博士 / 國立中山大學 / 資訊工程學系研究所 / 104 / Social network services (SNS) have become a major internet service for people to communicate with each other. It is full of complex relationships among people in the real-life and virtual world. Different social networks have different characteristics and varying levels of influence. To understand the message propagation process, the driving power behind it and its social influence, this paper presents a detailed analysis of message propagation models over the social networks by analyzing the relationships among nodes. This paper presents four proposed models which aim to analyze message propagation on social networks. We analyze the message propagation models and show how messages spread through the social networks. We also analyze and describe the significance and influence of clustering parameters in the model on the social clusters. Furthermore, we propose an social network analysis on Hadoop platform and implement a prototype to verify the social network characteristics. Hadoop platform can process massive data in a parallel manner on a large cluster built by commodity hardware. We also present a measurement study of messages collected from 900K users on Facebook, to verify our proposed models by means of big-data Hadoop platform. We believe that our research provides valuable insights for future social network service research.

博士 / 國立中山大學 / 資訊工程學系研究所 / 104 / In recent years, XML (eXtensible Markup Language) has become the standard code for data representation and data exchange on the Internet. In addition, data queries have constantly increased; however, it is getting harder to query efficiently and obtain the precise required results because of the huge amount of data. The main operation in XML query processing is finding nodes that match the given query tree pattern (QTP) in the document. The problem is that accessing too many useless nodes in order to match a query pattern is very time consuming. Meanwhile, the XML documents based on characteristics of structure can lead to results which lack clarity due to the query ambiguity of the structure. Therefore, determining how to ensure efficient query service based on a skillful representation that can support query diversification and solve ambiguity in order to achieve high precision search capability is an important issue. To overcome the time-consumption problem, we utilize the structural summary tree (SST) algorithm to optimize XML documents; the aim is to eliminate unnecessary paths that include nested structures and duplicate paths. The novel twig-join Swift (TJSwift) associated with adjacent linked (AL) lists for the provision of efficient XML query services is proposed herein, whereby queries can be versatile in terms of predicates. It can completely preserve hierarchical information, and the new index generated from SST is used to save semantic information in order to provide template-based indexing for fast data searches. At the same time, to the best of our knowledge, researches on query diversification, queries in single node and hierarchical level difference and intermediate nodes with ambiguity in regard to hierarchical level are insufficient. In terms of result relevance, effectiveness is the most crucial aspect of query search, which can be summarized as these issues. Hence, we also further propose extending twig join Swift (eXTJSwift) associated with AL lists to provide efficient XML query services, whereby queries can be versatile in terms of predicates. In order to evaluate the performance of the TJSwift and eXTJSwift approaches with that of TwigStack, TwigList and TJFast, we conducted two sets of performance evaluation. For the TJSwift, the performance evaluations were conducted in terms of total execution time, scalability and number of elements read, which indicates how many nodes must be read in a matching process. On the other hand, eXTJSwift is in terms of total execution time and number of paths matched and further to add various query criteria to compare the precise simulation in query diversification, target hierarchical level and the problem of ambiguity. Experiment of TJSwift results show that not only are these algorithms able to satisfy a query, but also has better time-saving efficiency compared with the existing twig-join algorithms such as the TJFast algorithm. Similarly, eXTJSwift achieved better accuracy than other approaches in terms of query diversification, target hierarchical level and the problem of ambiguity.

碩士 / 國立中山大學 / 資訊工程學系研究所 / 104 / Ciphertext-policy attribute-based encryption (CP-ABE) is an access control mechanism where a data provider encrypts a secret message and then sends the ciphertext to the receivers according to the access policy which she/he decides. If the attributes of the receivers match the access policy, then they can decrypt the ciphertext. This thesis shows a relation between ABE and identity-based encryption (IBE), and presents a bi-directional conversion between an access structure and identities. By the proposed conversion, the ABE scheme constructed from an IBE scheme will inherit the features, such as constant-size ciphertexts and anonymity, from the IBE scheme, and vice versa. It turns out that the proposed conversion also gives the first ABE achieving access structures with wildcard and constant-size ciphertexts/private keys. Finally, we prove the CCA security for confidentiality and anonymity.

碩士 / 國立中山大學 / 資訊工程學系研究所 / 104 / Smart grids have become a future trend due to the development of technology and increased energy demand and consumption. In smart grids, users’ electricity consumption is recorded by their smart meters, and the smart meters submit the data to the operation center in each time unit for monitoring. The operation center analyzes the data it receives to estimate users’ electricity usage in the next time unit and to ensure dynamic energy distribution. Compared to traditional grids, the electricity can be flexibly controlled, and waste is decreased in smart grids. However, details of users’ daily lives may be leaked out through the frequent monitoring of users’ electricity usage, which causes the problem of privacy preserving. To solve the problem, data aggregation mechanisms are adopted in this environment. The power usage data in the same units are aggregated before being sent to the operation center. This aggregation prevents personal electricity usage data from being shared with the operation center. Thus, users’ privacy is protected. Along with the increase in the number of research studies on smart grids, many studies on the privacy-preserving issues of power usage have been published. However, both power usage data and electricity payment data may jeopardize users’ privacy. The operation center is able to obtain users’ private information by analyzing users’ electricity payments. Therefore, we propose a verifiable privacy-preserving payment mechanism for smart grids. In our scheme, users can submit electricity payments without revealing any private information, and the operation center can verify the correctness of the payment.

碩士 / 國立中山大學 / 資訊工程學系研究所 / 104 / Lattice is widely used in cryptography since it has potential for defending quantum attacks. One of the significant problems in such cryptography is the shortest vector problem (SVP). This problem is to find the non-zero shortest vector in lattice. SVP is an NP-hard problem proven by Ajtai, and many cryptosystems are secure under the assumption that SVP is hard, such as NTRU. On the other hand, some primitives of lattice-based cryptography require relatively short vectors. In this thesis, we propose a new SVP algorithm which can be performed in time complexity O(n^3). We also prove that the Hermite factor of the proposed algorithm is polynomial-bounded.

博士 / 國立中山大學 / 資訊工程學系研究所 / 104 / To enable the immediate and efficient dispatch of relief to victims of disaster, this thesis proposes a greedy-search-based multi-objective genetic algorithm (GSMOGA) that is capable of regulating the distribution of available resources and automatically generating a variety of feasible emergency logistics schedules for decision-makers. The proposed algorithm merges the features of local search ability of the greedy method and the diversity of multi-objective genetic algorithm to enhance local search speed and diversity explore ability. It uses the Google Map to draw up the available roads which connect the demand points and supply points and applies the Dijkstra algorithm to find the shortest path between each demand point and supply point. It also dynamically adjust distribution schedules from various supply points according to the requirements at demand points, and adopts the NSGAII method to perform rank & sort procedure to find the feasible solution schedules on non-dominated Pareto front in order to minimize the following: unsatisfied demand for resources, time to delivery, and transportation costs. The sequence of three objectives are also applied to be the priority sequence to generate and order routing schedules for the decision maker. The algorithm uses the case of the Chi-Chi earthquake in Taiwan to verify its performance. Simulation results demonstrate that with a limited and unlimited number of available vehicles, the proposed algorithm outperforms the multi-objective genetic algorithm (MOGA) and the standard greedy algorithms in ‘time to delivery’ by 56.16% and 64.11%, respectively under the 10,000 generations and average situation. The final routing figures show that the GSMOGA is more comprehensive in the emergency logistics scheduling problem. We study the effect of different crossover methods on the performance of GSMOGA. The results show that order based crossover performs the best. We verify the correctness of GSMOGA by comparing the result using the brute force approach.

博士 / 國立中山大學 / 資訊工程學系研究所 / 104 / Since high mobility of vehicle and variable network topology are challenges in Vehicular Ad Hoc Networks (VANETs), routing protocols in Mobile Ad Hoc Network (MANET) are not suitable and need new strategy in VANETs. Geographic routing is beneficial to deal with these problems by collecting geographic information (such as vehicle’s position, velocity, direction, etc.) of vehicles to assist in establishing the stable routing paths. Therefore, in this dissertation we propose a Traffic-Aware routing strategy with Link Quality (TA-LQ) based on geographic routing protocol. The goal is to achieve high delivery ratio and low end-to-end delay. In road segment, we utilize the link quality between neighbors to select available relays for transmission. In the intersection, by collecting real-time traffic flow information, we derive a connectivity probability function to select road intersections efficiently for transmission and to reduce the path’s sensitivity to vehicle movements. Moreover, integrating the recovery and congestion aware mechanism, the high delivery ratio and low end-to-end delay can be achieved. The simulation results show that our mechanism has significant performance improvement in comparison with other existing geographic routing protocols.

碩士 / 國立中山大學 / 資訊工程學系研究所 / 104 / In this thesis, a multimedia conference system is proposed to support the high quality multimedia processing. This system adopts the client-server architecture and consists of three conference modules: simultaneous interpretation, multi-channel audio mixing and video streaming respectively. The goal of this thesis is to implement the proposed conference modules with a software solution. To perverse the performance of hardware solution, the multimedia data is encoded and decoded by Opus audio codec and H.264/AVC image codec. And the audio mixer uses Asymmetrical Wave-shrinking mixing algorithm to reduce the complexity. With the rigorous field trial, the results show the proposed system can be applied on a general conference to provide a good user experience.

碩士 / 國立中山大學 / 資訊管理學系研究所 / 104 / With the expending of Internet popularity, the competition between Internet Service Provider companies is achieving a peak. Because their products and services have highly similarity and also have close ranged pricing strategy. Hence, once ISP disappointed their customers, they will easily lose their market share. The purpose of this research is to understand the effect on different switch driver on customer’s emotion when they performed switching behavior. Second, after customer switched, what behavior can be influenced by customer’s emotion? This research combined Push-Pull-Mooring model and Expectation Disconfirmation Theory as drivers and post-switching problems. Moreover, the concept of Regret is used to investigate and explains the effect on customer’s behavior intention. The results of this research include: (1) Customers driven by Pull effects will cause higher Regret than driven by Push effects. (2) After switched, customers will sensed more Regret when negative consequences occurred. (3) Regret emotion created by post-switching behavior can affect customer’s negative behavior intention. The contribution of this research is to understand the causes of regret and the influences on customer’s post-switching behavior.

碩士 / 國立中山大學 / 資訊管理學系研究所 / 104 / Recently, the explosive development of mobile device has dramatically changed human life, mobile application becomes pervasive as well. Nowadays, there have released about 3 million mobile applications. Due to the tremendous and still increasing number of mobile application, user get harder to find needed apps. To tackle this problem, we propose a personalized recommender system based on the features of textual data. Specifically, we apply LDA to extract hidden topics of user reviews and use the probability of topic distribution to represent the features of app. Further, construct user profile based on his or her consumed apps in order to realize user requirements and preference. Eventually, for each app, we take account both the topic distribution and user preference to estimate a recommended score for target user and sort candidate apps by descending score to come out a personalized app recommended list. For the evaluation, we crawl the real-world dataset and adopt Recall@M as the measurement of performance. The experimental result shows our proposed mechanism outperforms the baseline and is able to enhance the performance of the state-of-the-art recommender systems. We then concluded that the user feedback is an effective variable to represent the features of app and plays significant role on app personalized recommender system.