Information flow based defensive chain for data leakage detection and prevention: a survey

Ning Xi, Chao Chen, Jun Zhang, Cong Sun, Shigang Liu, Pengbin Feng, Jianfeng MaarXiv 2021 · arXiv

TL;DR

This survey introduces an "Information Flow based Defensive Chain" (IFDC) framework to systematically analyze and categorize existing techniques for detecting and preventing data leakage in mobile and IoT applications. By organizing these techniques across the entire application lifecycle, from design to maintenance, the paper provides a comprehensive overview and identifies critical challenges and future research directions in this vital area of cybersecurity.

What Problem Does This Paper Address?

Mobile and IoT applications have become ubiquitous, offering immense convenience but also posing significant security risks due to their access to sensitive user data. Data leakage from these applications is a major concern, threatening user privacy, financial security, and even safety. While various defense mechanisms exist, they are often fragmented and focus on specific stages of an application's life cycle. This paper addresses the need for a holistic approach to data leakage prevention by proposing a unified framework that integrates Information Flow Control (IFC) techniques across all phases of an application's development and deployment.

Methods

The core contribution of this paper is the Information Flow based Defensive Chain (IFDC) framework. This framework structures IFC techniques into four main clusters corresponding to the application lifecycle:

  • Model-time Flow Control: Focuses on detecting vulnerabilities during the design phase using modeling techniques and formal verification.
  • Develop-time Flow Control: Employs techniques like type systems and static analysis on source code to catch leaks introduced during development.
  • Deploy-time Flow Control: Utilizes static analysis on compiled code (APKs, bytecode) to identify potential leaks before or during deployment. This is a heavily researched area, with techniques like taint analysis and data dependency analysis being prominent.
  • Run-time Flow Control: Monitors applications during execution using dynamic analysis, taint tracking, and secure multi-execution to detect and prevent leaks in real-time.

The authors systematically reviewed over 150 papers published in the last three years, categorizing them according to this IFDC framework and analyzing their techniques, performance, and limitations.

Datasets

The paper references a wide array of datasets used in the surveyed works, including:

  • DroidBench: A benchmark suite for Android information-flow analysis.
  • ICC Bench: A benchmark for inter-component communication analysis in Android.
  • GooglePlay: Large collections of real-world Android applications.
  • Malware datasets: Such as MalGenome and AMD dataset.
  • SmartThings datasets: Including commodity SmartApps and self-developed applications.
  • Benchmarks: Like CaffeineMark.

The paper does not introduce a new dataset but rather surveys existing research that utilizes these diverse datasets.

Key Insights and Ideas

  • Lifecycle is Crucial: Data leakage vulnerabilities can be introduced at any stage of an application's life cycle, not just during execution. A comprehensive defense requires addressing all phases.
  • IFC as a Unifying Framework: Information Flow Control (IFC) techniques offer a powerful and consistent approach to data leakage detection and prevention across different application types and development stages.
  • Trade-offs Exist: Each IFC technique has its strengths and weaknesses regarding precision, performance, and the complexity of code it can handle. The choice of technique depends on the specific phase and requirements.
  • Android Dominance, IoT Growth: The majority of research focuses on Android applications due to their widespread use, but there's a growing body of work addressing IoT applications.
  • Challenges Persist: Despite advancements, challenges remain in handling complex code (reflection, native code, external libraries), anti-analysis techniques, and ensuring the legality of detected flows, especially with the rise of AI-driven applications.

Key Results

The paper itself is a survey, so it doesn't present new experimental results. Instead, it synthesizes the results from over 150 surveyed papers. The key "results" are the categorization and analysis of these existing works. For example, the survey highlights that:

  • Static analysis techniques, particularly taint analysis, are heavily utilized in the deploy-time phase, with tools like FlowDroid achieving high precision and recall on large datasets.
  • Dynamic analysis techniques are essential for run-time detection, with systems like TaintART showing promising results in capturing leaks in newer Android versions with reduced overhead.
  • Many surveyed works report high precision and recall rates (often above 90%) on their respective benchmarks and datasets.

Limitations and Future Work

Limitations: * Focus on Android and IoT: While the survey covers mobile and IoT, the majority of reviewed works are on Android, with fewer studies on iOS. * Evolving Threat Landscape: The rapid growth of AI-based applications and IoT devices introduces new complexities and attack vectors that current IFC techniques may not fully address. * Challenges in Specific Areas: The paper identifies persistent challenges in analyzing complicated code (reflection, native code, external libraries), dealing with anti-analysis attacks, and ensuring the legality of detected flows, particularly with machine learning-based approaches.

Future Work: * More Precise Modeling: Developing more specific and granular IF properties for modeling and verification. * Improved Modeling Tools: Creating comprehensive modeling tools for IoT applications. * Richer Programming Language Support: Extending type systems to handle more complex language features and run-time behaviors. * Automated Typing Algorithms: Developing algorithms to automate the tedious process of assigning security types. * Analysis of Compiled Code: Enhancing techniques to analyze compiled code, which is crucial for app market owners who may not have source code access. * Advanced Static Analysis: Improving static analysis to handle complicated code, native calls, and external libraries more effectively. * Robustness Against Anti-Analysis: Developing techniques to resist side-channel and code-obfuscation attacks. * Enhanced Semantic-Based Validation: Addressing challenges in training data coverage and adversarial machine learning for more accurate flow legality detection. * Optimized Dynamic Analysis: Reducing the overhead of dynamic analysis and improving its practicality, especially for real-world devices. * SME Performance and IoT Adoption: Improving the performance of Secure Multi-Execution (SME) and exploring its adoption in IoT devices. * Richer Type Systems for Run-time: Enhancing type systems to support more complex run-time behaviors and languages.

Summary updated

BibTeX
@misc{xi2021information,
  author = {Ning Xi and Chao Chen and Jun Zhang and Cong Sun and Shigang Liu and Pengbin Feng and Jianfeng Ma},
  title = {{Information flow based defensive chain for data leakage detection and prevention: a survey}},
  year = {2021},
  eprint = {2106.04951},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2106.04951},
}

Loading related content...

Figure