Viet Hung Vu

With recent advancements in graph neural networks (GNN), GNN-based recommender systems (gRS) have achieved remarkable success in the past few years. Despite this success, existing research reveals that gRSs are still vulnerable to poison attacks , in which the attackers inject fake data to manipulate recommendation results as they desire. This might be due to the fact that existing poison attacks (and countermeasures) are either model-agnostic or specifically designed for traditional recommender algorithms (e.g., neighborhood-based, matrix-factorization-based, or deep-learning-based RSs) that are not gRS. As gRSs are widely adopted in the industry, the problem of how to design poison attacks for gRSs has become a need for robust user experience. Herein, we focus on the use of poison attacks to manipulate item promotion in gRSs. Compared to standard GNNs, attacking gRSs is more challenging due to the heterogeneity of network structure and the entanglement between users and items. To overcome such challenges, we propose GSPAttack —a generative surrogate-based poison attack framework for gRSs. GSPAttack tailors a learning process to surrogate a recommendation model as well as generate fake users and user-item interactions while preserving the data correlation between users and items for recommendation accuracy. Although maintaining high accuracy for other items rather than the target item seems counterintuitive, it is equally crucial to the success of a poison attack. Extensive evaluations on four real-world datasets revealed that GSPAttack outperforms all baselines with competent recommendation performance and is resistant to various countermeasures.

Rapid industrial expansion, urbanization, and traffic growth have led to a decline in air quality that significantly impacts human health and environmental sustainability, particularly in developing nations. Due to the limited number of monitoring stations, the air quality index is not gathered at numerous locations. To address the difficulty of predicting the air quality value at an arbitrary place, several studies, including statistical and machine learning approaches, have been proposed. The majority of existing research employs classic distance-based interpolation techniques. In this paper, we propose a novel attentive neural-based approach for estimating unmonitored air quality values. This method follows the encoder-decoder paradigm, with the encoder and decoder being learned independently utilizing distinct processes. In the encoder, we propose AGE, an inductive unsupervised learning methodology that integrates attention mechanisms. AGE learns a set of functions that generate spatial embeddings by aggregating features from the surrounding region. For the decoder, we utilize the Gated Recurrent Unit and a fully-connected layer to estimate the air quality index at the targeted location. We conduct extensive experiments to evaluate the performance of our proposed method and compare it to the state-of-the-art (SOTA). The experimental results show that our approach reduces the estimation error from 8.07% to 37.04% compared to the SOTA.