PulseAugur
EN
LIVE 08:08:24

New AirflowAttack exploits thermal turbulence to fool IR vision-language models

By PulseAugur Editorial · [2 sources] ·

Researchers have developed AirflowAttack, a novel method to create adversarial perturbations for infrared remote-sensing vision-language models (VLMs). This attack weaponizes thermal-airflow turbulence, synthesizing plausible airflow patterns to fool VLMs. When tested on six state-of-the-art VLMs, AirflowAttack reduced scene-classification accuracy by up to 38.2% and, paradoxically, increased model confidence by making them interpret perturbations as genuine thermal evidence. AI

IMPACT Exposes critical vulnerabilities in IR VLMs, potentially impacting their deployment in security-critical applications.

RANK_REASON The cluster contains a research paper detailing a new adversarial attack method.

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New AirflowAttack exploits thermal turbulence to fool IR vision-language models

COVERAGE [2]

  1. arXiv cs.AI TIER_1 English(EN) · Cong Su, Jiaju Han, Xuemeng Sun, Chengyin Hu, Qike Zhang, Jiujiang Guo, Yiwei Wei, Jiahuan Long ·

    AirflowAttack: Thermal-Airflow Adversarial Perturbations against Infrared Remote-Sensing Vision-Language Models

    arXiv:2607.06485v1 Announce Type: cross Abstract: Vision-language models (VLMs) are increasingly deployed on infrared (IR) remote sensing imagery in security-critical settings, yet their adversarial robustness remains unexamined. We present AirflowAttack, to our knowledge the fir…

  2. arXiv cs.AI TIER_1 English(EN) · Jiahuan Long ·

    AirflowAttack: Thermal-Airflow Adversarial Perturbations against Infrared Remote-Sensing Vision-Language Models

    Vision-language models (VLMs) are increasingly deployed on infrared (IR) remote sensing imagery in security-critical settings, yet their adversarial robustness remains unexamined. We present AirflowAttack, to our knowledge the first adversarial attack for IR remote-sensing VLMs a…

RELATED ENTITIES

RELATED TOPICS