eCLASS: Edge-Cloud-Log Assuring-Secrecy Scheme for Digital Forensics

Junyoung Park, Eui-Nam HuhSymmetry 2019 · DOI

TL;DR

This paper introduces eCLASS, a novel logging scheme designed for edge-cloud environments to address the unique challenges of digital forensics. By employing log segmentation and distributed storage, eCLASS enhances log confidentiality and integrity while minimizing the computational overhead on resource-constrained edge nodes, making it more efficient than traditional cloud logging methods.

What Problem Does This Paper Address?

The increasing adoption of edge computing, driven by advancements like 5G, presents significant challenges for digital forensics. Edge nodes, being closer to users and more exposed to security threats, are vulnerable to tampering and manipulation. Traditional cloud logging systems struggle with the distributed nature and limited resources of edge environments. Furthermore, collecting forensic data from these nodes is complicated by issues like virtualization, multitenancy, and the potential for malicious actors to compromise edge devices or collude with cloud service providers (CSPs). Existing solutions like CLASS, while addressing some cloud forensics issues, are not optimized for the specific constraints and security landscape of edge computing, particularly regarding privacy and the need for independent forensic data collection.

Methods

eCLASS proposes a multi-pronged approach to tackle these challenges:

  • Log Segmentation: Instead of full encryption, eCLASS divides log data into small, non-continuous blocks. This segmentation makes it difficult to extract meaningful information from individual blocks, thus protecting confidentiality without requiring significant computational resources on edge nodes.
  • Distributed Storage: Segmented logs are stored across a distributed-storage cluster (DSC). This decentralization enhances resilience and prevents a single point of failure.
  • Multi-Index Chain (MIC) Network: An MIC network is established among participants (including investigators and other edge nodes) to manage and share index files. These index files contain information about the segmented logs and their storage locations. The MIC network ensures the integrity and verifiability of these indices through a chain of hash values.
  • Zero-Trust Network Policy: The scheme operates under a "zero-trust" principle, meaning no party (eCSP, investigator, DSC, or user) is inherently trusted, necessitating robust verification mechanisms.
  • User Public Key Encryption: Index files are encrypted using the user's public key, ensuring that only authorized users can access and recover their logs, thereby preserving privacy.

Datasets

Not discussed in the paper. The study uses simulated edge nodes with varying capacities and log file sizes for performance evaluation.

Key Insights and Ideas

  • Log Segmentation as a Lightweight Alternative: The paper demonstrates that log segmentation can be a more efficient and effective method for protecting confidentiality on resource-constrained edge nodes compared to traditional encryption methods like RSA.
  • Decentralization for Forensics: Distributing log storage and index management across a network (DSC and MIC) significantly enhances the integrity and recoverability of forensic data, making it harder for malicious actors or colluding parties to tamper with evidence.
  • Empowering Investigators: eCLASS provides investigators with a mechanism to collect and verify log data independently of the CSP, addressing concerns about CSP collusion or non-cooperation.
  • Privacy by Design: Encrypting index files with user public keys ensures that user privacy is maintained throughout the logging and retrieval process.

Key Results

  • Performance Improvement: The proposed log-segmentation method in eCLASS performed 29.4% to 64.2% faster than the CLASS scheme using RSA encryption across different edge node capacities.
  • Reduced Log Size: eCLASS reduced the log size by approximately 58% compared to CLASS log encryption.
  • Lower CPU Usage: Edge-node CPU usage was reduced by 14% to 28% with eCLASS compared to CLASS.
  • Cost-Effectiveness: eCLASS was found to be more economically efficient than CLASS, particularly in low-capacity edge nodes, offering up to 169% more efficiency in GM edge service models.
  • Enhanced Security: eCLASS addresses key threats like log modification, tampering, and privacy violation more effectively than CLASS, particularly in scenarios involving collusion.

Limitations and Future Work

The authors acknowledge several limitations and suggest future research directions:

  • Authentication and Access Control: The current eCLASS design focuses on log integrity and confidentiality but lacks a robust authentication and access-control scheme for its three main entities (edge node, investigator, DSC). Future work should address this by designing such a scheme.
  • Log Format Standardization: Service logs often have diverse formats, making it difficult for users to understand. Future research could explore standardizing log formats to improve usability.
  • Real-World Prototype and Evaluation: The study was conducted in a simulated environment. Implementing and evaluating a prototype of eCLASS in collaboration with real-world eCSPs, storage providers, and forensic investigators is crucial for validating its utility in practical scenarios.

Summary updated

BibTeX
@article{park2019eclass,
  author = {Junyoung Park and Eui-Nam Huh},
  title = {{eCLASS: Edge-Cloud-Log Assuring-Secrecy Scheme for Digital Forensics}},
  journal = {Symmetry},
  year = {2019},
  doi = {10.3390/sym11101192},
}
Figure