# Smartphone barometer can hear, and sense finger taps

## TL;DR

Smartphone barometers can detect speaker activity and finger taps, revealing sensitive user information despite being considered harmless.

## Contribution

First demonstration that barometer data can detect speaker activity and touchscreen interactions with high accuracy.

## Key findings

- Barometer readings at 25 Hz can detect speaker activity with ≥95% accuracy.
- Barometer samples can detect touchscreen finger taps with nearly 100% accuracy.
- Barometer data can reveal approximate positions of finger taps on the screen.

## Abstract

Nearly all modern smartphones are now equipped with a barometer to sample air pressure. Accessing these samples is deemed harmless, hence does not require permission. In this work, we demonstrate that barometer samples can reveal sensitive information, particularly in smartphones with ingress protection. Using a support-vector machine (SVM) classifier, we demonstrate for the first time that barometer readings, even at a low sampling rate (25 Hz), can reveal smartphone speaker activity. In particular, our classifier achieves \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\ge 95\%$$\end{document}≥95% accuracy in detecting whether the speaker is silent or playing a ringtone. In addition, we show that low-rate barometer samples can be used to 1) detect touchscreen finger taps with nearly \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$100\%$$\end{document}100% accuracy, and 2) gain information about the approximate position of finger taps. Our findings underscore that the barometer sensor, often considered harmless, should be recognized as sensitive with regard to user privacy, and access to this sensor should be carefully managed.