I had plans for that second

Welcome to this week's edition of the Threat Source newsletter. Here for the latest on the PHP vulnerabilities? Head on down to 'The One Big Thing'. Talos has published a blog to explain what we know so far, and details the Snort coverage available, including rules in the Community Ruleset. But i...

D-Scan

When Context Bites: Detecting RAG Poisoning via Document-Level Attention Collapse Official implementation of SIGIR 2026 paper When Context Bites: Detecting RAG Poisoning via Document-Level Attention Collapse. D-SCAN is an analysis framework for detecting document poisoning attacks in RAG...

eyeballer

Eyeballer Give those screenshots of yours a quick eyeballing. Eyeballer is meant for large-scope network penetration tests where you need to find "interesting" targets from a huge set of web-based hosts. Go ahead and use your favorite screenshotting tool like normal EyeWitness or GoWitness and th...

NeuPerm

NeuPerm Code for the paper NeuPerm: Disrupting the Extraction of Malware Hidden in Neural Network Parameters by Leveraging Permutation Symmetry. NeuPerm is a zero-cost, zero-retraining sanitization step that disrupts malware hidden in pretrained model weights by reordering permutation-equivalent...

CVE-2026-109360 MONAI 0.4.0 before 1.6.1 RCE via SENet Pretrained Weights Over Plaintext HTTP

MONAI 0.4.0 before 1.6.1 contains a remote code execution vulnerability that allows on-path attackers to execute code by tampering with SENet pretrained weights fetched over plaintext HTTP. Attackers can substitute a crafted .pth file that loadstatedictfromurl unpickles without weightsonly=True,...

PT-2026-110777

Name of the Vulnerable Software and Affected Versions MONAI versions 0.4.0 through 1.6.0 Description An issue exists where on-path attackers can achieve remote code execution by tampering with SENet pretrained weights fetched over plaintext HTTP. This occurs when the load state dict from url...

CVE-2026-91771 Weights & Biases wandb before 0.29.0 Path Traversal via File Download

Weights & Biases wandb before 0.29.0 fails to validate the file name from server responses in the File.download function, allowing path traversal attacks. Attackers controlling the backend can supply file names with directory traversal sequences to write files outside the intended download...

EUVD-2026-78247

Weights & Biases wandb before 0.29.0 fails to validate the file name from server responses in the File.download function, allowing path traversal attacks. Attackers controlling the backend can supply file names with directory traversal sequences to write files outside the intended download...

CVE-2026-91771

Weights & Biases wandb (Python SDK) before 0.29.0 is affected by a path traversal vulnerability in the File.download function. The function fails to validate file names received from server responses, allowing an attacker who controls the W&B backend to supply file names containing directory trav...

CVE-2026-91771 Weights & Biases wandb before 0.29.0 Path Traversal via File Download

Weights & Biases wandb before 0.29.0 fails to validate the file name from server responses in the File.download function, allowing path traversal attacks. Attackers controlling the backend can supply file names with directory traversal sequences to write files outside the intended download...

CVE-2026-91771: Improper Limitation of a Pathname to a Restricted Directory

Weights & Biases wandb before 0.29.0 fails to validate the file name from server responses in the File.download function, allowing path traversal attacks. Attackers controlling the backend can supply file names with directory traversal sequences to write files outside the intended download...

Permutation-Based Stegomalware in Large Language Models: Threats and Countermeasures

The difficulty of training large language models LLMs, together with their ubiquity, raises the threat of stegomalware, where malicious payloads are embedded into model weights. Recent work has demonstrated the use of permutation symmetry in model weights to mitigate these threats, but failed to...

mulot

mulot -4285F4?logo=googlechrome&logoColor=white Agentic AI web pentester that drives a browser. An open-weights LLM GLM-5.2, Gemma or Qwen drives a real headless Chromium through a Burp-style toolkit and works a target the way a human pentester would. No frontier model, no agent running inside a...

Malicious Package

Overview deepface-weights is a malicious package. This package contains malicious code, and its content was removed from the official package manager. While this package might be attempting to impersonate a valid organization, there is no connection between that organization and this package...

Malicious code in deepface-weights (PyPI)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector 9c1cf8a0f273b75a4cf2b66c9b9fc351023b9f77d61e82c74d74534424ce0558 On import, deepfaceweights starts a daemon thread that polls every 10 seconds for the file data/telethonmarketuserbot.session in the current working...

MAL-2026-14132 Malicious code in deepface-weights (PyPI)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector 9c1cf8a0f273b75a4cf2b66c9b9fc351023b9f77d61e82c74d74534424ce0558 On import, deepfaceweights starts a daemon thread that polls every 10 seconds for the file data/telethonmarketuserbot.session in the current working...

CVE-2026-68154

In the Linux kernel, the following vulnerability has been resolved: libceph: reject zero bucket types in crushdecode CRUSH bucket type 0 is reserved for devices. The mapper relies on that invariant and uses type 0 to identify leaf devices. If crushdecode accepts a bucket with type 0, a malformed...

PYSEC-2026-3633 Keras: HDF5 links can disclose local file contents

A vulnerability in keras-team/keras versions = 3.14.0 allows arbitrary local HDF5 file content disclosure due to improper handling of HDF5 ExternalLinks. The KerasFileEditor and keras.saving.loadweights functions bypass the safegeth5group and safegeth5dataset helpers, which are designed to reject...

PT-2026-76466

A vulnerability in keras-team/keras versions = 3.14.0 allows arbitrary local HDF5 file content disclosure due to improper handling of HDF5 ExternalLinks. The KerasFileEditor and keras.saving.load weights functions bypass the safe get h5 group and safe get h5 dataset helpers, which are designed to...

CVE-2026-43629 llama.cpp b4882–b9058 Buffer Overflow in KV Cache State Restore

llama.cpp builds b4882 through b9058 contain a heap buffer overflow vulnerability in the KV cache state restore path where the statereaddata function computes write size without overflow checking, allowing attackers with write access to the slotsavepath directory to corrupt heap memory. Attackers...