Abstract The rise of the Internet of Things (IoT) has led to a huge amount of data beginning to emerge. Federated learning (FL) has received widespread attention and application as a new paradigm for data collection. However, data trading poses a threat to the privacy of data owners, and even participants in federated learning face the risk of data breaches. While many encryption methods have been proposed to mitigate these risks, the encrypted data negatively impacts the quality of the global model in federated learning. To this end, we propose an algorithm based on contract mechanisms to resolve the conflict between the privacy protection level of clients and the aggregation error on the federated learning server. Clients upload perturbed data according to their privacy protection levels, while mitigating the conflict between client data privacy protection and platform global model aggregation error. Through theoretical analysis and extensive experiments, our proposed trading method achieves desirable data utility while ensuring budget feasibility, individual rationality, and incentive compatibility.

Background information is very important for news reading and understanding, especially with today’s rapid new media development for news production and consumption. When researchers actively develop tools to automatically provide back-ground information for people, we still have little understanding of how people, from readers’ perspective, engage with news and what background information is needed. To address such a gap, we conducted a study with a news reading platform, observed participants’ news reading behaviors using the platform and interviewed them over the process. Our study highlighted that news reading is a learning activity between leisure and study, and their background information searching behavior is part of Gestalt knowing – to make their existing knowledge more complete, by deepening their prior understandings, broadening their horizons, filling the gaps, clearing their confusion, as well as tracking the progress of an event as a whole, so a more holistic understanding could be achieved. We end by discussing implications of our findings.

In the paper we focus on a promising research problem of data trading, under the scenario that data items can be reproduced easily and inexpensively. Apart from a Bayesian optimal mechanism based data trading approach, we also propose a prior-free data trading approach to organize the data trading process between selfish data owners and share-averse data consumers, with the goal of maximizing revenue of data owners and meanwhile determining the optimal number of data copies. Rigorous theoretical analysis and extensive experiment results are offered to verify the effectiveness of proposed methods in terms of sum of valuations, revenue, individual rationality and incentive compatibility.

In the era of big data, data is an extremely important and valuable asset. Data trading is significantly essential to unlocking the power of AI/ML and gaining competitive advantages. Different from a general concept of data sharing, data trading makes it possible for both data consumers to specify what data they want and data owners to determine whether or not as well as how to trade their data. Existing studies ignore a few important characteristics, e.g., difficulty on accurate valuation for every data item, correlated valuation for data items, and privacy concern on disclosing valuation information. To this end, we propose an iterative combinatorial auction based data trading approach (ICADT) to uncover the value of correlated mobile crowdsensing data on the fly without direct elicitation of private valuation information. ICADT is an efficient ascending price-based auction mechanism which proceeds in rounds and determines allocation of data items and trading prices iteratively. Both rigorous theoretical analysis and extensive simulations demonstrate good properties of ICADT, e.g., convergence, individual rationality, optimality on system efficiency.

Mobile crowd sensing (MCS) systems fully take advantage of wisdom of the crowd and benefit from low deployment cost and widely spatial coverage. We propose a practical decentralized MCS system based on a distributed auction process and the blockchain system, achieving the optimal social efficiency and individual rationality without disclosing privacy.

Nowadays most metro advertising systems schedule advertising slots on digital advertising screens to achieve the maximum exposure to passengers by exploring passenger demand models. However, our empirical results show that these passenger demand models experience uncertainty at fine temporal granularity (e.g., per min). As a result, for fine-grained advertisements (shorter than one minute), a scheduling based on these demand models cannot achieve the maximum advertisement exposure. To address this issue, we propose an online advertising approach, called FineUDM, based on the uncertain passenger demand modeling for both entering passengers and exiting passengers. FineUDM combines coarse-grained statistical demand modeling and fine-grained real-time demand modeling by leveraging historical passenger demands, real-time card-swiping records, and passenger mobility patterns. Based on this uncertain demand model, it schedules advertising time online based on robust receding horizon control to maximize the advertisement exposure. We evaluate the proposed approach based on an one-month sample from our 530 GB real-world metro fare dataset with 16 million cards. The results show that our approach provides a 61.5 percent lower traffic prediction error and 20 percent improvement on advertising efficiency on average.

Mobile crowd sensing, also called crowd sensing over smartphones, has been an appealing paradigm for collecting sensory data over a vast urban area, due to advantages of low deployment cost and widely spatial coverage of geographically distributed smartphones or other smart devices. In the paper, we focus on a nontrivial problem of making an agreement between crowdsourcers and smartphone users to find the most efficient assignment of sensing tasks to smartphone users. However, there exist several technical challenges such as the incentive issue to encourage participation of smartphone users, preserving private information and distributed implementation in practice. Existing approaches usually have several limitations, e.g. the absence of proper incentives, the assumption of a centralized auctioneer or platform. To this end, we propose a distributed auction framework that explicitly models the interaction between crowdsourcers and smartphone users, achieving the optimal social profit and providing proper incentives to entities without disclosing their privacy as well. We demonstrate that the proposed distributed auction algorithm satisfies a lot of good properties, including optimality of social profit, computation efficiency, convergence, individual rationality through both solid theoretical analysis and extensive experiments.

Understanding air travel choice behavior of air passengers is of great significance for various purposes such as travel demand prediction and trip recommendation. Existing approaches based on surveys can only provide aggregate level air travel choice behavior of passengers and they fail to provide comprehensive information for personalized services. In this paper we focus on modeling individual level air travel choice behavior of passengers, which is valuable for recommendations and personalized services. We employ a probabilistic model to represent individual level air travel choice behavior based on a large dataset of historical booking records, leveraging several key factors, such as takeoff time, arrival time, elapsed time between reservation and takeoff, price, and seat class. However, each passenger has only a limited number of historical booking records, causing a serious data sparsity problem. To this end, we propose a mixed kernel density estimation (mix-KDE) approach for each passenger with a mixture model that combines probabilistic estimation of both regularity of the individual himself and social conformity of similar passengers. The proposed model is trained and evaluated via the expectation-maximization (EM) algorithm with a huge dataset of booking records of over 10 million air passengers from a popular online travel agency in China. Experimental results demonstrate that our mix-KDE approach outperforms the Gaussian mixture model (GMM) and the simple kernel density estimation in the presence of the sparsity issue.

In this paper, we tackle the problem of stimulating smartphone users to join mobile crowdsourcing applications with smartphones. Different from existing work of mechanism design, we uniquely take into consideration the crucial dimension of location information when assigning sensing tasks to smartphones. However, the location awareness largely increases the theoretical and computational complexity. In this paper, we introduce a reverse auction framework to model the interactions between the platform and the smartphones. We rigorously prove that optimally determining the winning bids is NP hard. In this paper we design a mechanism called TRAC which consists of two main components. The first component is a near-optimal approximate algorithm for determining the winning bids with polynomial-time computation complexity, which approximates the optimal solution within a factor of 1 + ln(n), where n is the maximum number of sensing tasks that a smartphone can accommodate. The second component is a critical payment scheme which, despite the approximation of determining winning bids, guarantees that submitted bids of smartphones reflect their real costs of performing sensing tasks. Through both rigid theoretical analysis and extensive simulations, we demonstrate that the proposed mechanism achieves truthfulness, individual rationality and high computation efficiency.

Participatory sensing has gained increasing attention in recent years as it is promising for empowering large-scale monitoring and knowledge discovery, leveraging a crowd of public smartphone workers. In a participatory sensing system, the platform recruits workers to participate in multiple sensing processes. However, sensing processes have limited budgets. A smartphone is only satisfied when it receives a reward large enough to cover its cost caused, e.g., by resource consumption. Considering the rationality of smartphone workers, it is of great importance for the participatory sensing system to satisfy as many workers as possible, in order to stimulate more smartphone participation. In this paper, we propose a general sensing processes participation game framework with heterogenous workers and heterogenous sensing processes. We show that it is NP-hard to find a sensing processes participation solution which maximizes the number of satisfied workers. Inspired by the finite improvement property of the game, we design and implement an algorithm of sensing processes participation, which guarantees to reach a pure Nash equilibrium in polynomial time, and allows workers to change their strategy profiles asynchronously. Simulation results show that our algorithm is effective and efficient.

Vehicular networks have become a promising platform for large-scale urban sensing. On-demand data retrieval is a crucial operation for many applications of vehicular networks. It is particularly challenging, however, to achieve high performance of on-demand data retrieval in sparse vehicular networks. Data replication based on random linear network coding can solve the coupon collection problem but suffers the problem of unnecessary data replications which waste the precious communication opportunities of sparse vehicular networks. With real traces we reveal the existence of community structures in vehicular networks, which causes excessive linearly correlated blocks. Motivated by the important observation, we propose a community-aware data replication scheme based on random linear network coding. To reduce the excessive correlated coded blocks because of community structures, we make probabilistic control on the data replication phase, taking the community structures into account. It is demonstrated through extensive simulations that the community-aware scheme can effectively save up to 50% communication opportunities whiling achieving high performance of on-demand data retrieval.

Retrieving data from mobile source vehicles is a crucial routine operation for a wide spectrum of applications of vehicular networks, such as road surface monitoring and sharing. The key to improving the data retrieval performance is to increase the data availability such that a retrieving node can easily acquire its desired data from the vehicles that it contacts. Network coding has widely been exploited as an effective technique for defusing information over a network. In this paper we explore the exploitation of network coding for improving data availability in vehicular networks. By random linear codes, simple replication is avoided, and instead a node forwards a coded block which is a random combination of all data received by the node. We implement a network coding based approach for improving data availability in vehicular networks. To understand the performance of this approach, we have conducted empirical study with extensive simulations based on real vehicular GPS traces from around 2,000 taxis in Shanghai, China. We make the important observation that in spite of significant improvement up to 300% in data availability, there is a serious issue with linear correlation among the received codes, which degrades the performance of data retrieval success rate. By analyzing the real vehicular traces, we reveal that there is strong community structure with a real vehicular network and then explain that such community structure may contribute to the issue of linear dependence. Then, we point out the opportunities for improving the network coding based approach by developing community aware codes distribution techniques.