With the development of the deep learning technology, automatic lipreading based on deep neural network can achieve reliable results for speakers appeared in the training dataset. However, speaker-independent lipreading, i.e. lipreading for unseen speakers, is still a challenging task, especially when the training samples are quite limited. To improve the recognition performance in the speaker-independent scenario, a new deep neural network structure, named Disentangled Visual Speech Recognition Network (DVSR-Net), is proposed in this paper. DVSR-Net is designed to disentangle the identity-related features and the content-related features from the lip image sequence. To further eliminate the identity information that remained in the content features, a content feature refinement stage is designed in network optimization. By this way, the extracted features are closely related to the content information and irrelevant to the various talking style and thus the speech recognition performance for unseen speakers can be improved. Experiments on two widely used datasets have demonstrated the effectiveness of the proposed network in the speaker-independent scenario.

Compact E-cash achieves an efficient system by withdrawing 2n coins within O(1) operations and storing the coins in O(n) bits. For preventing a double-spender from cheating again, it is necessary to trace his e-coins. So full-tracing in compact E-cash system means tracing double-spender and tracing his coins. However, the efficiency problem caused by coin-tracing without TTP (trusted third party) has not been solved. For solving this problem, we introduce a non-standard construction into zero-knowledge proof of payment protocol, which leaks coin information when double-spending but is proven to be perfect zero-knowledge to verifier when spending a coin only once. Therefore, it achieves tracing dishonest users' coins and preserving the anonymity of honest users. Comparing with the existing most efficient method of coin-tracing without TTP, we improve computational complexity from O(k) to O(1) with less storage space. In addition, to improve efficiency and practicality further, batch-spending (spending any number of coins in one operation) and compact-spending (spending all coins in one operation) had been proposed. Based on the non-standard zero-knowledge proof, our scheme provides more efficient batch/compact-spending. Moreover, we also make a comparison with Bitcoin and Bitcoin Lightning Network, which have attracted considerable attention.

Extensive online knowledge sharing can be exploited to solve tasks better, faster and cheaper, and it garners considerable interest in institutional cooperation, learning communities, etc., while many technical problems, such as fair exchange in incentive design and security issues are waiting to be solved. Blockchain technique has high accountability and thus has potential to improve the transparency and security of knowledge sharing. In this paper, to address the above problems, first we propose a Reputation Based Knowledge Sharing system in blockchain, called RBKS. The aim of RBKS is to exploit the copyright protection of the knowledge owner using our proposed fine-grained access control system, and to achieve the paid-for content service which allows bystanders who are interested in the shared knowledge to pay a small fee for the access. Second, a reputation evaluation algorithm is introduced as the core of the incentive, and it may possibly be combined with stake in blockchain to form a hybrid stake for the RBKS system. Third, a blockchain and a trusted storage server are employed together for sharing and storing knowledge, and the main procedures are implemented with smart contract in blockchain to ensure secure execution and fairness. Finally, our analysis shows that the RBKS system is feasible and secure.

In a periodic K-times anonymous authentication system, user can anonymously show credential at most K times in one time period. In the next time period, user can automatically get another K-times authentication permission. If a user tries to show credential beyond K times in one time period, anyone can identify the dishonest user (the violator). But identifying violators is not enough for some systems, where it is also desirable to revoke violators' credentials for preventing them from abusing the anonymous property again. However, the problem of revoking credential without trusted third party has not been solved efficiently and practically. To solve it, we present an efficient scheme with efficient revocation of violator's credential. In fact, our method also solves an interesting problem-leaking information in a statistic zero-knowledge way, so our solution to the revocation problem outperforms all prior solutions. For achieving it, we use the special zero-knowledge proof with special information leak for revoking the violator's credential, but it can still be proven to be perfect statistic zero knowledge for guaranteeing the honest user's anonymity. Comparing with existing schemes, our scheme is efficient, and moreover, our method of revoking violator's credential is more practical with the least additional costs.

Anonymity is difficult to achieve in fully distributed Peer-to-Peer (P2P) reputation systems, because reputation systems often use identity to evaluate each peer's trustworthiness. In this paper, we propose a dynamic anonymous P2P reputation system (DARep) based on Trusted Computing (TC) technology to ensure all the peers' anonymity in a fully distributed structured manner. With the TC hardware, i.e., Trusted Platform Module (TPM), installed in each peer's computer, peers' anonymity is achieved by changing their pseudonyms. The simulation results show that DARep can be implemented in an efficient way and achieve considerably fine performance in terms of unlinkable rate, message overhead, time cost, final reputation value (FRV) computing error, and service selection success rate.