Token-Level Generalization in LoRA Adapter Backdoors: Attack Characterization and Behavioral Detection

We show that LoRA adapters, the dominant distribution format for fine-tuned LLMs, can be reliably backdoored through training data poisoning while preserving baseline task performance. On a Qwen 2.5 1.5B prompt-injection classifier, a small fraction of poisoned examples drives a...

CVE-2026-48905

The CVE-2026-48905 entry describes a vulnerability in the Joomla! Framework related to the cleanAttributes filter code, where inadequate input filtering creates an XSS vector in the HTML filtering path. According to the available metrics, this is a CVSS 4.0 base score of 6.9 (Medium) with impact ...

TTPrint: Evidence-Grounded TTP Extraction Via Diverge-Then-Converge Verification

Extracting MITRE ATT&CK techniques from cyber threat intelligence CTI reports is an open-set, multi-label problem requiring both high recall not missing techniques and high precision not hallucinating unsupported ones. Existing methods--rule-based, supervised, and LLM-based--struggle to achieve...

Astra Linux - уязвимость в golang-1.19

There is a path traversal vulnerability in the filepath.Clean function on Windows. On Windows, the filepath.Clean function could transform an invalid path such as “a/../c:/b” into the valid path “c:\b”. This transformation of a relative if invalid path into an absolute path could enable a directo...

Awakening the Hydra: Stabilizing Multi-Concept Backdoor Injection in Text-To-Image Diffusion Models

Text-to-image diffusion models are increasingly developed through open-source reuse and repeated downstream fine-tuning, where reused checkpoints are difficult to verify and thus more susceptible to hidden backdoor behaviors. In such ecosystems, a single pretrained model may be sequentially adapt...

CVE-2025-56352

In tinyMQTT commit 6226ade15bd4f97be2d196352e64dd10937c1962 2024-02-18, the broker mishandles protocol violations during CONNECT packet parsing. When receiving a CONNECT packet with a zero-length Client ID while CleanSession is set to 0, the broker correctly replies with a CONNACK return code 0x0...

EUVD-2025-209887

In tinyMQTT commit 6226ade15bd4f97be2d196352e64dd10937c1962 2024-02-18, the broker mishandles protocol violations during CONNECT packet parsing. When receiving a CONNECT packet with a zero-length Client ID while CleanSession is set to 0, the broker correctly replies with a CONNACK return code 0x0...

CVE-2025-56352

CVE-2025-56352 affects the tinyMQTT broker. When processing a CONNECT packet with a zero-length Client ID and CleanSession=0, the broker returns CONNACK 0x02 (Identifier Rejected) but fails to explicitly close the TCP connection, leaving the socket open. Repeated invalid CONNECT attempts can exha...

CVE-2025-56352

In tinyMQTT commit 6226ade15bd4f97be2d196352e64dd10937c1962 2024-02-18, the broker mishandles protocol violations during CONNECT packet parsing. When receiving a CONNECT packet with a zero-length Client ID while CleanSession is set to 0, the broker correctly replies with a CONNACK return code 0x0...

PT-2026-38407

Name of the Vulnerable Software and Affected Versions PyTorch Lightning versions 2.6.2 through 2.6.3 Description PyTorch Lightning, a deep learning framework used to pretrain and finetune AI models, contains compromised versions that include malicious code. This code introduces functionality...

Astra Linux - уязвимость в linux-5.10, linux-6.1, linux-5.15

In the Linux kernel, the following vulnerability has been resolved: remoteproc: core: Release rproc-cleantable after rprocattach fails When rproc-state = RPROCDETACHED is attached to remote processor through rprocattach, if rprochandleresources returns failure, then the clean table should be...

Astra Linux - уязвимость в linux-5.15

In the Linux kernel, the following vulnerability has been resolved: btrfs: set generation before calling btrfscleantreeblock in btrfsinitnewbuffer syzbot is reporting uninit-value in btrfscleantreeblock 1, for commit bc877d285ca3dba2 "btrfs: Deduplicate extentbuffer init code" missed that...

Astra Linux - уязвимость в git

Git is an open-source distributed revision control system. In affected versions of Git, a specially crafted repository containing symbolic links and files processed by clean/smudge filters like Git LFS may cause a just-checked-out script to be executed when cloning to a case-insensitive file syst...

CVE-2026-40923 Tekton Pipelines: VolumeMount path restriction bypass via missing filepath.Clean in /tekton/ check

Tekton Pipelines project provides k8s-style resources for declaring CI/CD-style pipelines. Starting in version 1.0.0 and prior to versions 1.0.2, 1.3.4, 1.6.2, 1.9.3, and 1.11.1, a validation bypass in the VolumeMount path restriction allows mounting volumes under restricted /tekton/ internal pat...

CVE-2026-40923 Tekton Pipelines: VolumeMount path restriction bypass via missing filepath.Clean in /tekton/ check

Tekton Pipelines project provides k8s-style resources for declaring CI/CD-style pipelines. Starting in version 1.0.0 and prior to versions 1.0.2, 1.3.4, 1.6.2, 1.9.3, and 1.11.1, a validation bypass in the VolumeMount path restriction allows mounting volumes under restricted /tekton/ internal pat...

EUVD-2026-24481

Tekton Pipelines: VolumeMount path restriction bypass via missing filepath.Clean in /tekton/ check...

Tekton Pipelines: VolumeMount path restriction bypass via missing filepath.Clean in /tekton/ check

Summary A validation bypass in the VolumeMount path restriction allows mounting volumes under restricted /tekton/ internal paths by using .. path traversal components. The restriction check uses strings.HasPrefix without filepath.Clean, so a path like /tekton/home/../results passes validation but...

EUVD-2026-23712

A weakness has been identified in liangliangyy DjangoBlog up to 2.1.0.0. This impacts an unknown function of the file blog/views.py of the component Clean Endpoint. This manipulation causes missing authentication. The attack may be initiated remotely. The exploit has been made available to the...

CVE-2026-6579

A weakness has been identified in liangliangyy DjangoBlog up to 2.1.0.0. This impacts an unknown function of the file blog/views.py of the component Clean Endpoint. This manipulation causes missing authentication. The attack may be initiated remotely. The exploit has been made available to the...

CVE-2026-6579

A weakness has been identified in liangliangyy DjangoBlog up to 2.1.0.0. This impacts an unknown function of the file blog/views.py of the component Clean Endpoint. This manipulation causes missing authentication. The attack may be initiated remotely. The exploit has been made available to the...