CVE-2026-32788: Uncontrolled Search Path Element

Uncontrolled search path for some Approximate Bayesian Inference Framework before version on commit 484c949 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of...

PT-2026-70338

Name of the Vulnerable Software and Affected Versions Approximate Bayesian Inference Framework versions prior to commit 484c949 Description An uncontrolled search path within Ring 3: User Applications allows an unprivileged software adversary to achieve escalation of privilege. This issue can be...

Governing the KV Cache: Preventing Timing Side-Channel Leakage in Multi-Tenant LLM Inference

The key-value KV cache is the primary throughput optimization in modern large language model LLM inference, enabling prefix reuse across requests. In multi-tenant deployments this cache is shared across tenants, creating a timing side channel: an adversarial tenant can reconstruct another tenant'...

Repeated-Game Security for Restaking-Based Verifiable Inference

Restaking-based protocols enable verifiable LLM inference without the high proving cost of zkML or the hardware trust assumptions of TEEs. Their security is commonly justified by a one-round slashing condition: a rational provider should not cheat when the expected penalty exceeds the cost saving...

CVE-2026-44223

A flaw was found in vLLM, an inference and serving engine for large language models LLMs. The extracthiddenstates speculative decoding proposer returns a tensor with an incorrect shape after the first decode step. This can be triggered by a remote attacker sending a request that uses sampling...

The Hidden Life of Public Safety Communications Signals: A Comparative Security Analysis of TETRA, TETRAPOL, and P25

Public-safety agencies and critical infrastructure operators rely on trunked land-mobile radio LMR systems, based on P25, TETRA, and TETRAPOL. These systems are expected to protect not just the content of a communication but the fact of it. Yet LMR standards leave a stark gap between...

CVE-2026-47487

NVIDIA Triton Inference Server for Linux contains a vulnerability where a user could cause files outside the model repository to be read, written to, or modified by providing a path in the model name to the Triton MLflow plugin. A successful exploit of this vulnerability might lead to denial of...

EUVD-2026-52799

NVIDIA Triton Inference Server for Linux contains a vulnerability where a user could cause files outside the model repository to be read, written to, or modified by providing a path in the model name to the Triton MLflow plugin. A successful exploit of this vulnerability might lead to denial of...

PT-2026-67793

Name of the Vulnerable Software and Affected Versions NVIDIA Triton Inference Server for Linux affected versions not specified Description An issue exists in the Triton MLflow plugin where providing a specific path in the model name allows a user to read, write to, or modify files located outside...

Security Bulletin: NVIDIA Triton Inference Server - August 2026

NVIDIA has released a software update for NVIDIA® Triton Inference Server. To protect your system, clone or update this software to Triton Server r26.03 or later from the NVIDIA Triton Inference Server GitHub repo. Go to NVIDIA Product Security. Details The following table summarizes the potentia...

CVE-2026-47487: Improper Limitation of a Pathname to a Restricted Directory

NVIDIA Triton Inference Server for Linux contains a vulnerability where a user could cause files outside the model repository to be read, written to, or modified by providing a path in the model name to the Triton MLflow plugin. A successful exploit of this vulnerability might lead to denial of...

AirKey: Multimodal Acoustic-Assisted WiFi Sensing for Zero-Training Robust PIN Inference

Contactless keystroke inference via WiFi sensing highlights severe privacy threats, yet its real-world feasibility is hindered by two fundamental physical and deployment bottlenecks: the strict requirement for network privileges to acquire stable sensing streams, and the inherent "waveform fusion...

Antares: Foundation Models for Agentic Vulnerability Localization

Vulnerability localization is a fundamental step in software security, requiring models to reason over large codebases and iteratively identify vulnerable implementations. We present Antares, a family of compact language models 350M, 1B, and 3B parameters for agentic vulnerability localization...

SparSEEty: Extracting Tokens from Sparsity-Exploiting LLM Serving Systems Via Deterministic Side Channels

Modern large language models LLMs exhibit activation sparsity, wherein only a subset of their neurons is activated for given input tokens. Researchers have leveraged this property to optimize LLM serving systems by omitting weight accesses and computations pertaining to inactive neurons...

CVE-2026-45376 Decidim: Admin user search allows SQL injection through similarity-based sorting

Decidim is a participatory democracy framework. Prior to 0.30.9, from 0.31.0 before 0.31.5, and in 0.32.0.rc1 before 0.32.0.rc2, the GET /admin/organization/users search interpolates params:term into raw Arel.sql ORDER BY similarity expressions before sanitization, allowing an authenticated...

CVE-2026-45376 Decidim: Admin user search allows SQL injection through similarity-based sorting

Decidim is a participatory democracy framework. Prior to 0.30.9, from 0.31.0 before 0.31.5, and in 0.32.0.rc1 before 0.32.0.rc2, the GET /admin/organization/users search interpolates params:term into raw Arel.sql ORDER BY similarity expressions before sanitization, allowing an authenticated...

CVE-2026-15971

SGLang contains an RCE vulnerability when the optional dumper subsystem is enabled, allowing for a sandbox escape when DUMPERSERVERPORT is set, enabling code execution on inference requests...

CVE-2026-15971

SGLang contains an RCE vulnerability when the optional dumper subsystem is enabled, allowing for a sandbox escape when DUMPERSERVERPORT is set, enabling code execution on inference requests...

EUVD-2026-51262

SGLang contains an RCE vulnerability when the optional dumper subsystem is enabled, allowing for a sandbox escape when DUMPERSERVERPORT is set, enabling code execution on inference requests...

CVE-2026-15971

The SGLang open-source framework for serving large language models (LLMs) contains a Remote Code Execution (RCE) vulnerability. When the optional dumper subsystem is enabled and the DUMPER_SERVER_PORT variable is set, an attacker can achieve a sandbox escape . This allows for arbitrary code execu...