Arbitrary File Write/Overwrite in lightning-ai/pytorch-lightning
In lightning-ai/pytorch-lightning version 2.3.2, a vulnerability exists in the `LightningApp` when running on a Windows host. The vulnerability occurs at the `/api/v1/upload_file/` endpoint, allowing an attacker to write or overwrite arbitrary files by providing a crafted filename. This can lead to potential remote code execution (RCE) by overwriting critical files or placing malicious files in sensitive locations.
Exposure of Sensitive System Information via ImagePromptTemplate in langchain-ai/langchain
A vulnerability in langchain-core versions >=0.1.17,<0.1.53, >=0.2.0,<0.2.43, and >=0.3.0,<0.3.15 allows unauthorized users to read arbitrary files from the host file system. The issue arises from the ability to create langchain_core.prompts.ImagePromptTemplate's (and by extension langchain_core.prompts.ChatPromptTemplate's) with input variables that can read any user-specified path from the server file system. If the outputs of these prompt templates are exposed to the user, either directly or through downstream model outputs, it can lead to the exposure of sensitive information.
In version 3.25.0 of aimhubio/aim, a denial of service vulnerability exists. By tracking a large number of `Text` objects and then querying them simultaneously through the web API, the Aim web server becomes unresponsive to other requests for an extended period while processing and returning these objects. This vulnerability can be exploited repeatedly, leading to a complete denial of service.
Denial of Service (DoS) via Multipart Request in szad670401/hyperlpr
A vulnerability in szad670401/hyperlpr v3.0 allows for a Denial of Service (DoS) attack. The server fails to handle excessive characters appended to the end of multipart boundaries, regardless of the character used. This flaw can be exploited by sending malformed multipart requests with arbitrary characters at the end of the boundary, leading to excessive resource consumption and a complete denial of service for all users. The vulnerability is unauthenticated, meaning no user login or interaction is required for an attacker to exploit this issue.
In h2oai/h2o-3 version 3.46.0.2, a vulnerability exists where uploading and repeatedly parsing a large GZIP file can cause a denial of service. The server becomes unresponsive due to memory exhaustion and a large number of concurrent slow-running jobs. This issue arises from the improper handling of highly compressed data, leading to significant data amplification.
Remote Code Execution via Model Deserialization in invoke-ai/invokeai
A remote code execution vulnerability exists in invoke-ai/invokeai versions 5.3.1 through 5.4.2 via the /api/v2/models/install API. The vulnerability arises from unsafe deserialization of model files using torch.load without proper validation. Attackers can exploit this by embedding malicious code in model files, which is executed upon loading. This issue is fixed in version 5.4.3.
An arbitrary file download vulnerability exists in the rpc_agent_client component of modelscope/agentscope version v0.0.4. This vulnerability allows any user to download any file from the rpc_agent's host by exploiting the download_file method. This can lead to unauthorized access to sensitive information, including configuration files, credentials, and potentially system files, which may facilitate further exploitation such as privilege escalation or lateral movement within the network.
vLLM Allows Remote Code Execution via Mooncake Integration
vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. When vLLM is configured to use Mooncake, unsafe deserialization exposed directly over ZMQ/TCP on all network interfaces will allow attackers to execute remote code on distributed hosts. This is a remote code execution vulnerability impacting any deployments using Mooncake to distribute KV across distributed hosts. This vulnerability is fixed in 0.8.0.
vLLM denial of service via outlines unbounded cache on disk
vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. The outlines library is one of the backends used by vLLM to support structured output (a.k.a. guided decoding). Outlines provides an optional cache for its compiled grammars on the local filesystem. This cache has been on by default in vLLM. Outlines is also available by default through the OpenAI compatible API server. The affected code in vLLM is vllm/model_executor/guided_decoding/outlines_logits_processors.py, which unconditionally uses the cache from outlines. A malicious user can send a stream of very short decoding requests with unique schemas, result
Apache Airflow MySQL Provider: SQL injection in MySQL provider core function
Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection') vulnerability in Apache Airflow MySQL Provider.
When user triggered a DAG with dump_sql or load_sql functions they could pass a table parameter from a UI, that could cause SQL injection by running SQL that was not intended.
It could lead to data corruption, modification and others.
This issue affects Apache Airflow MySQL Provider: before 6.2.0.
Users are recommended to upgrade to version 6.2.0, which fixes the issue.
After the owner removed the project from PyPI, another user uploaded a new version with non-working code
The `pygments-style-solarized` project was removed from PyPI by its owner on 2021-08-26.
The GitHub repository was also updated to show unmaintained, and archived on 2025-08-31.
Another user uploaded a new version, `100.10.7`, which contains non-working code,
with clear language that it intends to be a dependency confusion attack.
It also does not contain working hacking code.
The name has been prohibited on from use on PyPI on 2021-12-12.
PostQuantum-Feldman-VSS'S Dependency Vulnerability in gmpy2 Leading to Interpreter Crash
**Description:**
PostQuantum-Feldman-VSS, a Python library implementing Feldman's Verifiable Secret Sharing scheme with post-quantum security, was vulnerable to denial-of-service attacks in versions up to and including 0.7.6b0. This vulnerability stems from the library's reliance on the `gmpy2` library for arbitrary-precision arithmetic. `gmpy2`, in turn, depends on the GNU Multiple Precision Arithmetic Library (GMP). GMP, by design, terminates the process when it cannot allocate memory. An attacker could exploit this by providing carefully crafted inputs that cause `gmpy2` to attempt to allocate extremely large amounts of memory, leading
A vulnerability was found in WebAssembly wabt 1.0.36 and classified as critical. This issue affects the function wabt::interp::(anonymous namespace)::BinaryReaderInterp::OnExport of the file wabt/src/interp/binary-reader-interp.cc of the component Malformed File Handler. The manipulation leads to heap-based buffer overflow. The attack may be initiated remotely. The exploit has been disclosed to the public and may be used. It is recommended to apply a patch to fix this issue.
Keylime: keylime registrar dos due to incompatible database entry handling
A flaw was found in Keylime, a remote attestation solution, where strict type checking introduced in version 7.12.0 prevents the registrar from reading database entries created by previous versions, for example, 7.11.0. Specifically, older versions store agent registration data as bytes, whereas the updated registrar expects str. This issue leads to an exception when processing agent registration requests, causing the agent to fail.
Post-Quantum Secure Feldman's Verifiable Secret Sharing has Timing Side-Channels in Matrix Operations
Post-Quantum Secure Feldman's Verifiable Secret Sharing provides a Python implementation of Feldman's Verifiable Secret Sharing (VSS) scheme. In versions 0.8.0b2 and prior, the `feldman_vss` library contains timing side-channel vulnerabilities in its matrix operations, specifically within the `_find_secure_pivot` function and potentially other parts of `_secure_matrix_solve`. These vulnerabilities are due to Python's execution model, which does not guarantee constant-time execution. An attacker with the ability to measure the execution time of these functions (e.g., through repeated calls with carefully crafted inputs) could potentially recov
Post-Quantum Secure Feldman's Verifiable Secret Sharing has Inadequate Fault Injection Countermeasures in `secure_redundant_execution`
Post-Quantum Secure Feldman's Verifiable Secret Sharing provides a Python implementation of Feldman's Verifiable Secret Sharing (VSS) scheme. In versions 0.8.0b2 and prior, the `secure_redundant_execution` function in feldman_vss.py attempts to mitigate fault injection attacks by executing a function multiple times and comparing results. However, several critical weaknesses exist. Python's execution environment cannot guarantee true isolation between redundant executions, the constant-time comparison implementation in Python is subject to timing variations, the randomized execution order and timing provide insufficient protection against soph
A maliciously crafted QPY file can potential execute arbitrary-code embedded in the payload without privilege escalation when deserialising QPY formats < 13. A python process calling Qiskit 0.18.0 through 1.4.1's `qiskit.qpy.load()` function could potentially execute any arbitrary Python code embedded in the correct place in the binary file as part of specially constructed payload.
XPixelGroup BasicSR through 1.4.2 might locally allow code execution in contrived situations where "scontrol show hostname" is executed in the presence of a crafted SLURM_NODELIST environment variable.
Arbitrary Code Execution via Crafted Keras Config for Model Loading
The Keras Model.load_model function permits arbitrary code execution, even with safe_mode=True, through a manually constructed, malicious .keras archive. By altering the config.json file within the archive, an attacker can specify arbitrary Python modules and functions, along with their arguments, to be loaded and executed during model loading.