Enhancing the Anonymity in Information Diffusion Based on Obfuscated Coded Data
Enhancing the Anonymity in Information Diffusion Based on Obfuscated Coded Data
复制标题
DOI：
10.1109/tnse.2018.2888848
复制
发表时间：
2019-10
影响因子：
6.6
通讯作者：
Jin Wang;K. Lu;Jianping Wang;Chuan Wu;Naijie Gu
中科院分区：
文献类型：
--
作者：
Jin Wang;K. Lu;Jianping Wang;Chuan Wu;Naijie Gu
线性网络编码（LNC）是一种很有前途的方法，以促进匿名的信息扩散，因为每个数据包是由多个传入的数据包线性组合。由于线性组合中使用的系数将揭示节点处传入和传出数据包之间的相关性，因此大多数匿名LNC设计的现有研究都集中在加密这些系数上。尽管这些研究的重要性，编码内容的相关性仍然可以被分析，未加密的LNC的潜力还没有得到充分利用。在本文中，我们解决了这些问题，我们提出了一种新的ALNCode计划，可以提高匿名性，通过生成传出的数据包，相关的传入编码数据包的多个流。通过坚实的理论分析，我们首先证明了来自不同流的输入编码分组是相关的概率。然后，我们证明，如果这种相关性存在，我们可以设计确定性LNC混淆数据包的相关性。在相同的条件下，我们还证明了随机生成的编码数据包与其他流中的编码数据包相关的概率。除了理论研究，我们进行了大量的数值实验，以了解各种编码参数和ALNCode的性能在真实的场景的影响。
Linear network coding (LNC) is a promising approach to facilitate anonymity in information diffusion because each packet is generated by linearly combining multiple incoming packets. Since the coefficients used in the linear combination would reveal the correlation between incoming and outgoing packets at a node, most existing studies on anonymous LNC design focus on encrypting these coefficients. Despite the importance of these studies, the correlation of coded content can still be analyzed and the potential of un-encrypted LNC has not been fully exploited. In this paper, we tackle these issues and we propose a novel ALNCode scheme that can enhance anonymity by generating outgoing packets that are correlated to incoming coded packets of multiple flows. With solid theoretical analysis, we first prove the probability that incoming coded packets from different flows are correlated. Then, we prove that, if such correlation exists, we can design deterministic LNC to obfuscate the correlation of packets. With the same condition, we also prove the probability that a randomly generated coded packet is correlated to coded packets in other flows. Besides the theoretical study, we conduct extensive numerical experiments to understand the impacts of various coding parameters and the performance of ALNCode in real scenarios.