Indirect File Leaks in Mobile Applications

Daoyuan Wu, Rocky K. C. Chang

Abstract

Today, much of our sensitive information is stored inside mobile applications (apps), such as browsing histories and chatting logs. To safeguard these privacy files, modern mobile systems, notably Android and iOS, use sandboxes to isolate apps' file zones from one another. However, we show in this paper that these private files can still be leaked by indirectly exploiting components that are trusted by the victim apps. In particular, we devise new indirect file leak (IFL) attacks that exploit browser interfaces, command interpreters, and embedded app servers to leak data from very popular apps, such as Evernote and QQ. Unlike previous attacks, we demonstrate that these IFLs can affect both Android and iOS. Moreover, our IFL methods allow an adversary to launch the attacks remotely, without implanting malicious apps in victim's smartphones. We finally compare the impacts of four different types of IFL attacks on Android and iOS, and propose several mitigation methods.

Introduction: The Confused Deputy Problem

Modern mobile security relies heavily on sandboxing—the architectural principle that App A cannot access App B’s files. While this prevents direct theft, it does not stop an app from being tricked into giving its own data away. This research addresses the "confused deputy" problem in the context of mobile file systems, identifying a gap in previous research: while local attacks on Android were known, the industry lacked a systematic, cross-platform analysis of how these leaks happen remotely and how they affect the supposedly more "locked-down" iOS ecosystem.

The IFL Attack Model

The researchers developed a high-level model for IFL attacks, where an external adversary leverages a trusted component within the victim application to bypass sandbox restrictions.

The IFL attack model illustrates how an adversary (a) exploits a deputy (d) to access private files (s).

The core challenge is that apps often use trusted components (like WebViews or internal servers) that have broader access to the app's private sandbox than an external adversary does. The attack flow follows a specific path: Adversary (a) → Deputy (d) → Private files (s) → Adversary (a).

Attack Vectors

The study categorizes these leaks into four distinct attack vectors:

  1. sopIFL (Same-Origin Policy IFL): Exploits flaws in how mobile browsers handle the file:// scheme. On mobile, this scheme often treats all local files as having the same origin, allowing a malicious HTML file to bypass origin restrictions and steal other local files.
  2. aimIFL (Active Injection Method IFL): Instead of bypassing SOP, the attacker injects malicious JavaScript directly into files the app renders (like cookie files or history logs). When the app displays these files in a WebView, the script executes and exfiltrates data.
  3. cmdIFL (Command IFL): Exploits apps that contain command interpreters, such as terminal emulators or SSH clients, to execute unauthorized shell commands.
  4. serverIFL (Server IFL): Targets apps with embedded web or FTP servers (common in "WiFi File Transfer" utilities) that have weak or non-existent authentication.

Cross-Platform Impact and Results

The researchers performed manual and semi-automated vulnerability testing on high-profile apps across both Android and iOS. The findings were significant: three out of the four IFL types were found to affect both operating systems.

Summary of IFL attack channels and affected popular applications across Android and iOS.

Key results include the discovery of a serious SOP issue in iOS 8 and vulnerabilities in massive apps like Evernote and Tencent QQ. In a test of the top 10 server apps, 70% failed to enforce any meaningful authentication, and 80% did not use encrypted transmissions. Crucially, these are not just local attacks; an adversary can trigger an IFL remotely by sending a malicious email attachment or enticing a user to click a link in a browser.

Mitigation and Future Work

The authors propose several mitigation methods, including NoJS (disabling JavaScript for local schemes) and AuthAccess (user-authorized access to private zones). However, they note that implementing these at the OS level is challenging for remote attacks where the context of the request is harder to verify. Future work should focus on building automated IFL detectors using static and dynamic analysis to scan app stores at scale.

Summary updated

Abstract

Today, much of our sensitive information is stored inside mobile applications (apps), such as the browsing histories and chatting logs. To safeguard these privacy files, modern mobile systems, notably Android and iOS, use sandboxes to isolate apps' file zones from one another. However, we show in this paper that these private files can still be leaked by indirectly exploiting components that are trusted by the victim apps. In particular, we devise new indirect file leak (IFL) attacks that exploit browser interfaces, command interpreters, and embedded app servers to leak data from very popular apps, such as Evernote and QQ. Unlike the previous attacks, we demonstrate that these IFLs can affect both Android and iOS. Moreover, our IFL methods allow an adversary to launch the attacks remotely, without implanting malicious apps in victim's smartphones. We finally compare the impacts of four different types of IFL attacks on Android and iOS, and propose several mitigation methods.
BibTeX
@misc{wu2015indirect,
  author = {Daoyuan Wu and Rocky K. C. Chang},
  title = {{Indirect File Leaks in Mobile Applications}},
  year = {2015},
  eprint = {1511.00104},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1511.00104},
}

Loading related content...

Figure