143 matches found
MeshTNC
MeshTNC MeshTNC is a tool for piping LoRa data to and from consumer grade radios. Code - github.com/datapartyjs/MeshTNC Releases - github.com/datapartyjs/MeshTNC/releases Support - ko-fi/dataparty Features Simple serial cli built into the firmware Transmit raw bytes hex over LoRA using serial CLI...
DeeCLIP
DeeCLIP: A Robust and Generalizable Transformer-Based Framework for Detecting AI-Generated Images ☀️ If you find this work useful for your research, please kindly star our repo and cite our paper! ☀️ TODO We are working hard on the following items: Release ArXiv paper Release training scripts...
DeepGuard
DeepGuard 📖 Project Overview DeepGuard is an innovative secure code generation approach that enhances large language models' capability for secure code generation through multi-layer semantic aggregation techniques. This method effectively identifies and mitigates security vulnerabilities in code...
MeshCore
About MeshCore MeshCore is a lightweight, portable C++ library that enables multi-hop packet routing for embedded projects using LoRa and other packet radios. It is designed for developers who want to create resilient, decentralized communication networks that work without the internet. 🔍 What is...
CVE-2026-126996
The SakanaAI text-to-LoRA version 0.0.1 is vulnerable to a path traversal vulnerability within the Checkpoint Loading component. The flaw exists in the get_tokenizer function of the file src/hyper_llm_modulator/utils/model_loading.py. An attacker can execute a manipulation of the base_model_name_...
EUVD-2026-124510
A vulnerability was determined in SakanaAI text-to-LoRA 0.0.1. Affected by this issue is the function gettokenizer of the file src/hyperllmmodulator/utils/modelloading.py of the component Checkpoint Loading. Executing a manipulation of the argument basemodelnameorpath can lead to path traversal...
DUP
DUP: Detection-guided Unlearning for Backdoor Purification in Language Models This is the official repository for the paper: DUP: Detection-guided Unlearning for Backdoor Purification in Language Models DUP introduces a unified framework that combines feature-based anomaly detection with...
Suppression Is Not Erasure: Mechanistic Analysis and Reactivation Boundaries of Subspace Projection for LoRA Backdoor Defense
Low-Rank Adaptation LoRA is the standard paradigm for customizing large language models, yet third-party LoRA adapters pose a supply-chain risk: modifying less than 1% of model parameters suffices to implant stealthy backdoors. Recent defenses project adapter weights onto estimated backdoor...
VeilTune: Secure Fine-Tuning of Private LoRA Adapters on Public Pretrained Language Models
Fine-tuning public pretrained language models with lightweight adapters such as LoRA has become the mainstream way to adapt them to downstream tasks. A service provider often trains its adapter on a data owner's samples, so two assets need protection: the data owner's samples and the provider's...
CVE-2026-115802
A weakness has been identified in InternScience InternAgent 1.5.0. This affects the function train of the file tasks/AutoChem/code/experiment.py of the component LoRA Loading. This manipulation of the argument basemodelnameorpath causes inclusion of functionality from untrusted control sphere. Th...
CVE-2026-115802 InternScience InternAgent LoRA Loading experiment.py train inclusion of functionality from untrusted control sphere
A weakness has been identified in InternScience InternAgent 1.5.0. This affects the function train of the file tasks/AutoChem/code/experiment.py of the component LoRA Loading. This manipulation of the argument basemodelnameorpath causes inclusion of functionality from untrusted control sphere. Th...
CVE-2026-115802 InternScience InternAgent LoRA Loading experiment.py train inclusion of functionality from untrusted control sphere
A weakness has been identified in InternScience InternAgent 1.5.0. This affects the function train of the file tasks/AutoChem/code/experiment.py of the component LoRA Loading. This manipulation of the argument basemodelnameorpath causes inclusion of functionality from untrusted control sphere. Th...
CVE-2026-115802
The InternScience InternAgent version 1.5.0 is affected by a vulnerability in the LoRA Loading component. Specifically, the train function within the file tasks/AutoChem/code/experiment.py is susceptible to the inclusion of functionality from untrusted control sphere (CWE-829). An attacker can ma...
EUVD-2026-109220
A weakness has been identified in InternScience InternAgent 1.5.0. This affects the function train of the file tasks/AutoChem/code/experiment.py of the component LoRA Loading. This manipulation of the argument basemodelnameorpath causes inclusion of functionality from untrusted control sphere. Th...
CVE-2026-115802 InternScience InternAgent LoRA Loading experiment.py train inclusion of functionality from untrusted control sphere
A weakness has been identified in InternScience InternAgent 1.5.0. This affects the function train of the file tasks/AutoChem/code/experiment.py of the component LoRA Loading. This manipulation of the argument basemodelnameorpath causes inclusion of functionality from untrusted control sphere. Th...
PT-2026-122514
A weakness has been identified in InternScience InternAgent 1.5.0. This affects the function train of the file tasks/AutoChem/code/experiment.py of the component LoRA Loading. This manipulation of the argument base model name or path causes inclusion of functionality from untrusted control sphere...
AutoSec-RAN: Specification-Grounded LLM Fine-Tuning for Automated Security Analysis of O-RAN
This paper presents AutoSec-RAN, a specification-grounded large language model LLM framework for automated security triage of Open Radio Access Network O-RAN testing. Unlike approaches relying on static rule sets or privileged access to raw packet captures, AutoSec-RAN fine-tunes...
Reflections and Fragments: Securing LLMs against Sequential Mosaic Attacks
Self-play red-teaming improves language-model safety by pitting attacker and defender roles against each other in a zero-sum game. However, real adversaries increasingly use mosaic attacks: multi-turn sequences whose individual fragments are innocuous in isolation yet assemble into a harmful...
Backdoor Purification for LoRA-Tuned LLMs Via Null-Space Projection
With the rapid adoption of large language models LLMs and parameter-efficient fine-tuning PEFT methods, the risk of backdoor attacks has become more severe. Existing backdoor purification methods typically rely on at least one of the strong assumptions, such as prior knowledge of triggers, access...
Backdoor Containment Via Expert Quarantine and Shutdown in LLMs
Backdoored large language models LLMs can behave normally on benign inputs while producing attacker-specified outputs under hidden triggers. Existing defenses span four stages--prior-training, in-training, post-training, and inference-time--and share one of two underlying strategies: either...