Strengthening trust in computing technologies for detecting illegal deception online
summary
A team reports building a browser extension, curating a new dataset and running LLMs with retrieval augmented generation to detect dark patterns; existing public datasets prove unreliable, so the curated set and models produce results comparable to experts. Regulators use automated scans, cookie-config analysis and open tools (ETH Zurich code) for mass checks; one NGO files 637 complaints with 320 pending. Mobile apps and multimodal (video/voice) detection remain technically hard and unresolved.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:11Panelist backgrounds
- 12:27Dark pattern detection with LLMs
- 26:08AI's role in deception detection
- 38:53Regulatory challenges
- 51:17Cookie banners and consent
- 1:05:09Future of detection tools
- 1:14:27Q&A
related talks
37mFraud prevention in the age of online privacy
42mPrivacy Toolkit: Privacy Awareness In The Digital Age
48mSecuring the web with AI
35mBreaking fraud & bot detection solutions
52mBreaking Fraud and Bot Detection Solutions
01h 13mEnhancing privacy and security through technological innovation
01h 12mInvestigating algorithmic systems - Algorithm auditing.
28mFortifying Digital Trust: Next-Gen Strategies in Fraud Prevention