Conceptual evidence collection and analysis methodology for Android devices

Ben Martini, Quang Do, Kim-Kwang Raymond ChooThe Cloud Security Ecosystem 2015 · DOI · DBLP

Introduction

Digital forensics on mobile platforms, particularly Android, presents unique challenges compared to traditional desktop environments. Authored by Ben Martini, Quang Do, and Kim-Kwang Raymond Choo, this research proposes a standardized, forensically sound framework that moves beyond "quick and dirty" methods like rooting. The goal is a device-agnostic methodology that prioritizes data integrity and legal admissibility.

The Forensic Challenge

Android devices employ aggressive sandboxing, hardware-backed encryption, and suffer from significant hardware fragmentation. Historically, investigators often relied on "rooting" to gain deep access. However, rooting typically involves unverified third-party exploits that modify the "userdata" or "system" partitions. Such modifications can taint evidence, potentially making it inadmissible in court. There is a critical need for a methodology that minimizes changes to the original evidence while working across diverse manufacturers.

Proposed Methodology

The researchers outline a rigorous methodology divided into four high-level stages: Identification, Collection, Examination/Analysis, and Reporting. This workflow is designed to maintain the "Minimal Handling" and "Account for Change" rules of digital forensics.

Proposed Evidence Collection and Analysis Methodology for Android Devices

Key Technical Approaches

  • Physical Imaging via Live OS: Instead of relying on the device's stock OS, the authors advocate booting a custom "Live OS" into RAM via the bootloader (using fastboot). This enables a bit-for-bit physical image of partitions without triggering stock security constraints.
  • System Framework Injection: To access sensitive account data (passwords and auth tokens) in the AccountManager, the authors decompile services.jar, modify PackageManagerService to disable signature checks, and re-inject it. This allows a forensic APK to "impersonate" any app to retrieve stored credentials.
  • Dynamic Memory Analysis: For apps using code obfuscation, the authors use DDMS to capture memory heaps (HPROF files), analyzing them with Object Query Language (OQL) to identify authentication and network activity strings.

Key Insights

  • The "Rooting" Trap: Rooting is an exploit, not a forensic tool. Its non-open-source nature makes it difficult for practitioners to explain its impact on a device in court.
  • Bypassing the Sandbox: By modifying the compareSignatures method, the researchers turned a security feature into a forensic gateway, accessing "secure" data without cracking individual app database encryption.
  • Live OS as Triage: Treating the Android device like a PC by booting from a trusted external environment circumvents passcode locks and prevents the stock OS from wiping data.

Results and Validation

The methodology was validated using a case study of seven popular cloud computing apps. Key results include:

  • Credential Recovery: Successful retrieval of plaintext passwords and authentication tokens usually inaccessible via standard logical backups.
  • Integrity Verification: Reliable physical imaging via adb forward with MD5/SHA1 hash verification, demonstrating that imaging can occur without stock OS interference.

Limitations and Future Work

The methodology faces hurdles such as "oem unlock" commands that may trigger data wipes and Full Disk Encryption (FDE) which requires a passcode if the device is powered off. Additionally, the complex process of de-odexing and recompiling system files requires further automation to be accessible to non-specialist practitioners.

Summary updated

BibTeX
@inproceedings{martini2015conceptual,
  author = {Ben Martini and Quang Do and Kim-Kwang Raymond Choo and Ben Martini and Quang Do},
  title = {{Conceptual evidence collection and analysis methodology for Android devices}},
  booktitle = {The Cloud Security Ecosystem},
  year = {2015},
  doi = {10.1016/B978-0-12-801595-7.00014-8},
}
Figure