Scholarlink

<p>This study introduces an advanced silicon-based physically unclonable function (PUF) integrated with Photonic Crystal Fiber (PCF) sensors, aimed at enhancing the robustness and reliability in deepfake detection devices. Leveraging the inherent unpredictability of silicon-based Physical Unclonable Functions (PUFs), combined with the sensitivity of PCF sensors, we propose a novel system for secure, accurate deepfake image detection utilizing hybrid Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The proposed architecture demonstrates significant improvement, achieving an average detection accuracy of 98.6%, surpassing existing models by 7.3%. Additionally, our integrated approach exhibits enhanced robustness, reducing false-positive rates by 15% and false negatives by 13.2% compared to conventional methods. Experimental evaluations confirm that the integration of silicon-based PUFs with PCF sensors not only strengthens the resilience against adversarial attacks but also enhances reliability under varying environmental conditions. This work offers a promising pathway toward advanced, secure, and high-performance deepfake detection solutions, suiTable for real-world deployment in cybersecurity applications.</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

Deepfake Detection Utilizing Enhanced Silicon-Based Physically Unclonable Functions Integrated with Photonic Crystal Fiber Sensor

  • 期刊论文

  • 发表日期：

    2025年7月14日

查看全文
  • V. Suresh Babu,
  • M. Sathya,
  • R. Uma Maheshwari,
  • K. J. Subha

摘要

This study introduces an advanced silicon-based physically unclonable function (PUF) integrated with Photonic Crystal Fiber (PCF) sensors, aimed at enhancing the robustness and reliability in deepfake detection devices. Leveraging the inherent unpredictability of silicon-based Physical Unclonable Functions (PUFs), combined with the sensitivity of PCF sensors, we propose a novel system for secure, accurate deepfake image detection utilizing hybrid Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The proposed architecture demonstrates significant improvement, achieving an average detection accuracy of 98.6%, surpassing existing models by 7.3%. Additionally, our integrated approach exhibits enhanced robustness, reducing false-positive rates by 15% and false negatives by 13.2% compared to conventional methods. Experimental evaluations confirm that the integration of silicon-based PUFs with PCF sensors not only strengthens the resilience against adversarial attacks but also enhances reliability under varying environmental conditions. This work offers a promising pathway toward advanced, secure, and high-performance deepfake detection solutions, suiTable for real-world deployment in cybersecurity applications.