SocInf: Membership Inference Attacks on Social Media Health Data With Machine Learning

Liu, Gaoyang; Wang, Chen; Peng, Kai; Huang, Haojun; Li, Yutong; Cheng, Wenqing

doi:10.1109/tcss.2019.2916086

Cited by 80 publications

(37 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A comprehensive survey presented in [37] shows the various detection methods for suicide ideation spanning several clinical methods, feature engineering based machine learning approaches or deep learning based methods. To detect suicidal ideation and provide early intervention to at-risk individuals, many researchers conducted psychological and clinical studies [81] using the classified responses of questionnaires [21,73] and analysis of social media data [46] based on feature engineering [52,69], sentiment analysis [63,89], and deep learning [6,38,36] techniques.…”

Section: Related Workmentioning

confidence: 99%

A Multitask Framework to Detect Depression, Sentiment and Multi-label Emotion from Suicide Notes

2021

View full text Add to dashboard Cite

The significant rise in suicides is a major cause of concern in public health domain. Depression plays a major role in increasing suicide ideation among the individuals. Although most of the suicides can be avoided with prompt intercession and early diagnosis, it has been a serious challenge to detect the at-risk individuals. Our current work focuses on learning three closely related tasks, viz. depression detection, sentiment citation, and to investigate their impact in analysing the mental state of the victims. We extend the existing standard emotion annotated corpus of suicide notes in English, CEASE, with additional 2539 sentences collected from 120 new notes. We annotate the consolidated corpus with appropriate depression labels and multi-label emotion classes. We further leverage weak supervision to annotate the corpus with sentiment labels. We propose a deep multitask framework that features a knowledge module that uses SenticNet's IsaCore and AffectiveSpace vector-spaces to infuse external knowledge specific features into the learning process. The system models emotion recognition (the primary task), depression detection and sentiment classification (the secondary tasks) simultaneously. Experiments show that our proposed multitask system obtains the highest cross-validation MR of 56.47 %. Evaluation results show that all our multitask models perform better than their single-task variants indicating that the secondary tasks (depression detection and sentiment classification) improve the performance of the primary task (emotion recognition) when all tasks are learned jointly.

show abstract

Section: Related Workmentioning

confidence: 99%

A Multitask Framework to Detect Depression, Sentiment and Multi-label Emotion from Suicide Notes

2021

View full text Add to dashboard Cite

show abstract

“…Narain et al found that attackers can use the gyroscope, accelerometer and magnetometer data on the smartphone of online social network users to infer the user's location [67], the comparison between datasets and measurements of other geo-location inference methods is given in Table 10. Twitter, Facebook, Weibo, and other online social networks are widely used to recommend "people you may know", that is, to predict users' friends from social media data [68]. Many online users are not willing to publish their friend networks.…”

Section: Othersmentioning

confidence: 99%

Privacy Inference Attack Against Users in Online Social Networks: A Literature Review

Piao

Cui

2021

IEEE Access

View full text Add to dashboard Cite

With the rapid development of social networks, users pay more and more attention to the protection of personal information. However, the transmission of users' personal information through social networks will inevitably lead to privacy leakage and make users attacked. A large amount of privacy information can be inferred from the content and social traces published by users, which leads to the rise of privacy inference technology for users in social networks. Social relationship inference and attribute inference are two basic attacks on users' privacy in social networks. This is the first systematic review of privacy inference attacks in social networks. The purpose of this retrospective study is to provide a global summary, summarizing the time trend of the topic, and show the evolution of the topic in the past 15 years. In addition, this paper discusses the trend and development of this topic and finally looks forward to future work. The contribution of our work is helpful for researchers to continue their research in this field.

show abstract

“…The shadow models mimic the target model's prediction behavior. To improve accuracy, Liu et al [15] and Hayes et al [10] leverage Generative Adversarial Networks (GAN) to generate shadow models with increasingly similar outputs to the target model. Salem et al [32] relax the attack assumptions mentioned in the work [39], demonstrating that shadow models are not necessary to launch the membership inference attack.…”

Section: Membership Inference Attackmentioning

confidence: 99%

The Audio Auditor: User-Level Membership Inference in Internet of Things Voice Services

Miao

Xue

Chen

et al. 2020

Proceedings on Privacy Enhancing Technologies

View full text Add to dashboard Cite

With the rapid development of deep learning techniques, the popularity of voice services implemented on various Internet of Things (IoT) devices is ever increasing. In this paper, we examine user-level membership inference in the problem space of voice services, by designing an audio auditor to verify whether a specific user had unwillingly contributed audio used to train an automatic speech recognition (ASR) model under strict black-box access. With user representation of the input audio data and their corresponding translated text, our trained auditor is effective in user-level audit. We also observe that the auditor trained on specific data can be generalized well regardless of the ASR model architecture. We validate the auditor on ASR models trained with LSTM, RNNs, and GRU algorithms on two state-of-the-art pipelines, the hybrid ASR system and the end-to-end ASR system. Finally, we conduct a real-world trial of our auditor on iPhone Siri, achieving an overall accuracy exceeding 80%. We hope the methodology developed in this paper and findings can inform privacy advocates to overhaul IoT privacy.

show abstract

SocInf: Membership Inference Attacks on Social Media Health Data With Machine Learning

Cited by 80 publications

References 33 publications

A Multitask Framework to Detect Depression, Sentiment and Multi-label Emotion from Suicide Notes

A Multitask Framework to Detect Depression, Sentiment and Multi-label Emotion from Suicide Notes

Privacy Inference Attack Against Users in Online Social Networks: A Literature Review

The Audio Auditor: User-Level Membership Inference in Internet of Things Voice Services

Contact Info

Product

Resources

About