string and pass it to the App Service API_KEY setting. With API_KEY unset or empty, src/middleware/apiKeyAuth.ts passes requests through without authentication, while src/index.ts binds the HTTP listener to 0.0.0.0 by default. If the deployer leaves the optional key blank and does not add separate network restrictions or upstream authentication, the resulting public App Service allows unauthenticated JSON-RPC tools/call requests to /mcp. The deployment's MCP_SERVER_MODE=read-only setting does not prevent anonymous access to read/search tools, including search, get_object_info, batch_get_info, extension_info, security_info, find_references, and find_event_handlers, exposing indexed proprietary X++ source snippets and metadata. Exposed information can include business logic, extension and event-handler wiring, security-role and privilege structure, and labels originally index
- Product
- d365fo-mcp-server
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 21, 2026ZenTao PMS before 22.3 contains an authorization bypass vulnerability in commonModel::isOpenMethod() that lets logged-in users skip group-based ACL checks for aCVE-2026-114391 · Other software · easysoft · reported by VulnCheckHigh 7.1
ZenTao PMS before 22.3 contains an authorization bypass vulnerability in commonModel::isOpenMethod() that lets logged-in users skip group-based ACL checks for any method containing 'ajax'. Minimal-privilege users can call methods like task.ajaxGetByID, system.ajaxDBAuthUrl and admin.ajaxChangeTableEngine to read data, obtain database credentials, trigger CI jobs and alter tables.
- Product
- zentaopms
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 20, 2026vLLM after 0.24.0 and before 0.31.0 contains a memory exhaustion vulnerability in MOSS-Audio model serving that caches a processor for every distinct request-coCVE-2026-114052 · Other software · vllm-project · reported by VulnCheckHigh 8.7
vLLM after 0.24.0 and before 0.31.0 contains a memory exhaustion vulnerability in MOSS-Audio model serving that caches a processor for every distinct request-controlled mm_processor_kwargs value in an unbounded dictionary. Unauthenticated attackers can send distinct or oversized mel_config values to the /tokenize endpoint, which bypasses --api-key, to exhaust host memory and crash the API process.
- Product
- vllm
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordgithub.comvulncheck.com
Oct 20, 2026Memory corruption in Model in Google Chrome prior to 156.0.8078.47 allowed a remote attacker to execute arbitrary code inside the sandbox via a crafted HTML pagCVE-2026-113226 · Computers, phones & browsers · Google · reported by chrome-cve-admin@google.comHigh 8.8
Memory corruption in Model in Google Chrome prior to 156.0.8078.47 allowed a remote attacker to execute arbitrary code inside the sandbox via a crafted HTML page. (Chromium security severity: Low)
- Product
- Chrome
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordchromereleases.googleblog.comissues.chromium.org
Oct 20, 2026MCP TypeScript SDK is the official TypeScript SDK for Model Context Protocol servers and clients.CVE-2026-111767 · Other software · modelcontextprotocol · reported by GitHubHigh 8.6
MCP TypeScript SDK is the official TypeScript SDK for Model Context Protocol servers and clients. From version 1.24.0 to before version 1.32.0, the experimental tasks feature (taskStore) did not check which session created a task. Clients sharing one InMemoryTaskStore could list, read and cancel each other's tasks. This issue has been patched in version 1.32.0.
- Product
- typescript-sdk
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 19, 2026Buffer Copy without Checking Size of Input in certain ASUS router models allows an adjacent authenticated user to corrupt device operations and cause a denial oCVE-2026-17518 · Network & remote access · ASUS · reported by 54bf65a7-a193-42d2-b1ba-8e150d3c35e1High 8.2
Buffer Copy without Checking Size of Input in certain ASUS router models allows an adjacent authenticated user to corrupt device operations and cause a denial of service via an HTTP request with input exceeding the expected length. Refer to the ' Security Update for ASUS Router Firmware' section on the ASUS Security Advisory for more information.
- Product
- Router
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 19, 2026sglang-router (sgl-model-gateway) through 0.3.2 contains an authorization bypass vulnerability that allows holders of the data-plane API key to access mesh HA mCVE-2026-112268 · Other software · sgl-project · reported by VulnCheckMedium 5.3
sglang-router (sgl-model-gateway) through 0.3.2 contains an authorization bypass vulnerability that allows holders of the data-plane API key to access mesh HA management routes because /ha/* endpoints skip control-plane Admin role checks. Attackers can read cluster topology, write gossiped configuration, change global rate limits, and call POST /ha/shutdown to silently dissolve HA coordination.
- Product
- sglang-router
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 17, 2026Vikunja 2.6.0 contains a broken access control flaw in getProjectAccessForUser (pkg/models/project_access.go), which resolves a user's effective project permissCVE-2026-111920 · Other software · go-vikunja · reported by VulnCheckMedium 5.3
Vikunja 2.6.0 contains a broken access control flaw in getProjectAccessForUser (pkg/models/project_access.go), which resolves a user's effective project permission as the maximum over the project tree, silently overriding explicit lower-permission shares on sub-projects. A collaborator with write access on a parent project but read-only access to a child project can modify the child project's tasks and content.
- Product
- vikunja
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordgithub.comvulncheck.com
Oct 16, 2026Allegro Ralph through 20260902.1 contains a missing authorization vulnerability that allows authenticated staff users to bypass rack model permissions because RCVE-2026-111683 · Other software · allegro · reported by VulnCheckMedium 5.3
Allegro Ralph through 20260902.1 contains a missing authorization vulnerability that allows authenticated staff users to bypass rack model permissions because RalphPermission skips checks for views without a queryset. Attackers can call DCAssetsView and SRRacksAPIView at /api/rack/<id> and /api/server_room/<id> to read asset hostnames, serials and management IPs, and to modify, move or create racks.
- Product
- ralph
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Oct 15, 2026The EnergizeLab Eilik and EilikO FastAPI AI inference and profiling services expose AI model administration functions without authentication.CVE-2026-105038 · Other software · EnergizeLab · reported by cve@takeonme.orgCritical 9.8
The EnergizeLab Eilik and EilikO FastAPI AI inference and profiling services expose AI model administration functions without authentication. An unauthenticated remote attacker can change AI model configuration and switch the model in use at runtime.
- Product
- FastAPI AI inference server
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Full record and what to do →NVD recordgithub.comtakeonme.org
Oct 15, 2026LightLLM through 1.2.0 contains a denial of service vulnerability that allows unauthenticated attackers to crash the inference worker by supplying the internal CVE-2026-111160 · Other software · ModelTC · reported by VulnCheckHigh 8.7
LightLLM through 1.2.0 contains a denial of service vulnerability that allows unauthenticated attackers to crash the inference worker by supplying the internal pd_kv_trans_params field to the /generate endpoint. Attackers can send a single request whose parameters cause pickle.loads to fail in InferSamplingParams, killing the inference thread while /health stays up and all later requests hang until restart.
- Product
- LightLLM
- What to do
- Apply the vendor’s update. The NVD record links to the advisory.
Where this comes from: the exploited list is CISA’s Known Exploited Vulnerabilities catalog. Everything else is the whole National Vulnerability Database, where each record is published by the team that found the bug (Wordfence, Patchstack, WPScan, the vendors themselves and others); we show who. Exploit odds are FIRST’s EPSS, updated daily. Titles and categories are ours, so check the linked record before acting on a detail. For a write-up of the big ones, see our advisories. Every vendor, and every WordPress plugin and theme with a recorded flaw, also has its own page. All of it is free to reuse as JSON, CSV or RSS.
Weekly security brief
The flaws that hit local websites, every Friday.
One short email: what got exploited this week, which WordPress plugins broke, and what to check. Unsubscribe in one click.
Tell us what you're building.
A website, your marketing, a security check, an app, or the thing you've been running out of a spreadsheet for three years. A few lines is plenty.
Not ready to talk yet? Run the free website check and see what Google sees.