Feature Analysis for Detecting Mobile Application Review Generated by AI-Based Language Model
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Mobile applications can be easily downloaded and installed via markets. However, malware and maliciousapplications containing unwanted advertisements exist in these application markets. Therefore, smartphoneusers install applications with reference to the application review to avoid such malicious applications. Anapplication review typically comprises contents for evaluation; however, a false review with a specific purposecan be included. Such false reviews are known as fake reviews, and they can be generated using artificialintelligence (AI)-based text-generating models. Recently, AI-based text-generating models have beendeveloped rapidly and demonstrate high-quality generated texts. Herein, we analyze the features of fake reviewsgenerated from Generative Pre-Training-2 (GPT-2), an AI-based text-generating model and create a model todetect those fake reviews. First, we collect a real human-written application review from Kaggle. Subsequently,we identify features of the fake review using natural language processing and statistical analysis. Next, wegenerate fake review detection models using five types of machine-learning models trained using identifiedfeatures. In terms of the performances of the fake review detection models, we achieved average F1-scores of0.738, 0.723, and 0.730 for the fake review, real review, and overall classifications, respectively.
Abstract
Mobile applications can be easily downloaded and installed via markets. However, malware and maliciousapplications containing unwanted advertisements exist in these application markets. Therefore, smartphoneusers install applications with reference to the application review to avoid such malicious applications. Anapplication review typically comprises contents for evaluation; however, a false review with a specific purposecan be included. Such false reviews are known as fake reviews, and they can be generated using artificialintelligence (AI)-based text-generating models. Recently, AI-based text-generating models have beendeveloped rapidly and demonstrate high-quality generated texts. Herein, we analyze the features of fake reviewsgenerated from Generative Pre-Training-2 (GPT-2), an AI-based text-generating model and create a model todetect those fake reviews. First, we collect a real human-written application review from Kaggle. Subsequently,we identify features of the fake review using natural language processing and statistical analysis. Next, wegenerate fake review detection models using five types of machine-learning models trained using identifiedfeatures. In terms of the performances of the fake review detection models, we achieved average F1-scores of0.738, 0.723, and 0.730 for the fake review, real review, and overall classifications, respectively.
참고문헌 (34)
- Yun Liu, "Information cascades prediction with attention neural network", Human-centric Computing and Information Sciences, 10(1) (2020), Springer Science and Business Media LLC[DOI]
- Zhiyong Zhang, "A crowdsourcing method for online social networks security assessment based on human-centric computing", Human-centric Computing and Information Sciences, 10(1) (2020), Springer Science and Business Media LLC[DOI]
- M. Harman, "App store mining and analysis: MSR for app stores", Proceedings of 2012 9th IEEE Working Conference on Mining Software Repositories (MSR), pp.108~111 (2012)
- M. Talal, "Comprehensive review and analysis of anti‑malware apps for smartphones", Telecommunication Systems, 72(2), pp.285~337 (2019)
- 최상용, "Automated Link Tracing for Classification of Malicious Websites in Malware Distribution Networks", JIPS(Journal of Information Processing Systems), 15(1), pp.100~115 (2019), 한국정보처리학회[DOI]
- H. Chen, "Toward detecting collusive ranking manipulation attackers in mobile app markets", Proceedings of the 2017 ACM on Asia Conference on Computer and Communications Security, pp.58~70 (2017)
- N. Genc-Nayebi, "A systematic literature review : opinion mining studies from mobile app store user review", Journal of Systems and Software, 125, pp.207~219 (2017)
- D. He, "Fake review detection based on PU learning and behavior density", IEEE Network, 34(4), pp.298~303 (2020)
- 정영식, "Learning Algorithms in AI System and Services", JIPS(Journal of Information Processing Systems), 15(5), pp.1029~1035 (2019), 한국정보처리학회[DOI]
- "Do massively pretrained language models make better storytellers?" (2019)[URL]
- "Language models are unsupervised multitask learners" (2019)[URL]
- Yonghun Jang, "Fake News Analysis Modeling Using Quote Retweet", Electronics, 8(12), pp.1377~ (2019), MDPI AG[DOI]
- M. Ott, "Finding deceptive opinion spam by any stretch of the imagination", Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics:Human Language Technologies, pp.309~319 (2011)
- "Improving language understanding by generative pre-training" (2018)[URL]
- T. Brown, "Language models are few-shot learners", Advances in Neural Information Processing Systems, 33, pp.1877~1901 (2020)
- "BERT: pre-training of deep bidirectional transformers for language understanding" (2019)[URL]
- "GLTR: statistical detection and visualization of generated text" (2019)[URL]
- L. Zhang, "CSLabel : an approach for labelling mobile app reviews", Journal of Computer Science and Technology, 32(6), pp.1076~1089 (2017)
- D. Martens, "Towards understanding and detecting fake reviews in app stores", Empirical Software Engineering, 24(6), pp.3316~3355 (2019)
- K. Ouazzane, "An artificial Intelligence-based language modeling framework", Expert Systems with Applications, 39(5), pp.5960~5970 (2012)
- R. Zellers, "Defending against neural fake news", Advances in Neural Information Processing Systems, 32, pp.9054~9065 (2019)
- "Automatic text summarization of COVID-19 medical research articles using BERT and GPT-2" (2020)[URL]
- "Writing the next American hit: using GPT-2 to explore the possibility of creating successful AIgenerated song lyrics" (2020)[URL]
- D. I. Adelani, "Generating sentiment-preserving fake online reviews using machine language models and their human-and machine-based detection", Advanced Information Networking and Applications, pp.1341~1354 (2020)
- Y. Nishi, "Agents and Multi-Agent Systems: Technologies and Applications 2020", pp.313~320 (2020), Springer
- S. Kreps, "All the news that’s fit to fabricate : AI-generated text as a tool of media misinformation", Journal of Experimental Political Science, 9(1), pp.104~117 (2022)
- "Can GPT-2 replace a Sex and the City writers’ room?" (2019)[URL]
- "GPT-based generation for classical Chinese" (2019)[URL]
- "TweepFake: about detecting deepfake tweets" (2021)[URL]
- Z. Horvitz, "Context-driven satirical headline generation", Proceedings of the 2nd Workshop on Figurative Language Processing, pp.40~50 (2020)
- Jieh-Sheng Lee, "Patent classification by fine-tuning BERT language model", World Patent Information, 61, pp.101965~ (2020), Elsevier BV[DOI]
- W. Huang, "Generating reasonable legal text through the combination of language modeling and question answering", Proceedings of the 29th International Joint Conference on Artificial Intelligence (IJCAI), pp.3687~3693 (2020)
- J. Peng, "Automatic generation of electronic medical record based on GPT2 model", Proceedings of 2019 IEEE International Conference on Big Data, pp.6180~6182 (2020)
- Joni Salminen, "Developing an online hate classifier for multiple social media platforms", Human-centric Computing and Information Sciences, 10(1) (2020), Springer Science and Business Media LLC[DOI]