Biblio

Social media have become increasingly popular for the sharing and spreading of user-generated content due to their easy access, fast dissemination, and low cost. Meanwhile, social media also enable the wide propagation of cyber frauds, which leverage fake information sources to reach an ulterior goal. The prevalence of untrustworthy information sources on social media can have significant negative societal effects. In a trustworthy social media system, trust calculation technology has become a key demand for the identification of information sources. Trust, as one of the most complex concepts in network communities, has multi-criteria properties. However, the existing work only focuses on single trust factor, and does not consider the complexity of trust relationships in social computing completely. In this paper, a multi-criteria trustworthiness calculation mechanism called Info-Trust is proposed for information sources, in which identity-based trust, behavior-based trust, relation-based trust, and feedback-based trust factors are incorporated to present an accuracy-enhanced full view of trustworthiness evaluation of information sources. More importantly, the weights of these factors are dynamically assigned by the ordered weighted averaging and weighted moving average (OWA-WMA) combination algorithm. This mechanism surpasses the limitations of existing approaches in which the weights are assigned subjectively. The experimental results based on the real-world datasets from Sina Weibo demonstrate that the proposed mechanism achieves greater accuracy and adaptability in trustworthiness identification of the network information.

The disclosure of an important yet sensitive link may cause serious privacy crisis between two users of a social graph. Only deleting the sensitive link referred to as a target link which is often the attacked target of adversaries is not enough, because the adversarial link prediction can deeply forecast the existence of the missing target link. Thus, to defend some specific adversarial link prediction, a budget limited number of other non-target links should be optimally removed. We first propose a path-based dissimilarity function as the optimizing objective and prove that the greedy link deletion to preserve target link privacy referred to as the GLD2Privacy which has monotonicity and submodularity properties can achieve a near optimal solution. However, emulating all length limited paths between any pair of nodes for GLD2Privacy mechanism is impossible in large scale social graphs. Secondly, we propose a Walk2Privacy mechanism that uses self-avoiding random walk which can efficiently run in large scale graphs to sample the paths of given lengths between the two ends of any missing target link, and based on the sampled paths we select the alternative non-target links being deleted for privacy purpose. Finally, we compose experiments to demonstrate that the Walk2Privacy algorithm can remarkably reduce the time consumption and achieve a very near solution that is achieved by the GLD2Privacy.

Artificial Intelligence (AI) has attracted a large amount of attention in recent years. However, several new problems, such as privacy violations, security issues, or effectiveness, have been emerging. Differential privacy has several attractive properties that make it quite valuable for AI, such as privacy preservation, security, randomization, composition, and stability. Therefore, this paper presents differential privacy mechanisms for multi-agent systems, reinforcement learning, and knowledge transfer based on those properties, which proves that current AI can benefit from differential privacy mechanisms. In addition, the previous usage of differential privacy mechanisms in private machine learning, distributed machine learning, and fairness in models is discussed, bringing several possible avenues to use differential privacy mechanisms in AI. The purpose of this paper is to deliver the initial idea of how to integrate AI with differential privacy mechanisms and to explore more possibilities to improve AIs performance.

Multimedia data available in various disciplines are usually heterogeneous, containing representations in multi-views, where the cross-modal search techniques become necessary and useful. It is a challenging problem due to the heterogeneity of data with multiple modalities, multi-views in each modality and the diverse data categories. In this paper, we propose a novel multi-view cross-modal hashing method named Multi-view Collective Tensor Decomposition (MCTD) to fuse these data effectively, which can exploit the complementary feature extracted from multi-modality multi-view while simultaneously discovering multiple separated subspaces by leveraging the data categories as supervision information. Our contributions are summarized as follows: 1) we exploit tensor modeling to get better representation of the complementary features and redefine a latent representation space; 2) a block-diagonal loss is proposed to explicitly pursue a more discriminative latent tensor space by exploring supervision information; 3) we propose a new feature projection method to characterize the data and to generate the latent representation for incoming new queries. An optimization algorithm is proposed to solve the objective function designed for MCTD, which works under an iterative updating procedure. Experimental results prove the state-of-the-art precision of MCTD compared with competing methods.

In big data applications such as healthcare data mining, due to privacy concerns, it is necessary to collect predictions from multiple information sources for the same instance, with raw features being discarded or withheld when aggregating multiple predictions. Besides, crowd-sourced labels need to be aggregated to estimate the ground truth of the data. Due to the imperfection caused by predictive models or human crowdsourcing workers, noisy and conflicting information is ubiquitous and inevitable. Although state-of-the-art aggregation methods have been proposed to handle label spaces with flat structures, as the label space is becoming more and more complicated, aggregation under a label hierarchical structure becomes necessary but has been largely ignored. These label hierarchies can be quite informative as they are usually created by domain experts to make sense of highly complex label correlations such as protein functionality interactions or disease relationships. We propose a novel multi-source hierarchical prediction consolidation method to effectively exploits the complicated hierarchical label structures to resolve the noisy and conflicting information that inherently originates from multiple imperfect sources. We formulate the problem as an optimization problem with a closed-form solution. The consolidation result is inferred in a totally unsupervised, iterative fashion. Experimental results on both synthetic and real-world data sets show the effectiveness of the proposed method over existing alternatives.