Terms of Deception: Exposing Obscured Financial Obligations in Online Agreements with Deep Learning (Extended Abstract

Published

May 23, 2024

Author(s)

Elisa Tsai, Anoop Singhal, Atul Prakash

Abstract

This paper investigates one type of social engineering scam, where unsuspecting users inadvertently consent to hidden financial obligations by performing routine online actions, such as making a purchase. Terms and conditions, often dense and overlooked, can be a vehicle for these scams, embedding deceptive or confusing terms to manipulate users. This paper highlights the suitability of a deep learning approach to address the wordplay and nuanced language used in these terms. We propose the design of TermLens, a browser plugin that leverages Large Language Models (LLMs) to detect obscured financial terms hidden within the fine print, a task that traditional security checks often miss. We show the feasibility of Termlens detecting obscured financial terms through a case study. We also discuss challenges and future plans.

Proceedings Title

IEEE 7th Deep Learning Security and Privacy Workshop

Conference Dates

May 23, 2024

Conference Location

San Francisco, CA, US

Pub Type

Conferences

Download Paper

Local Download

Keywords

Social Engineering, Fraud Detection, Machine Learning, Large Language Models

Citation

Tsai, E. , Singhal, A. and Prakash, A. (2024), Terms of Deception: Exposing Obscured Financial Obligations in Online Agreements with Deep Learning (Extended Abstract, IEEE 7th Deep Learning Security and Privacy Workshop, San Francisco, CA, US, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=957559 (Accessed April 23, 2026)

Additional citation formats