隐形像素已死,隐形像素万岁!
Invisible Pixels Are Dead, Long Live Invisible Pixels!
摘要:
自20世纪90年代以来,万维网的隐私状况一直在恶化。不同第三方对浏览习惯的跟踪是这种恶化的核心。网络cookie和所谓的网络信标是实现第三方跟踪的经典方法。由于采用了更复杂的技术跟踪解决方案和其他根本性的转变,使用传统的基于图像的网络信标可能会失去吸引力。本文通过对3万多张流行网站图片的抽样分析,证明了这种假设是错误的:经典的1 x 1图片在当代万维网中仍然被广泛用于第三方跟踪。虽然广告拦截器似乎无法完全阻止这些经典的基于图像的跟踪信标,但该论文进一步证明,即使是有限的信息也可以用于准确地将第三方1 x 1图像与其他图像进行分类。实验结果表明,该方法的平均分类精度为0.956.有了这些结果的文件有助于正在进行的尝试,以更好地了解在万维网上缺乏隐私,并通过这种情况可能最终得到改善的手段。
英文摘要:
Privacy has deteriorated in the world wide web ever since the 1990s. The tracking of browsing habits by different third-parties has been at the center of this deterioration. Web cookies and so-called web beacons have been the classical ways to implement third-party tracking. Due to the introduction of more sophisticated technical tracking solutions and other fundamental transformations, the use of classical image-based web beacons might be expected to have lost their appeal. According to a sample of over thirty thousand images collected from popular websites, this paper shows that such an assumption is a fallacy: classical 1 x 1 images are still commonly used for third-party tracking in the contemporary world wide web. While it seems that ad-blockers are unable to fully block these classical image-based tracking beacons, the paper further demonstrates that even limited information can be used to accurately classify the third-party 1 x 1 images from other images. An average classification accuracy of 0.956 is reached in the empirical experiment. With these results the paper contributes to the ongoing attempts to better understand the lack of privacy in the world wide web, and the means by which the situation might be eventually improved.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。