Abstract

With the growing ubiquity of the internet and internet enabled devices, the issue of online risk and the need for safety measures against such risks has grown paramount. Many researchers have analyzed the scope and characteristics of online risk, especially with respect to demographics, yet few have studied the risky media itself. This work sets out to move the conversation from surveys and interviews to content analysis and automation through a comprehensive thematic analysis of online multi-media risks (images and videos) sent to and from teens in private messages on Instagram. These messages, along with demographic information such as age and gender, were collected to form the Instagram Data Donation (IGDD) dataset. The findings from this analysis were used to inform semi- and self-supervised state-of-the-art vision transformers to produce validated labeled data for use in downstream risk detection and classification tasks. The results from this study show that online multi-media risks can be characterized by a myriad of content and contextual factors including media types (memes, screenshots, etc.), sender information (friend, acquaintance, etc.), and use of comedy and directed language. Our results also show that vision transformers are capable of learning complex image features and are capable of correctly attending to key, differentiable aspects of the images for use in risk detection and classification. Trigger Warning: This work includes graphic images of pornography, violence, self-harm, and bullying. We recognize that this material may not be appropriate for all readers and encourage readers to proceed with caution. Ethical Considerations: Our dataset consists of primarily private media messages that may contain personally identifiable information (PII). Because of this, we have chosen to use only the images that were publicly available on the web prior to the participant joining the study or have recreated the images ourselves with fully redacted PII.

Notes

If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu

Graduation Date

2022

Semester

Spring

Advisor

Wisniewski, Pamela

Degree

Master of Science (M.S.)

College

College of Engineering and Computer Science

Department

Computer Science

Degree Program

Computer Science

Format

application/pdf

Identifier

CFE0008986; DP0026319

URL

https://purls.library.ucf.edu/go/DP0026319

Language

English

Release Date

May 2023

Length of Campus-only Access

1 year

Access Status

Masters Thesis (Open Access)

Download

Share

COinS