Original title: When Ad Networks Misbehave: Understanding Risks of Semi-Drive-By Splash Ads
Song Wu, Bo Wang, Yifan Zhang, Yinfeng Cao +1 · September 2026
When you open an app, you often see a full-screen splash ad. Advertisers pay ad networks when people interact with these ads. However, verifying these clicks is hard. Many ad software development kits, or SDKs, secretly turn tiny accidental phone movements or touches into fake ad engagement. The researchers call this fraud semi-drive-by splash ads. To study this trick, the team built AdHive. AdHive acts like an automated honeypot trap. It mimics real human behavior by using large language models to generate realistic usage patterns and sensor shifts. This setup triggers tricky ad code that usually stays hidden from basic security tools. The researchers tested thousands of popular Android apps and found this fraud is widespread. Splash ads frequently trigger from minor sensor variations. Working with one of China's largest advertisers, the team caught multiple ad networks committing this fraud, forcing them to repay about 4 million Yuan (approximately US$600,000).
Imagine you're at a store checkout and the salesperson holds out a tablet for you to sign. You accidentally brush the screen with your sleeve, and the signature registers as a real purchase, charging you for something you never meant to buy. In this paper, the same thing happens with mobile ads: a full-screen ad appears when you open an app, and an accidental touch or a tiny movement of the phone gets counted as a genuine tap. The party showing the ad is the salesperson, and the accidental brush is the accidental touch that gets billed as a real click.
Advertisers pay for engagement that never happened, so their money goes to ad networks that count accidental touches as real clicks. Users get interrupted by ads that seem to appear on their own, eroding trust in apps and the ad industry. The problem is widespread — thousands of popular Android apps — and has already forced repayments of about $600,000.
The study covers only Android apps and one advertiser in China, so it doesn't show whether the same fraud happens on iOS or elsewhere. It also doesn't measure total industry-wide losses, only the single confirmed repayment.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
When Ad Networks Misbehave: Understanding Risks of Semi-Drive-By Splash Ads — Song Wu, Bo Wang, Yifan Zhang et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.