The Performance Optimization For Multi-user ORAM

Oblivious Random Access Machine(ORAM)allows user to access cloud data without revealing the access pattern and protecte data's privacy.However,existing efficient ORAM schemes either have O(logN)bandwidth consumption or rely on expensive homomorphic encryption to achieve the so-called O(1)bandwidth blowup.Moreover,most ORAM schemes are are single-user oriented,and the existing multi-user scenario ORAM are difficult to satisfy the actual deployment requirements due to the security or performance problems.A more practical tree-based ORAM schemes—SORAM(Shamir-based ORAM)has been proposed,that can be actual deployed.We designed a new"no eviction"data write-back method in SORAM.Combined with Shamir secret sharing algorithm,SORAM can achieve O(1)bandwidth blowup,single round communication and O(1)client storage cost,without complex homomorphic encryption,also doesn't need to execute data eviction.At the same time,the security of the scheme is enhanced.Based on SORAM,the access security of SORAM in the multi-user scenario is guaranted by combining trusted third party proxy and drawing on the"sequencer"idea previously studied.and implemented ASORAM(Asynchronous multi-user shamir-based ORAM),a practical and deployable asynchronous access scheme that used in asynchronous network scenario.Comprehensive experiments has been conducted to evaluate the access performance and bandwidth cost of ASORAM.Experimental results show that,compared with~3,SORAM can reach 1.6x-2x bandwidth performance optimization,while the cost of client computing and storage consumption are almost negligible.In terms of overall performance,ASORAM has excellent concurrency scalability.When accessed 4K block with 10 clients at the same time,the average response time of ASORAM was half of TaoStore which is the latest research progress in this field.Moreover,as the number of client load increases to 100,ASORAM's average response time barely increases.