Han Qilong

Memory Confidentiality and Integrity Protection Method Based on Variable Length Counter

Focusing on the problem of high overhead and frequent overflow of the counter mode encryption, this paper proposed an efficient scheme to protect data confidentiality and integrity. Based on the local characteristic of data accessing, the scheme set different counter length for memory area according to different accessing frequencies and the counter length can be dynamically adjusted. The analysis and the simulation results indicated that compared with the counter mode encryption, the scheme can decrease memory space overhead and the number of overflow. The proposed scheme can be applied to other schemes of protecting confidentiality and integrity based on counters and can satisfy performance requirements for most applications.

Symmetry theory based classification algorithm in CT image database

CT imaging shows that it is approximately symmetrical about the perpendicular bisector. Based on this medical knowledge guidance, symmetry theory based classification algorithm in CT image database is presented in this paper. First of all, the definitions of the weak symmetry and strong symmetry were given. Then, the weak symmetry was applied to the first stage classification of the CT images. Secondly, we proposed the combination of weak symmetry and strong symmetry for the second stage classification. Finally, according to the tumor edge profile, tumors are divided into benign and malignant lesions by extracting some features of the tumor in the third stage classification. In this paper, sample size requirements of SVM (Support Vector Machine) classifier were selected to classify the CT images. Experimental results show that symmetry theory based classification algorithm in CT image database can increase the accuracy of the classification and reduce the time of the doctors diagnosis.

A Memory Integrity Protection Method Based on Bloom Filter

Today, attackers are capable of using software or hardware methods to intrude computer network system. The phenomena of confidential information being tampered with or disclosed are emerging everywhere. Hardware-based attacks have emerged and caused tremendous losses. Focus on this question, this paper proposed new architectural mechanisms to ensure data integrity which improved previous work in this area without compromising security. The new integrity mechanism offers significant higher performance, space advantages and adjustable level of security over existing ones. This paper evaluated the overhead of the scheme and compared it with the conventional integrity checking scheme CHTree. For most benchmarks, the overhead of the new method is much better.

Page Write-Related Sector Translation for Solid State Drivers

Solid-state disks (SSD) are expected to replace magnetic disks in servers and supercomputers due to its low electronic power, nonvolatile storage, and stability. But solid-state drivers must be erased before fresh data can be written, and erase operations are too consuming, which affects the performance of solid state drivers. In order to hide the erase-before-write characteristics of NAND flash memory, an intermediate software layer called a flash translation layer (FTL) is proposed. However, most existing flash translation schemes such as BAST and FAST will increase the number of full merge operations and degrade the performance of SSD. In this paper, we propose a new solid state driver management scheme called Page Write-Related Sector Translation which is based on workload characteristic. PRST finds write-related pages from I/O request history set and writes these pages into the same block, which reduces garbage collection overhead and response time. For Postmark and IOzone workload, Experiment results indicate that PRST reduces response time by 35% over BAST and 26% over FAST. For TPC-C workload, PRST reduces response time by 12% compared to BAST and 10% compared to FAST.