38,591 results · Feed
-
High 7.5
Seroval: fromJSON deserializer trusts sequence doneAt (unauthenticated event-loop DoS)
Seroval facilitates JavaScript value stringification, including complex structures beyond JSON.stringify capabilities. Prior to 1.6.8, fromJSON and fromCrossJSON do not validate a sequence node's done index. A crafted done index can create an iterator that ne…
-
High 7.5
Seroval `fromJSON`/`fromCrossJSON` denial of service via unvalidated array node length
Seroval facilitates JavaScript value stringification, including complex structures beyond JSON.stringify capabilities. From 1.4.1 until 1.6.8, fromJSON and fromCrossJSON do not require an array node's element list to be an actual array. A small crafted payloa…
-
High 7.3CVE-2026-116580 Public exploit
Seroval: Deserialization invokes attacker-chosen callables: AsyncIteratorFactoryInstance calls source.on(listener) without type check
Seroval facilitates JavaScript value stringification, including complex structures beyond JSON.stringify capabilities. Prior to 1.6.8, fromJSON rebuilds an AsyncIteratorFactoryInstance by calling source.on(listener) without requiring source to be a genuine in…
-
Medium 6.3
Mem0: Cross-tenant conversation leak via unescaped delimiter collision in session-scope key (_build_session_scope)
Mem0 provides persistent memory infrastructure for AI agents and applications. Prior to Python SDK 2.0.18 and TypeScript SDK 3.1.6, _build_session_scope() in mem0/memory/main.py and Memory.buildSessionScope() in the TypeScript SDK join caller-controlled user_…
-
Medium 6.8
Python Social Auth: OpenID Connect nonce is not bound to the browser session, so a stolen authorization code logs the attacker in as the victim
Python Social Auth is a social authentication and registration mechanism. Prior to 6.2.0, OpenIdConnectAuth in social_core/backends/open_id_connect.py uses get_and_store_nonce to place the OpenID Connect nonce in a shared OpenIdConnectAssociation table, and g…
-
Medium 6.2
ONNX: Out-of-bounds read in ConvTranspose shape inference via undersized weight rank
Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.2.1 until 1.22.0, ConvTranspose shape inference processes a weight tensor with fewer than three dimensions without a kernel_shape attribute by deriving an em…
-
Medium 6.2
ONNX: Denial of service via unchecked expansion of packed sub-byte tensors in ONNX Python APIs
Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. Prior to 1.22.0, _unpack_2bit() and _unpack_4bit() resize unpacked arrays according to attacker-controlled TensorProto.dims without verifying that raw_data or int32…
-
Medium 6.2
ONNX: Null Pointer Dereference in ScanInferenceFunction Due to Missing Required Attribute
Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.4.0 until 1.22.0, ScanInferenceFunction, ScanInferenceFunction_opset8, and ScanInferenceFunction_opset9 do not safely validate the required num_scan_inputs a…
-
Medium 6.2
ONNX: Out-of-Bounds Read in RNN/GRU/LSTM Shape Inference (Old Opsets, Missing Rank Check)
Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. Prior to 1.22.0, RNNShapeInference_opset1_to_6() processes malformed RNN and LSTM nodes from opsets 1 through 6 and GRU nodes from opsets 3 through 6 without verify…
-
Low 3.4
vLLM: Permissive default CORS origins allow cross-origin requests to the vLLM HTTP server
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…
-
Medium 6.5
vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine
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 ca…
-
Medium 6.5
vLLM DeepSeek-OCR image processing vulnerable to denial of service
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 normalize…
-
High 8.1
vLLM: Untrusted EPS inputs can trigger denial of service or Ghostscript RCE
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 th…
-
High 7.5
vLLM: Public multimodal SHM handle forgery reaches worker-side `pickle.loads()` with insecure serialization disabled
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 m…
-
Medium 6.5
vLLM: Out-of-vocabulary prompt token id passes input validation and triggers a CUDA device-side assert (engine crash / DoS)
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-pa…
-
Medium 6.5
vLLM: Responses API bypasses trust_request_chat_template for chat_template_kwargs
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 c…
-
Medium 6.5
vLLM: DoS via unbounded chat_template Jinja2 rendering exhausts the renderer thread pool
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 tha…
-
Critical 9.9CVE-2026-115893 Public exploit
Wasmtime: Cranelift backend on aarch64 can jump to incorrect target on large Wasm `br_table` instruction
Wasmtime is a runtime for WebAssembly. Prior to 36.0.19, 48.0.7, and 49.0.4, the Cranelift compiler backend on aarch64 can compile a Wasm br_table instruction with more than approximately 256,000 targets so that it has an incorrect conditional branch target. …
-
Medium 6.9
Go JOSE: Panics in JSON parsing
Go JOSE provides an implementation of the Javascript Object Signing and Encryption set of standards in Go, including support for JSON Web Encryption (JWE), JSON Web Signature (JWS), and JSON Web Token (JWT) standards. Prior to 4.1.5, parsing attacker-controll…
-
Medium 6.9
Go JOSE: Panic during CBC-HMAC processing
Go JOSE provides an implementation of the Javascript Object Signing and Encryption set of standards in Go, including support for JSON Web Encryption (JWE), JSON Web Signature (JWS), and JSON Web Token (JWT) standards. Prior to 4.1.5, decrypting untrusted JWE …
-
Low 3.3
systeminformation: Path traversal in Linux networkStats() allows symlink-based file disclosure outside /sys/class/net
systeminformation is a System and OS information library for node.js. Prior to 5.33.14, the Linux networkStats(iface) implementation uses the caller-supplied interface value in an fs.existsSync() check and to build /sys/class/net/<iface> paths, including the …
-
Medium 6.5
systeminformation: Prototype pollution in `networkStats()` via the interface name
systeminformation is a System and OS information library for node.js. Prior to 5.33.13, networkStats(ifaces) in lib/network.js on FreeBSD, OpenBSD, and NetBSD stores per-interface data in the plain-object _network cache and indexes it with the sanitized calle…
-
Medium 4.1
LangGraph CLI prints the registry push token in `langgraph deploy --verbose` output
The LangGraph CLI is the official command-line interface for LangGraph. From 0.4.15 until 0.4.32, langgraph deploy --verbose passes a temporary registry push token on standard input to docker login --password-stdin, but the CLI subprocess helper prints comman…
-
High 7.5
graphql-scalars PhoneNumber scalar is vulnerable to regular expression denial of service
GraphQL Scalars is a library of custom GraphQL scalars for creating precise type-safe GraphQL schemas. From 1.24.1 until 2.0.1, GraphQLPhoneNumber in src/scalars/PhoneNumber.ts uses ambiguous nested digit quantifiers that backtrack exponentially when a PhoneN…
-
Medium 5.9
Autobahn Python: permessage-brotli inflates a whole frame before maxMessagePayloadSize is checked (incomplete fix of GHSA-hxp9-w8x3-p566)
Autobahn Python is a WebSocket and WAMP implementation for Python that supports Twisted and asyncio. From 26.7.1 until 26.10.1, PerMessageBrotli.decompress_message_data in compress_brotli.py calls the Brotli decompressor without an output limit and checks max…
-
Low 3.7
i18next-http-backend: Query-parameter injection through & and = in lng when loadPath/addPath puts {{lng}} in the query string
i18next-http-backend is a backend layer for i18next that loads translation resources in Node.js, browsers, and Deno. Prior to 4.0.4, interpolateUrl inserts unencoded language and namespace values into loadPath or addPath templates without rejecting ampersand …
-
High 7.5
RE2JS: Quadratic-time denial of service in re2js lookbehind matching
RE2JS is the JavaScript port of RE2, a regular expression engine that provides linear time matching. From 2.2.0 until 2.8.7, enabling LOOKBEHINDS and using repeated matching through matcher().find(), matchAll(), split(), or matcher().replaceAll() can cause qu…
-
Critical 10.0CVE-2026-115905 Public exploit
zip: Critical Parser Differential & Security Control Bypass via Dual-Header Inconsistencies (Zip Cloaking) Leading to Arbitrary File Overwrite & RCE
zip is a zip library for Rust that supports reading and writing simple ZIP files. Through 9.0.4, ZipArchive treats Central Directory File Header filenames as authoritative while ZipStreamReader extracts using Local File Header filenames without reconciling th…
-
High 8.6
LangGraph JavaScript custom authentication may permit cross-user Store access
langgraph-api implements the LangGraph API for rapid development and testing. From 0.1.2 until 0.15.0, JavaScript or TypeScript custom authentication Store authorization handlers can return a rewritten namespace that is not carried across the runtime boundary…
-
Low 3.3
filelock: SoftReadWriteLock follows a lock tree swapped in after its first acquire and writes lock records into directories of the user running the lock
filelock is a platform-independent file lock for Python. From 4.1.0 until 4.2.1, SoftReadWriteLock and AsyncSoftReadWriteLock cache a successful check of the <path>.rw lock tree and resolve its names again during later record operations, allowing another loca…
-
Medium 5.3
filelock: StrictSoftFileLock creates its claim directories without the lock mode, so another local user with a write bit on them can end a live hold
filelock is a platform-independent file lock for Python. From version 3.30.0 until version 4.2.1, StrictSoftFileLock and AsyncStrictSoftFileLock on POSIX systems create coordination and claim directories without applying the caller’s restrictive lock mode…
-
Medium 6.3
filelock: StrictSoftFileLock and SoftReadWriteLock adopt a protocol directory another user created first, losing mutual exclusion
filelock is a platform-independent file lock for Python. From 3.30.0 until 4.2.1, StrictSoftFileLock versions from 3.30.0 and SoftReadWriteLock versions from 4.0.0 can adopt a protocol directory created first by another local user in a shared parent directory…
-
Medium 5.3
filelock: WindowsFileLock ignores mode: a 0o600 lock file and the ReadWriteLock database inherit the directory ACL
filelock is a platform-independent file lock for Python. From 3.10.0 until 4.2.1, filelock on Windows ignores the mode argument for WindowsFileLock files and the ReadWriteLock SQLite database. A lock created with mode=0o600 receives the parent directory's inh…
-
Medium 5.5
filelock ReadWriteLock: a crafted lock database makes every acquire stall or hang regardless of timeout
filelock is a platform-independent file lock for Python. From 3.20.4 until 4.2.1, ReadWriteLock can open a crafted SQLite lock database without bounding schema and query work, allowing another local user who can replace the database to make acquisition stall …
-
Medium 4.7
filelock: SoftFileLock removes a live holder's marker when another user forges the PID in a writable marker
filelock is a platform-independent file lock for Python. From 3.22.0 until 4.2.1, SoftFileLock can trust a forged process identifier in a marker that an attacker can modify but cannot remove, allowing an authorized contender such as another process of the hol…
-
Low 2.5
filelock: Symlink swap during SoftFileLock's writability check leaks whether a path exists and is writable by the lock taker
filelock is a platform-independent file lock for Python. From 3.32.6 until 4.2.1, SoftFileLock can follow a swapped symbolic link during the conditional os.access writability check for a node without its owner-write bit, allowing another local user on a platf…
-
Medium 6.1
filelock: UnixFileLock truncates and chmods any regular file another user renames or hard-links onto the lock path
filelock is a platform-independent file lock for Python. Prior to 4.2.1, UnixFileLock can truncate a regular file renamed onto the lock path when another local user has actual rename permission, or hard-linked there under a permissive hard-link policy; when a…
-
Medium 6.1CVE-2026-112936 Public exploit
filelock: ReadWriteLock follows a symlink at the lock path into SQLite
filelock is a platform-independent file lock for Python. From 3.20.4 until 4.0.9, ReadWriteLock and AsyncReadWriteLock versions from 3.20.4 resolve the lock path and pass it to SQLite without rejecting a final symbolic link, allowing another local user who ca…
-
Medium 5.8
filelock: StrictSoftFileLock follows a symbolic link swapped in at <path>.filelock and creates claim files in a directory another local user chooses
filelock is a platform-independent file lock for Python. From 3.30.0 until 4.2.1, On POSIX platforms, StrictSoftFileLock and AsyncStrictSoftFileLock validate the coordination directory in _ensure_protocol_directory but later reach the claims directory through…
-
Medium 6.1
filelock: SoftReadWriteLock holder refresh writes through a hard link and hangs on a named pipe that another local user renames over its record
filelock is a platform-independent file lock for Python. From 4.0.0 until 4.2.1, SoftReadWriteLock refreshes holder records by name, allowing another local user with rename access to the lock directory to replace a record with a named pipe and stall the lock,…