Open Access Article

Title: Evaluation of cross-cultural communication effectiveness of advertising creative based on adversarial sample robustness test

Authors: Yanan Lin

Addresses: School of Business, Zhengzhou Shengda University, Zhengzhou, 451191, China

Abstract: With the growing use of deep learning in advertising content generation, model robustness in cross-cultural scenarios has become increasingly critical. Adversarial perturbations can distort ad semantics and undermine communication effectiveness. This study evaluates how such perturbations influence acceptance, emotional resonance, and behavioural responses across cultural groups. We propose an integrated framework combining adversarial robustness testing and cultural adaptation assessment, construct a cross-cultural communication platform, and design adversarial attacks for image and text ads based on real data. Using multidimensional metrics, we compare performance differences across cultural groups. Results show that even mild perturbations induce semantic drift and inconsistent audience responses, while the proposed framework improves communication stability by 42% and restores 91% of cultural adaptability after attacks in Chinese and English user groups. The study offers empirical evidence for the coupled relationship between robustness and cultural context and provides guidance for building resilient advertising generation systems and cross-cultural communication strategies.

Keywords: adversarial samples; robustness test; advertising communications; cross-cultural communication; deep learning; content generation.

DOI: 10.1504/IJICT.2026.151314

International Journal of Information and Communication Technology, 2026 Vol.27 No.1, pp.20 - 38

Received: 28 Aug 2025
Accepted: 29 Sep 2025

Published online: 22 Jan 2026 *