Article
Multidisciplinary Sciences
Tugba Turkoglu Kaya, Cihan Kaleli
Summary: Recommendation systems use popular methods to generate predictions and create product lists based on user feedback, increasing customer satisfaction through accurate recommendations. The level of personalization is crucial for successful predictions, which can be improved by collecting more detailed user data. Multi-criteria recommender systems, using user-item matrix to evaluate items in terms of multiple criteria, aim to achieve higher personalization level. However, vulnerability against shilling attacks is a significant challenge for recommender systems.
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Julio Barbieri, Leandro G. M. Alvim, Filipe Braida, Geraldo Zimbrao
Summary: The use of a generative model like Variational Autoencoder (VAE) to create fake profiles for Shilling Attacks in Collaborative Filtering (CF) system shows superior results compared to traditional attack models, especially at lower attack sizes. The ratings pattern of the generated profiles is very similar to real profiles, indicating a potential difficulty in detection.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Li Li, Zhongqun Wang, Chen Li, Linjun Chen, Yong Wang
Summary: In this study, a novel collaborative filtering recommendation technique (CFR-F) is proposed to defend against shilling attacks. Experimental results demonstrate that the approach can recommend accurate information resources with lower Mean Absolute Error (MAE) and Average Prediction Shift (APS) compared to traditional techniques.
CONNECTION SCIENCE
(2022)
Article
Computer Science, Artificial Intelligence
Li Yang, Xinxin Niu
Summary: This paper explores the benefits of using trust to resist shilling attacks in collaborative filtering recommender systems. It introduces a new approach called "genre trust degree" that considers both trust values and user credibility to improve the accuracy of recommendation models. Experimental results demonstrate the superiority of this method in defending against different types of shilling attacks.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Fatemeh Rezaimehr, Chitra Dadkhah
Summary: With the increasing amount of data, the use of recommender systems has increased, emphasizing the importance of recommendation quality for users. Most studies focus on collaborative filtering recommender systems, categorizing attack detection methods into clustering, classifying, feature extraction, and probabilistic approaches.
ARTIFICIAL INTELLIGENCE REVIEW
(2021)
Article
Computer Science, Information Systems
Reda A. Zayed, Lamiaa Fattouh Ibrahim, Hesham A. Hefny, Hesham A. Salman, Abdulaziz AlMohimeed
Summary: This paper emphasizes the importance of detecting shilling attacks in recommender systems and proposes various detection methods, including supervised, semi-supervised, unsupervised, deep learning, and hyper deep learning. The study also compares the existing attack detection methods and explores the exploitation of injected profile traits.
Article
Mathematical & Computational Biology
Chen Shao, Yue Zhong Yi Sun
Summary: This paper proposes a gradient boosting method named XGB-SAD, which combines double-view and gradient boosting to achieve attack detection. Experimental results demonstrate that XGB-SAD outperforms comparison methods in terms of small-scale attack detection and overall detection.
MATHEMATICAL BIOSCIENCES AND ENGINEERING
(2022)
Article
Computer Science, Information Systems
Yaojun Hao, Guoyan Meng, Jian Wang, Chunmei Zong
Summary: To simultaneously detect model-generative shilling attacks and group shilling attacks, we propose a detector based on graph convolutional networks (GCN). Our method characterizes these attacks by extracting user features from item popularity sequence and rating values, and constructs a user graph based on user distances. A two-stage scheme is developed for detecting shilling profiles using user features and the user graph. Experimental results show the efficacy of our method in detecting hybrid attacks.
INFORMATION SYSTEMS
(2023)
Article
Automation & Control Systems
Lei Liu, Lifeng Ma, Jie Zhang, Yuming Bo
Summary: This paper investigates the secure distributed set-membership filtering problem for general nonlinear systems over wireless sensor networks. Bias injection attacks and channel fading of wireless communication are taken into account in filter design. By using LMI technique and Taylor's expansion formula, nonlinearity, channel fading, bias injection attacks, and non-fragility are handled simultaneously to address the filter design problem. Sufficient conditions are obtained for pre-specified filtering performance, and an optimal algorithm is proposed for locally best performance. Simulation example demonstrates the effectiveness of the proposed secure filtering algorithm.
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE
(2021)
Article
Computer Science, Artificial Intelligence
Yaojun Hao, Fuzhi Zhang
Summary: This study proposes an unsupervised method based on deep learning and community detection for shilling attack detection. Experimental results show that the proposed method outperforms some baseline detectors in detecting simulated attacks and actual attacks.
Article
Computer Science, Artificial Intelligence
Pranpaveen Laykaviriyakul, Ekachai Phaisangittisagul
Summary: With the aim of increasing the robustness of deep learning models, this paper proposes a novel defense method against adversarial samples by filtering the perturbation noise in these samples. The method utilizes DiscoGANs to discover the relationship between attackers and defenders, and the defender model is trained to reconstruct original samples from the adversarial ones. In comparative experiments, the proposed method demonstrates promising results in improving the robustness against various attack models.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Fan Wu, Min Gao, Junliang Yu, Zongwei Wang, Kecheng Liu, Xu Wang
Summary: This paper introduces a novel shilling attack model called GOAT, which combines primitive attack methods and upgraded attack ideas, using a generative adversarial network and graph convolution structure to generate fake ratings, balancing feasibility and effectiveness.
INFORMATION SCIENCES
(2021)
Article
Mathematics, Applied
Yang Lou, Lin Wang, Guanrong Chen
Summary: This paper investigates network controllability robustness from the perspective of malicious attacks by proposing a hierarchical attack framework and demonstrating its effectiveness through experiments. The study suggests that critical edges and nodes should be hidden to protect the network from destructive attacks.
COMMUNICATIONS IN NONLINEAR SCIENCE AND NUMERICAL SIMULATION
(2021)
Article
Computer Science, Information Systems
Reda A. Zayed, Lamiaa Fattouh Ibrahim, Hesham A. Hefny, Hesham A. Salman, Abdulaziz Almohimeed
Summary: Shill attacks pose a threat to the stability of filtering and recommendation systems. This paper proposes an enhanced method that combines statistical and machine learning techniques to detect attacks in collaborative recommender systems, achieving higher accuracy.
Article
Computer Science, Information Systems
Jialie Shen, Neil Robertson
Summary: Recent years have seen rapid development of deep neural networks (DNN) and increasing interest in adversarial example attacks. Researchers have proposed an ensemble-based approach to enhance the robustness and reliability of DNN models, and introduced the BBAS scheme for diverse adversarial example generation.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Zeynep Batmaz, Burcu Yilmazel, Cihan Kaleli
JOURNAL OF AMBIENT INTELLIGENCE AND HUMANIZED COMPUTING
(2020)
Article
Computer Science, Artificial Intelligence
Ali Yurekli, Alper Bilge, Cihan Kaleli
Summary: The study shows that playlist titles can serve as an auxiliary data source to address the cold-start problem, with titles related to specific themes providing highly accurate recommendations. However, the correlation between title and recommendation usability is weak.
JOURNAL OF AMBIENT INTELLIGENCE AND HUMANIZED COMPUTING
(2021)
Article
Computer Science, Artificial Intelligence
Emre Yalcin, Firat Ismailoglu, Alper Bilge
Summary: Group recommender systems aim to suggest appropriate products/services to a group of users by determining group preferences. However, current aggregation techniques have limitations in capturing group members' propensities, and new methods are needed to discover items on which most members provided a consensus.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Information Systems
Emre Yalcin, Alper Bilge
Summary: This study focuses on investigating popularity bias in group recommender systems, analyzing various aggregation techniques, and proposing two strategies to mitigate bias towards popular items while maintaining reasonable ranking accuracy. Experiment results show that both strategies are more efficient than the adapted approach, significantly reducing bias towards popular items.
INFORMATION PROCESSING & MANAGEMENT
(2021)
Article
Chemistry, Multidisciplinary
Jehan Al-Safi, Cihan Kaleli
Summary: The article proposes an algorithm that utilizes item genre information to measure user similarity for more accurate recommendations, especially for customers who leave few ratings. The algorithm measures user relationships, finds nearest neighbors with similar preferences, and predicts item ratings using a defined procedure. Empirical results show improved accuracy, rating level, and user reliability compared to existing collaborative filtering algorithms.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Artificial Intelligence
Ali Yurekli, Cihan Kaleli, Alper Bilge
Summary: The paper presents a multi-stage retrieval system utilizing user-generated playlist titles to address the cold-start problem in music recommender systems. The system employs clustering, latent semantic indexing, and singular value decomposition to enhance recommendation accuracy and user experience in music listening.
INTERNATIONAL JOURNAL OF MULTIMEDIA INFORMATION RETRIEVAL
(2021)
Article
Computer Science, Artificial Intelligence
Tugba Kaya, Cihan Kaleli
Summary: This paper proposes a top -n recommender system method based on intuitionistic fuzzy sets for multi-criteria datasets, which focuses on determining the relational structure between products and investigating user tendencies. The method utilizes association rule mining and entropy measure to determine the rating distribution and relational structure, while analyzing user attitudes and tendencies with intuitionistic fuzzy sets.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Tugba Turkoglu Kaya, Emre Yalcin, Cihan Kaleli
Summary: Recommender systems are emerging techniques that improve predictive accuracy by considering users' past rating behaviors. However, these systems are more vulnerable to malicious attacks. In this study, a classification method based on user attributes is proposed to detect shill profiles in multi-criteria recommender systems, showing successful results in detecting attack profiles.
COMPUTATIONAL INTELLIGENCE
(2023)
Article
Computer Science, Software Engineering
Emre Yalcin, Alper Bilge
Summary: Recommender systems often exhibit popularity bias, favoring frequently rated items over less-rated ones in recommendation lists. This bias can affect users differently based on their gender, age, or rating behavior, leading to reduced overall satisfaction. This paper investigates the impact of user personalities on popularity bias and unfairness concerns, categorizing users into high, moderate, and low clusters based on personality traits. The experiments show that less-extroverted individuals and those avoiding new experiences are more likely to receive unfair recommendations in terms of popularity, despite being significant contributors to the system. However, algorithm choice affects discrepancies in other recommendation qualities such as accuracy, diversity, and novelty.
CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE
(2023)
Review
Computer Science, Artificial Intelligence
Muhammad Zaman Fakhar, Emre Yalcin, Alper Bilge
Summary: Smart homes provide a comfortable living environment and aim to reduce energy consumption. This review study provides an overview of previously proposed energy conservation techniques and recommendations, and includes comparative and statistical analyses. Possible research directions in energy conservation are also highlighted.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Multidisciplinary Sciences
Tugba Turkoglu Kaya, Cihan Kaleli
Summary: Recommendation systems use popular methods to generate predictions and create product lists based on user feedback, increasing customer satisfaction through accurate recommendations. The level of personalization is crucial for successful predictions, which can be improved by collecting more detailed user data. Multi-criteria recommender systems, using user-item matrix to evaluate items in terms of multiple criteria, aim to achieve higher personalization level. However, vulnerability against shilling attacks is a significant challenge for recommender systems.
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
(2023)
Article
Computer Science, Information Systems
Muhammad Zaman Fakhar, Emre Yalcin, Alper Bilge
Summary: This article presents a novel off-peak scheduling technique that provides instant energy scheduling recommendations by monitoring appliances in real-time and following user-devised criteria. The simulation results show significant cost-saving performance and outperform existing methods.
Article
Computer Science, Artificial Intelligence
Emre Yalcin, Alper Bilge
TURKISH JOURNAL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCES
(2020)