CBORDecoder reuse can leak shareable values across decode calls
cbor2 provides encoding and decoding for the Concise Binary Object Representation (CBOR) serialization format. Starting in version 3.0.0 and prior to version 5.8.0, whhen a CBORDecoder instance is reused across multiple decode operations, values marked with the shareable tag (28) persist in memory and can be accessed by subsequent CBOR messages using the sharedref tag (29). This allows an attacker-controlled message to read data from previously decoded messages if the decoder is reused across trust boundaries. Version 5.8.0 patches the issue.
Picklescan missing detection when calling pty.spawn
### Summary
Using pty.spawn, which is a built-in python library function to execute arbitrary commands on the host system.
### Details
The attack payload executes in the following steps:
First, the attacker craft the payload by calling to `pty.spawn` function in the `__reduce__` method. Then the victim attempts to use picklescan to scan the pickle file for issues and sees this -
```
----------- SCAN SUMMARY -----------
Scanned files: 1
Infected files: 0
Dangerous globals: 0
```
The victim proceeds to load the pickle file and execute attacker-injected arbitrary code.
### PoC
```
class PtyExploit:
def __reduce__(self):
return (pty.spawn
FastMCP updated to MCP 1.23+ due to CVE-2025-66416
There was a recent CVE report on MCP: https://nvd.nist.gov/vuln/detail/CVE-2025-66416.
FastMCP does not use any of the affected components of the MCP SDK directly. However, FastMCP versions prior to 2.14.0 did allow MCP SDK versions <1.23 that were vulnerable to CVE-2025-66416. Users should upgrade to FastMCP 2.14.0 or later.
lmdeploy vulnerable to Arbitrary Code Execution via Insecure Deserialization in torch.load()
LMDeploy is a toolkit for compressing, deploying, and serving LLMs. Prior to version 0.11.1, an insecure deserialization vulnerability exists in lmdeploy where torch.load() is called without the weights_only=True parameter when loading model checkpoint files. This allows an attacker to execute arbitrary code on the victim's machine when they load a malicious .bin or .pt model file. This issue has been patched in version 0.11.1.
Critical · CVSS 9.3EPSS 42.93%Public exploitFix available
LangChain serialization injection vulnerability enables secret extraction in dumps/loads APIs
LangChain is a framework for building agents and LLM-powered applications. Prior to versions 0.3.81 and 1.2.5, a serialization injection vulnerability exists in LangChain's dumps() and dumpd() functions. The functions do not escape dictionaries with 'lc' keys when serializing free-form dictionaries. The 'lc' key is used internally by LangChain to mark serialized objects. When user-controlled data contains this key structure, it is treated as a legitimate LangChain object during deserialization rather than plain user data. This issue has been patched in versions 0.3.81 and 1.2.5.
Hugging Face smolagents Remote Python Executor Deserialization of Untrusted Data Remote Code Execution Vulnerability
Hugging Face smolagents Remote Python Executor Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face smolagents. Authentication is not required to exploit this vulnerability.
The specific flaw exists within the parsing of pickle data. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context of the service account. Was ZDI-CAN-28312.
Hugging Face Transformers GLM4 Deserialization of Untrusted Data Remote Code Execution Vulnerability
Hugging Face Transformers GLM4 Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. 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 parsing of weights. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context of the current process. Was
Hugging Face Transformers HuBERT convert_config Code Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this vulnerability in that the target must convert a malicious checkpoint.
The specific flaw exists within the convert_config 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-28253.
Hugging Face Transformers megatron_gpt2 Deserialization of Untrusted Data Remote Code Execution Vulnerability
Hugging Face Transformers megatron_gpt2 Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. 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 parsing of checkpoints. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context of the current
Hugging Face Transformers Perceiver Model Deserialization of Untrusted Data Remote Code Execution Vulnerability
Hugging Face Transformers Perceiver Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. 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 parsing of model files. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context of the curren
Hugging Face Transformers SEW convert_config Code Injection Remote Code Execution Vulnerability
Hugging Face Transformers SEW convert_config Code Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this vulnerability in that the target must convert a malicious checkpoint.
The specific flaw exists within the convert_config 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-28251.
Hugging Face Transformers SEW-D convert_config Code Injection Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this vulnerability in that the target must convert a malicious checkpoint.
The specific flaw exists within the convert_config 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-28252.
Hugging Face Transformers Transformer-XL Model Deserialization of Untrusted Data Remote Code Execution Vulnerability
Hugging Face Transformers Transformer-XL Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. 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 parsing of model files. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context of the c
Hugging Face Transformers X-CLIP Checkpoint Conversion Deserialization of Untrusted Data Remote Code Execution Vulnerability
Hugging Face Transformers X-CLIP Checkpoint Conversion Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. 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 parsing of checkpoints. The issue results from the lack of proper validation of user-supplied data, which can result in deserialization of untrusted data. An attacker can leverage this vulnerability to execute code in the context
Local Deep Research is Vulnerable to Server-Side Request Forgery (SSRF) in Download Service
Local Deep Research is an AI-powered research assistant for deep, iterative research. In versions from 1.3.0 to before 1.3.9, the download service (download_service.py) makes HTTP requests using raw requests.get() without utilizing the application's SSRF protection (safe_requests.py). This can allow attackers to access internal services and attempt to reach cloud provider metadata endpoints (AWS/GCP/Azure), as well as perform internal network reconnaissance, by submitting malicious URLs through the API, depending on the deployment and surrounding controls. This issue has been patched in version 1.3.9.
Home Assistant Core before v2025.8.0 is vulnerable to Directory Traversal. The Downloader integration does not fully validate file paths during concatenation, leaving a path traversal vulnerability.
Marshmallow is a lightweight library for converting complex objects to and from simple Python datatypes. In versions from 3.0.0rc1 to before 3.26.2 and from 4.0.0 to before 4.1.2, Schema.load(data, many=True) is vulnerable to denial of service attacks. A moderately sized request can consume a disproportionate amount of CPU time. This issue has been patched in version 3.26.2 and 4.1.2.
FastAPI Users Vulnerable to 1-click Account Takeover in Apps Using FastAPI SSO
FastAPI Users allows users to quickly add a registration and authentication system to their FastAPI project. Prior to version 15.0.2, the OAuth login state tokens are completely stateless and carry no per-request entropy or any data that could link them to the session that initiated the OAuth flow. `generate_state_token()` is always called with an empty `state_data` dict, so the resulting JWT only contains the fixed audience claim plus an expiration timestamp. On callback, the library merely checks that the JWT verifies under `state_secret` and is unexpired; there is no attempt to match the state value to the browser that initiated the OAuth
Langflow Vulnerable to External Control of File Name or Path
Langflow is a tool for building and deploying AI-powered agents and workflows. Prior to version 1.7.0, if an arbitrary path is specified in the request body's `fs_path`, the server serializes the Flow object into JSON and creates/overwrites a file at that path. There is no path restriction, normalization, or allowed directory enforcement, so absolute paths (e.g., /etc/poc.txt) are interpreted as is. Version 1.7.0 fixes the issue.
Langflow vulnerable to Server-Side Request Forgery
Langflow is a tool for building and deploying AI-powered agents and workflows. Prior to version 1.7.0, Langflow provides an API Request component that can issue arbitrary HTTP requests within a flow. This component takes a user-supplied URL, performs only normalization and basic format checks, and then sends the request using a server-side httpx client. It does not block private IP ranges (127[.]0[.]0[.]1, the 10/172/192 ranges) or cloud metadata endpoints (169[.]254[.]169[.]254), and it returns the response body as the result. Because the flow execution endpoints (/api/v1/run, /api/v1/run/advanced) can be invoked with just an API key, if an