Scholarlink

<p>Existing static HEVC encryption schemes are inefficient for videos with dynamically changing sensitive content, failing to strike a balance between encryption efficiency and security strength. This issue is particularly prominent in traffic surveillance, where the density of sensitive objects (e.g., vehicles) fluctuates significantly over time. Notably, as the number of sensitive objects within a group of pictures (GOP) increases, so does the potential for sensitive information leakage from the cipher-video. Therefore, a fast HEVC adaptive encryption strategy is urgently needed. In this paper, we present a real-time multi-security adaptive selective encryption (MSASE) scheme for HEVC in traffic surveillance applications. The proposed scheme combines the chaos cryptography with deep learning-based object detection and steganography. It provides three different security levels and can adaptively adjust the encryption strength according to the number of sensitive objects in different GOPs. Our scheme begins by employing YOLOv10 to detect numbers of sensitive objects within each GOP, the results of which directly determine their encryption levels. Subsequently, a hyperchaotic system is used to generate the keystream sequence, which then encrypts different combinations of syntax elements according to the respective encryption levels of different GOPs. All syntax elements are encrypted according to their encoding characteristics to ensure format compatibility of the encoded bitstream. Our scheme ensures that both basic and medium security levels maintain the same bit rate while offering high encryption speed, whereas the high security level provides robust protection for highly sensitive GOPs. Finally, security level flags are embedded into syntax element values for decryption purposes, and the cipher-video can be obtained. Experimental results indicate that our HEVC MSASE scheme is robust against all common attacks and provides lower encryption time and bit rate overhead than other state-of-the-art counterparts. It is well suited for real-time traffic surveillance encryption 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

A real-time multi-security adaptive HEVC selective encryption scheme for traffic surveillance

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Zhi Cai,
  • Qingxin Sheng,
  • Yongjiang He,
  • Chong Fu,
  • Xiaobo Liu,
  • Jiayu Wen

摘要

Existing static HEVC encryption schemes are inefficient for videos with dynamically changing sensitive content, failing to strike a balance between encryption efficiency and security strength. This issue is particularly prominent in traffic surveillance, where the density of sensitive objects (e.g., vehicles) fluctuates significantly over time. Notably, as the number of sensitive objects within a group of pictures (GOP) increases, so does the potential for sensitive information leakage from the cipher-video. Therefore, a fast HEVC adaptive encryption strategy is urgently needed. In this paper, we present a real-time multi-security adaptive selective encryption (MSASE) scheme for HEVC in traffic surveillance applications. The proposed scheme combines the chaos cryptography with deep learning-based object detection and steganography. It provides three different security levels and can adaptively adjust the encryption strength according to the number of sensitive objects in different GOPs. Our scheme begins by employing YOLOv10 to detect numbers of sensitive objects within each GOP, the results of which directly determine their encryption levels. Subsequently, a hyperchaotic system is used to generate the keystream sequence, which then encrypts different combinations of syntax elements according to the respective encryption levels of different GOPs. All syntax elements are encrypted according to their encoding characteristics to ensure format compatibility of the encoded bitstream. Our scheme ensures that both basic and medium security levels maintain the same bit rate while offering high encryption speed, whereas the high security level provides robust protection for highly sensitive GOPs. Finally, security level flags are embedded into syntax element values for decryption purposes, and the cipher-video can be obtained. Experimental results indicate that our HEVC MSASE scheme is robust against all common attacks and provides lower encryption time and bit rate overhead than other state-of-the-art counterparts. It is well suited for real-time traffic surveillance encryption applications.