Abstract
Data security is a primary consideration when users choose cloud storage, and it is also an important factor that hinders the development of cloud storage. In order to meet the requirements of secure storage of user-differentiated data and guarantee a good user service experience, we propose a multi-level security task scheduling scheme that comprehensively considers the data security level, user service level, and task waiting time. The scheme adopts the idea of multi-level security. The security requirements of different security level data are ensured by the key strength of the cryptographic algorithm. The mapping between the data security level and the key length under the selected cryptographic algorithm achieves the task's priority order. Scheduling appropriate encryption modules to implement differentiated encryption is suitable for cloud storage environments where data sensitivity and importance are diversified.
Introduction
As cloud storage becomes the backbone of both corporate and personal data management, "Data Leakage" has emerged as a top security threat. However, current cloud systems often apply a uniform encryption method to all data. This is inefficient: low-sensitivity data (like public videos) wastes expensive computing resources on heavy encryption, while high-sensitivity data (like medical records) might not receive the specialized protection it requires.
This paper identifies a critical gap: the need for a scheduling system that can handle "user-differentiated data." The challenge lies in balancing three competing factors: the security level of the data, the service level of the user (SLA), and the waiting time in the system queue.
Methodology
The proposed "Priority-based Multi-level Security Task Scheduling Scheme" is designed for heterogeneous multi-core processors. The methodology consists of three main pillars:
- Data Classification: Data is categorized into four levels, ranging from Level 1 (Public) to Level 4 (High security, e.g., medical records).
- Cryptographic Mapping: These security levels are mapped to specific algorithms and key lengths. For example, Level 1 might use 3DES with a 112-bit key, while Level 4 utilizes AES with a 256-bit key. This ensures that the "key strength" matches the data's importance.
- Priority Calculation: To decide which task to encrypt next, the system uses a weighted formula: $$Priority_i = \alpha \cdot Prior_i + \beta \cdot Wait_i$$ Where $Prior_i$ is the user's service level and $Wait_i$ is the time the task has already spent in the queue. The coefficients $\alpha$ and $\beta$ (which sum to 1) allow administrators to tune the system to favor either high-tier users or overall queue throughput.
The architecture features a "Cipher Machine" containing dedicated modules for 3DES, SM4, and AES. As shown in the system flow above, user data enters a First-Come-First-Served (FCFS) queue before being mapped to a security level and dispatched by a schedule module to specific waiting queues for each encryption algorithm.
Experimental Setup
The researchers evaluated their scheme using CloudSim, a standard open-source framework for simulating cloud computing environments.
- Task Parameters: Simulated task sets ranged from 50 to 1000 tasks with lengths between 100 and 300 units.
- User Service Levels: Assigned on a scale of 1 to 4.
- Comparison Baselines: The proposed algorithm was compared against standard Round Robin (RR) and Greedy scheduling algorithms.
Key Results and Discussion
The core innovation is treating security as a tunable resource. By linking key length directly to data sensitivity, the system avoids "over-encrypting" trivial data. Furthermore, by including "waiting time" in the priority formula, the system prevents the starvation of low-priority tasks.
In simulations, the proposed algorithm significantly outperformed the Round Robin (RR) approach in total task completion time. While the completion time was similar to the Greedy algorithm, the proposed scheme provided a better user experience by prioritizing tasks based on service levels and urgency rather than just arrival order. Additionally, the average waiting time remained competitive as task numbers increased, validating the effectiveness of the priority formula.
Conclusion
This multi-level security task scheduling scheme offers a flexible approach to cloud data protection. While currently validated via CloudSim, future work should explore real-world network latencies and hardware bottlenecks. Additionally, the manual setting of weights ($\alpha$ and $\beta$) could be replaced by machine learning models to dynamically optimize performance based on real-time traffic patterns.
Summary updated