Hark: A Deep Learning System for Navigating Privacy Feedback at Scale

Harkous, Hamza; Peddinti, Sai Teja; Khandelwal, Rishabh; Srivastava, Animesh; Taft, Nina

doi:10.1109/sp46214.2022.9833729

Cited by 20 publications

(16 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Based on analyses of Android app reviews [39,57,58], one of the main privacy concerns of end-users relates to app permissions and what access apps have to data and resources on their smartphones. However, when a permission request is contextualized, matches end-users' expectations, and is perceived by end-users as necessary for an app to function, end-users are likely to grant the permission request [21,50,54,73,98], which is rooted in the theory of Contextual Integrity [98].…”

Section: Related Work 21 App Permissionsmentioning

confidence: 99%

Stuck in the Permissions With You: Developer & End-User Perspectives on App Permissions & Their Privacy Ramifications

Tahaei,

Abu-Salma,

Rashid

2023

Preprint

View full text Add to dashboard Cite

While the literature on permissions from the end-user perspective is rich, there is a lack of empirical research on why developers request permissions, their conceptualization of permissions, and how their perspectives compare with end-users' perspectives. Our study aims to address these gaps using a mixed-methods approach.Through interviews with 19 app developers and a survey of 309 Android and iOS end-users, we found that both groups shared similar concerns about unnecessary permissions breaking trust, damaging the app's reputation, and potentially allowing access to sensitive data. We also found that developer participants sometimes requested multiple permissions due to confusion about the scope of certain permissions or third-party library requirements. Additionally, most end-user participants believed they were responsible for granting a permission request, and it was their choice to do so, a belief shared by many developer participants. Our findings have implications for improving the permission ecosystem for both developers and end-users. CCS CONCEPTS• Security and privacy → Software and application security;Human and societal aspects of security and privacy; Usability in security and privacy; • Human-centered computing → Human computer interaction (HCI); HCI design and evaluation methods; • Software and its engineering;

show abstract

Section: Related Work 21 App Permissionsmentioning

confidence: 99%

Stuck in the Permissions With You: Developer & End-User Perspectives on App Permissions & Their Privacy Ramifications

Tahaei,

Abu-Salma,

Rashid

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Leveraging OPP-115 and deep learning, Polisis extracts segments from privacy policies and presents them to users in a visualized format [6]. According to the paper, the union-based gold standard is used for experiments; 65 privacy policies were considered for training and 50 policies were kept for the test set.…”

Section: Related Workmentioning

confidence: 99%

“…In order to establish a firm foundation, we attempt to reproduce the work of [6] with additional improvements. To do that, we conduct experiments using word embeddings and a Convolutional Neural Network (CNN).…”

Section: Approachmentioning

confidence: 99%

“…Architecture To tackle the multi-label classification problem, we follow the work of [6] by using a CNN (displayed in Figure 2). The previously explained word embeddings are provided as input to the neural network.…”

Section: Convolutional Neural Networkmentioning

confidence: 99%

“…The first is the OPP-115 dataset, which contains 115 privacy policies at paragraph level, each of which includes fine-grained annotations from 3 experts [23]; e.g., the paragraph in Figure 1 from the Amazon policy 1 is annotated with two classes: User Access, Edit & Deletion and Data Retention. The second study which inspired our research is the effort by Polisis to build a Convolutional Neural Network (CNN) model exploiting OPP-115 [6]. Despite the valuable contribution of these earlier studies, they exhibit one major weakness: reproducibility.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Establishing a Strong Baseline for Privacy Policy Classification

Nejad

Jabat

Nedelchev

et al. 2020

IFIP Advances in Information and Communication Technology

View full text Add to dashboard Cite

Digital service users are routinely exposed to Privacy Policy consent forms, through which they enter contractual agreements consenting to the specifics of how their personal data is managed and used. Nevertheless, despite renewed importance following legislation such as the European GDPR, a majority of people still ignore policies due to their length and complexity. To counteract this potentially dangerous reality, in this paper we present three different models that are able to assign pre-defined categories to privacy policy paragraphs, using supervised machine learning. In order to train our neural networks, we exploit a dataset containing 115 privacy policies defined by US companies. An evaluation shows that our approach outperforms state-of-the-art by 5% over comparable and previously-reported F1 values. In addition, our method is completely reproducible since we provide open access to all resources. Given these two contributions, our approach can be considered as a strong baseline for privacy policy classification.

show abstract