Data Security and Privacy Protection
Data security is bound to become a key research and development direction of all sectors of society. As far as concerned, the ability of data leakage prevention and traceability under the Internet sharing environment is still relatively weak, and data leakage has been happening all the time. "Insider" often has a variety of ways and means to disclose high-risk data, especially unstructured data leakage such as screen recapture, screenshots and screen snapping.
We define unstructured data as data with irregular or incomplete data structure and no predefined data model, and this kind of data is difficult to be represented by the two-dimensional logical table of the database. It includes all forms of office documents, text, pictures, XML, HTML, various reports, images and audio / video information. Among the risk points of different types of unstructured data mentioned above, the most common scenario is to disclose high-risk content through screen recapture and snapping. Since everyone has mobile-phone, the high fidelity format of high-risk and sensitive information data can be obtained in a very short time through screenshots or screen recapture. Although some digital watermarking technologies can be used to track and trace the source of information leakage, the scanning ability of the current security infrastructure for unstructured data content is still insufficient, which may lead to some high-risk information can be transmitted externally by using image, audio and other carriers to penetrate the current protection ability of data security. Especially for the leaker photographers with special disclosure purpose, the shooting action is usually relatively secret. After the leaker shooting, it is likely to carry out a series of attacks on the image and erase the watermark marks contained in the high-risk content. Or further, the candid photographer is likely to disguise the images taken and recorded on the screen by means of image fusion and information hiding, so as to realize the behavior of secret transmission. At present, many enterprises allow arbitrary transmission and copying of photos and images, and there are obvious "security vulnerabilities" in image unstructured data.
Unstructured data leakage such as recapture and recording will have a very serious and bad negative impact on society, enterprises and individuals. Detection of mobile phone image/video recapture is an important topic related to data security. By studying efficient and accurate real-time detection of screen recapture data leakage, we can plug the data security loopholes in this aspect in the society to a great extent.
This project proposes to conduct an in-depth research on two technical problems:
1）Efficient and accurate screen recapture detection technology. It can resist composite post-processing attacks including re-compression, cropping, scaling, multiple social media transmission and so on. 高效准确的屏幕翻拍、翻录检测技术。能够抵抗包含重压缩、裁剪、缩放、多次社交媒体传输等各种强度的复合后处理攻击。
2）Effectively combine the data content and recapture behavior to judge whether the recapture behavior is suspicious. When the recapture behavior is detected, the content of the recapture images and the behavior before and after the shooting are actively detected and analyzed through multi-modal machine learning technology and OCR technology, so as to effectively judge whether the recapture and the recapture behavior involve data leakage.
Related Research Topics
- Efficient and accurate screen recapture detection 高效、准确的屏幕 翻拍、翻录检测手段
- Analysis on the screen recapture behavior and leaked data content, effectively detect data leakage 结合数据内容、拍摄行为，有效检测数据泄漏的发生
- To reveal the post-processing trace in the process of screen recapture 揭示屏幕翻拍翻录过程中的后处理痕迹