Research On Key Technology And Its Applications Of Boundary Limited Differential Privacy

Unprecedented amount of data is generated in the era of Mobile Internet,the flooding data provides service providers with good opportunates of digging users' preferences.However,from the perspective of information security,data exposure is a privacy threat to users,thus,plaintext should not be revealed.Differential privacy plays an important role in the area of privacy protection,it tries to preserve individual privacy from global view.It is proven that differential privacy can provide not only fine privacy protection,but formal mathematical proof as well.This paper focuses on boundary limited differential privacy as well as its applications in smart grid and data auction mechanism design.Firstly,two boundary limited differentially private schemes are proposed,then they can be proved to be in accordance to differential privacy requirements.Afterwards,applications of boundary limited differential privacy in smart grid and data auction mechanism design are discussed.In smart grid,users' data is collected by smart meters first,then reported to electricity supplier.The attacker can get users' data by eavesdropping,then discover real-time state of each appliance by applying skills like non-intrusive load monitor,as a result,privacy is leaked.Most of the existing solutions are based on a chargable-battery load hiding where charge or discharge of the battery is used to perturb the real consumption,unfortunately,these solutions either lack considerations of the battery's natural limitions or have potentials in privacy protection.Based on these observations,this paper proposes a differentially private meter reading report scheme(Priv Meter).The proposed scheme introduces the novel switch and penalty mechanism,decreases privacy loss by denying reporting meter readings that violate differential privacy,and take punishment from electricity supplier.Next,we decrease the probability of invalid noises with respect to charge or discharge rate and capacity by linear adaptation and multi-armed bandit mechanism.It can be proved that Priv Meter satisfies differential privacy.Experimental results indicate that Priv Meter could achieve 5 times better privacy protection than related solutions,meanwhile,extra costs of Priv Meter is tolerable.Data acution is an effective approach for elevating the value of data in the era of big data,it enables data analysis more possibilities.Designing practical data auction mechanism is crucial to the development of data trade,but privacy is still a problem.Existing schemes employ differential privacy to address privacy issues,however,they ignore boundary problems dealing with bounded data,which leads to low utility analysis results.In order to resolve utility issue,we proposes two utility-friendly differentially private schemes(UDPS1 and UDPS2).Theoretically,UDPS1 and UDPS2 satisfy differential privacy and their noises are bounded.According to experimental results,UDPS1 and UDPS2 achieve better data utility than existing solutions,in addition,privacy loss of UDPS2 outperforms other solutions.