Abstract: The conventional deep packet inspection technologies usually need to inspect complete payloads of the network packets, and this will bring hidden dangers to the users privacy. Privacy aware deep packet inspection (PaDPI) can protect users privacy while getting higher accuracy of identification. Firstly, this paper studies several privacy levels that deep packet inspection involves, and classifies the privacy of packet payload into different levels according to the depth in payloads. Secondly, this paper proposes several schemes to truncate the payloads of network packets according to the privacy levels. Thirdly, this paper evaluates 12 widely used application protocols with those truncated payloads through PaDPI for application protocol identification. The results show that privacy aware deep packet inspection can gain above 80% accuracy on truncated payloads, and can limit the leakage of users privacy strictly.

Key words: deep packet inspection, privacy, application protocol identification

摘要： 传统深包检测技术需要检测网络包的完整有效载荷，从而给用户隐私安全带来隐患。隐私敏感的深包检测技术在实现较高协议识别率情况下能够有效保护用户隐私。对深包检测技术中涉及的用户隐私进行了研究，根据有效载荷深度对隐私级别进行了划分。设计了根据隐私级别对网络包有效载荷进行截断预处理的方法。针对12种常用协议，使用隐私敏感的深包检测技术进行协议识别，结果表明，隐私敏感的深包检测技术可以获得80%以上的准确率，并严格地限制了用户隐私的泄露。

关键词: 深包检测技术, 隐私, 应用层协议识别