In order to better maintain the data integrity and help auditors find anomalous reimbursement records in relational databases, the algorithm AAC-Hunter (Aggregation Algebraic Constraints Hunter), which discovered Aggregation Algebraic Constraints (AACs) automatically, was proposed. An AAC is a fuzzy constraint defined between the aggregation results of two columns in the database and acts on most but not all records. Firstly, joining, grouping and algebraic expressions were enumerated to generate candidate AACs. Secondly, the value range sets of these candidate AACs were calculated. Finally, the AAC results were output. However, this method was not able to face the performance challenges caused by massive data, so that a set of heuristic rules were applied to decrease the size of candidate constraint space and the optimization strategies based on intermediate results reuse and trivial candidate AACs elimination were employed to speed up the value range set calculation for candidate AACs. Experimental results on TPC-H and European Soccer datasets show that AAC-Hunter reduces the constraint discovery space by 95.68% and 99.94% respectively, and shortens running time by 96.58% and 92.51% respectively, compared with the baseline algorithm without heuristic rules or optimization strategies. As the effectiveness of AAC-Hunter is verified, it can be seen that AAC-Hunter can improve the efficiency and capability of auditing application.

In view of the fact that the rapid development of quantum computer technology poses serious threat to the security of the traditional public-key cryptosystem, a new authenticated key exchange protocol scheme based on Ring Learning With Errors (RLWE) problem was proposed. By using Peikert error reconciliation mechanism, both parties of communication can directly obtain the shared bit value of the uniform distribution and get the same session key. The encoding bases of lattice was used to analyze the error tolerance, and reasonable parameters were selected to ensure that both parties can get the same session key with significant probability. The security of the protocol was proved in the BR (Bellare-Rogaway) model with weak perfect forward secrecy. The security of the protocol was attributed to the difficult RLWE problem of lattice, so that the protocol can resist quantum attacks. Compared with the existing authenticated key exchange protocols based on RLWE, the size of the parameter value modulus decreases from sub-exponential to polynomial magnitude, thus the corresponding amount of computation and communication are also significantly reduced. The results show that the proposed scheme is a more concise and efficient post quantum authenticated key exchange protocol.

In order to improve the performance of the Particle Swarm Optimization (PSO), an adaptive range PSO with the Cauchy distributed population named ARPSO/C was proposed. The algorithm used the median and scale parameters to adjust self-adaptively the search range in population under the suppose of the individuals obeying the Cauchy distribution, thus balanced between local search and global search. The numerical comparison results on the proposed algorithm, ARPSO and PSO show that the presented algorithm has higher convergence speed and can overcome the prematurity.

As a new media, social network gains a wide range of social influence, and the social network based e-commerce becomes more and more popular, which make the study of information propagation of great significance. A trust based information propagation model was presented in this paper according to the trust relationship between people in daily life. First, the algorithm of data mining was employed on personal information to make the classification of the users. And then the value of trust between the users was calculated according to their relationships. At last, this paper predicted the range of information propagation by using the trust value between the users and messages. Compared with two basic methods, the final experiment shows that the results generated by the model are enhanced by 15% in precision, while the time used decreases more than 50%. The results differ with the statistic results on the dataset at 5%, which shows that the model preforms well in practice.