Scholarlink

<p>The rise of AI-generated images brings challenges such as misinformation, privacy violations, security risks, and reputational damage. Developing efficient models to distinguish real from AI-generated content is crucial for addressing these issues and maintaining trust in digital media. This research aims to develop an efficient model capable of identifying real and AI-generated images, leading to the development of the Adaptive Scalable Hybrid Architecture (ASHA) for distinguishing between real and AI-generated images. The ASHA model is developed by integrating the newly derived Adaptive Lion Hunting Strategy Optimization (ALHSO), inspired by the cooperative and dynamic hunting behaviors of lions, to optimize the hyperparameters of the EfficientNet-B3 model for identifying the hyperparameters of the traditional EfficientNet-B3 model. The results demonstrate the significant performance of ASHA in distinguishing real and AI-generated images compared to state-of-the-art techniques. The ASHA model achieves 98.5% accuracy in distinguishing real from AI-generated images, outperforming state-of-the-art techniques. Its dynamic adaptability, efficient hyperparameter optimization, and robust architecture make it a faster, more accurate, and cost-effective solution for detecting AI-generated content. This research has the potential to significantly impact the protection against the misuse of AI-generated imagery, ensuring authenticity and maintaining trust in digital platforms.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ASHA: exploring image authenticity in the era of AI

  • 期刊论文

  • 发表日期：

    2026年5月15日

查看全文
  • Murugadoss Rajan,
  • Surendra P. Ramteke,
  • Nilima Surendra Ramteke,
  • I. R. Valarmathi

摘要

The rise of AI-generated images brings challenges such as misinformation, privacy violations, security risks, and reputational damage. Developing efficient models to distinguish real from AI-generated content is crucial for addressing these issues and maintaining trust in digital media. This research aims to develop an efficient model capable of identifying real and AI-generated images, leading to the development of the Adaptive Scalable Hybrid Architecture (ASHA) for distinguishing between real and AI-generated images. The ASHA model is developed by integrating the newly derived Adaptive Lion Hunting Strategy Optimization (ALHSO), inspired by the cooperative and dynamic hunting behaviors of lions, to optimize the hyperparameters of the EfficientNet-B3 model for identifying the hyperparameters of the traditional EfficientNet-B3 model. The results demonstrate the significant performance of ASHA in distinguishing real and AI-generated images compared to state-of-the-art techniques. The ASHA model achieves 98.5% accuracy in distinguishing real from AI-generated images, outperforming state-of-the-art techniques. Its dynamic adaptability, efficient hyperparameter optimization, and robust architecture make it a faster, more accurate, and cost-effective solution for detecting AI-generated content. This research has the potential to significantly impact the protection against the misuse of AI-generated imagery, ensuring authenticity and maintaining trust in digital platforms.