Realtime Privacy Monitoring on Smartphones

Update for those interested in installing TaintDroid: Tracking how apps use sensitive information required integrating our software into the Android platform at a low level. As a result, it was not possible to implement TaintDroid as a stand-alone app. Instead, to use TaintDroid you must flash a custom-built firmware to your device, similar to a number of popular community-supported Android ROMs. Please see the instructions for building, installing, and running Taintdroid on your phone.


A joint study by Intel Labs, Penn State, and Duke University has identified that publicly available cell-phone applications from application markets are releasing consumers' private information to online advertisers. Researchers at the participating institutions have developed a realtime monitoring service called TaintDroid that precisely analyses how private information is obtained and released by applications "downloaded" to consumer phones. In a study of 30 popular applications, TaintDroid revealed that 15 send users' geographic location to remote advertisement servers. The study also found that seven of the 30 applications send a unique phone (hardware) identifier, and, in some cases, the phone number and SIM card serial number to developers.

Smartphones offer a convenient way to download and install third-party applications. Over 200,000 applications are currently available in Apple's App Store and over 70,000 in Android's Market. Many of these applications access users' personal data such as location, phone information, and usage history to enhance their experience. But users must trust that applications will only use their privacy-sensitive information in a desirable way. Unfortunately, applications rarely provide privacy policies that clearly state how users' sensitive information will be used, and users have no way of knowing where applications send the information given to them.

The study was led by Jaeyeon Jung (a research scientist at Intel Labs, Seattle) and William Enck (a doctoral student at Penn State University). Their peer-reviewed report will be presented at the USENIX Symposium on Operating Systems Design and Implementation (OSDI) Oct. 4-6 in Vancouver, BC, Canada. Co-authors on the paper are Peter Gilbert (a doctoral student at Duke University), Landon Cox (an assistant professor at Duke University); Byung-Gon Chun (a research scientist at Intel Labs, Berkeley), Anmol Sheth (a research scientist at Intel Labs, Seattle); and Patrick McDaniel (an associate professor at Penn State University).

February 6, 2014: We updated the TaintDroid source code for Android 4.3. Please see the build instructions for details.
October 6, 2012: We released the TaintDroid source code for Android 4.1. Please see the build instructions for details.
August 30, 2011: We updated the TaintDroid source code for Android 2.3. Please see the announcement and build instructions for details.
October 13, 2010: We released the source code of the TaintDroid system. Please see the instructions for details.
September 29, 2010: See TaintDroid press release by Duke University and Penn State University and coverage by many online media outlets including Network World, CNET, The Register, Wired, BBC MSNBC, and Slashdot.
August 16, 2010: Look for our paper "TaintDroid: An Information-Flow Tracking System for Realtime Privacy Monitoring on Smartphones" at the USENIX Symposium on Operating Systems Design and Implementation (OSDI) in Vancouver, BC on October 6.
October 1, 2010: We fixed the following errors in the Table 2 of our OSDI paper:
(1) Moved "Barcode Scanner" to the "Camera only" row. See our letter to the author of the Barcode Scanner application.
(2) Renamed "Layer (Productivity)" to "Layar (Lifestyle)".
(3) Added a footnote to the "Applications" header that says, "Listed names correspond to the name displayed on the phone and not necessarily the name listed in the Android Market".
(4) Moved "3001 Wisdom Quotes Lite" to the "(Productivity)" category in the same row.

Research Contributions




Acknowledgements and sponsors

This material is based upon work supported by the National Science Foundation under Grant No. CNS-0905447, CNS-0721579 and CNS-0643907. Landon Cox and Peter Gilbert's participation was partially supported by NSF CAREER award CNS-0747283, NSF Grant No. CNS-1018547 and NSF Grant No. CNS-0910653.