Abstract
Cloud computing acts as a computing paradigm that aims to provide a huge amount of computing in a fully virtualized manner by aggregating resources and thus offering a single system view. Cloud Computing is also delivered as a utility, assuring customized and quality-of-service guaranteed computation environments for cloud users. While an enterprise organization is composed of different departments like finance and admin, these departments are often segregated as sub-network zones interconnected via a network. Security is essential for the authorization of storage and computing. In this paper, we propose a privacy cheating discouragement and computation auditing approach that bridges secure storage and computation auditing in the cloud. Privacy cheating discouragement is designated by verifier signatures, batch verification, and probabilistic sampling techniques.
Introduction: The Trust Gap in Cloud Computing
As enterprises move sensitive data and heavy workloads to the cloud, they face a "trust but verify" dilemma. Cloud Service Providers (CSPs) are incentivized to save on I/O and CPU cycles. In some cases, a server might suffer a "Byzantine failure" or an external attack, leading it to return incorrect results or pretend it has performed a computation it actually skipped to save resources.
While public auditing for storage—checking if files are still intact—is a well-researched area, auditing computation (checking if the math was done correctly) is significantly more complex. If a user has to re-run the entire computation to verify the result, they lose the benefit of outsourcing it in the first place—a problem known as the "double-click" issue. This research seeks a middle ground: a way to audit both storage and computation efficiently without compromising user privacy.
Proposed Framework and Methodology
The proposed system architecture involves three main actors: the Cloud User (CU), the Cloud Service Provider (CSP), and a third-party Verification Agency (VA). The interaction between these entities is designed to ensure that the CSP remains honest through a series of verification requests and results returning.
Key techniques utilized in this framework include:
- Commitment-Based Sampling (CBS): Borrowed from grid computing, this technique allows the auditor to check only a fraction of the work (probabilistic sampling) to infer the correctness of the whole.
- Designated Verifier Signatures: To protect privacy, the user signs data in a way that only the specific Verification Agency can verify it. This prevents unauthorized third parties from snooping on the auditing process.
- Batch Verification: This allows the auditor to handle multiple verification requests simultaneously, reducing the computational overhead on the auditor's side.
- Master/Slave Model: Within the cloud environment, a master server controls multiple slave servers to execute tasks, while a dedicated security module handles data encapsulation and signing.
Performance Evaluation
The system was evaluated using simulated transactions of varying data sizes: 4 GB, 8 GB, and 16 GB. The performance was measured based on the number of threads and the transaction efficiency (Transactions per Dollar).
Key Results
- Scalability: The system was tested with up to 128 threads. The results indicate that throughput improves as more threads are utilized, though it eventually hits a plateau, which is typical for multi-threaded database and server environments.
- Data Volume Impact: While increasing the data size from 4GB to 16GB naturally increases the processing time per transaction, the curves for different data sizes follow a similar predictable pattern. This suggests the overhead of the security layer remains relatively stable as workloads grow.
- Efficiency: By using probabilistic sampling (checking random blocks of data/computation), the system provides a high statistical assurance of correctness without the "double-click" overhead of re-running the entire job.
Conclusion and Future Work
The core philosophy of this approach is "cheating discouragement." By making the chance of getting caught high and the auditing process low-cost for the user, the system creates a financial and reputational risk that discourages providers from cutting corners.
However, limitations remain. The model relies heavily on the "Verification Agency" being trustworthy; collusion between the VA and the CSP would break the security model. Additionally, while the system shows scalability, future work should involve more rigorous testing against specific "cheating" strategies—such as a server that only skips a tiny percentage of computations—to determine the exact sensitivity of the probabilistic sampling methods.
Summary updated