ENERGY THEFT DETECTION IN MULTI-TENANT DATA CENTERS WITH DIGITAL PROTECTIVE RELAY DEPLOYMENT

 

ABSTRACT

High performance data centers serve as the backbone of the prevailing cloud computing paradigm. A data centerswith different operational structures, multi-tenant data centers (MTDCs) are increasingly popular a various internet serviceproviders for the ease of deployment. Despite the offered benefits, MTDCs are vulnerable to various cyberattacks. An importantcyberattack is energy theft which can be launched by malicious tenants to reduce monetary cost of the electricity consumption. It canbe achieved through attacking a smart meter in the data center to undercount its energy usage. By alleviating the financial burden ofthe cloud service providers in MTDCs, energy theft discourages frugality in terms of energy consumption, which is highly undesirable inthe era of sustainable computing. Despite fruitful research results on MTDCs, none of them address energy theft. When energy theftoccurs, it might be necessary for the data center operator to examine smart meter of all tenants to find the compromised ones whichcould induce excessive labor cost.Localization of energy theft detection is an effective way to limit the labor cost in detecting energy thefts in MTDCs. It can be facilitatedthrough deploying Digital Protective Relays (DPR) in the data center where a DPR is a microprocessor based device for fault detectionand event logging in the power system. In this paper, we propose an anomaly rate range based dynamic programming algorithm forinserting DPRs into the data center while minimizing the deployment cost. Our algorithm optimizes the DPR deployment throughexploring an innovative aggregated anomaly rate range which accounts for the long term effect of energy theft in an MTDC. In addition,given the historical records of energy usage for all tenants, we calculate the anomaly rate range for each tenant, leveraging theMinimum Covariance Determinant (MCD) based anomaly identification algorithm. To the best of our knowledge, this is the first workaddressing the energy theft issue in multi-tenant data centers. The simulation results demonstrate that our algorithm inserts 18:9% lessDPRs into the data center compared to a natural baseline algorithm. Meanwhile, in an attempt to identify all energy theft cases, ourDPR insertion solution requires 14:7% less tenants to be checked compared with a natural heuristic baseline algorithm.

EXISTING SYSTEM:

There is significant research on smart grid cybersecurity. A set of code optimization techniquesare proposed in to improve the efficiency of energyconsumption in smart grid security algorithms. A partiallyobservable Markov decision process based framework is developedin to detect pricing cyberattacks on prevailingelectricity billing models and detect energy theft in the smarthome system. Despite these, none of them discusses thedetection of cyberattacks in MTDCs.In order to detect energy theft, a trivial solution is toexamine the smart meter of every tenant to find the attackedones. This is inefficient and is potentially associated withhigh labor costs. Narrowing the search zone is the keyto reducing the manual effort of energy theft detection. Itcan be achieved by inserting Digital Protective Relay (DPR)into the power distribution network of the data center. ADPR is a microprocessor based power system controller,integrated with the Ethernet communication module. It isusually deployed in the power system for event logging andon-site detection of electrical faults .Leveraging the metering capability of DPRs, our solutionof localization of energy theft detection is to insert DPRsinto the distribution network of the data center. Each DPRindependently collects and reports the aggregated electricityusage of its downstream network to the billing center inreal-time. Meanwhile, the electricity consumption for eachtenant is measured and reported by each smart meter. Foreach DPR, if the aggregated electricity usage does not matchthe summation of the energy usage reported by all downstreamsmart meters, energy theft is identified. In this case,only the smart meters in the downstream of the DPR are tobe examined in order to find the attacked ones.

PROPOSED SYSTEM:

we propose a new dynamicprogramming based DPR insertion algorithm. Differentfrom, we do not assume that anomaly rateof each tenant is given. In contrast, they are calculatedfrom the anomalies identified in historical record of energyusage by an intelligent Minimum Covariance Determinant(MCD)-based anomaly identification algorithm. Given theconstraints that limit the number of smart meters to check inthe worst case, our algorithm inserts the minimum numberof DPRs into the power distribution network of the datacenter. In addition, our DPR insertion solution explores aninnovative aggregated anomaly rate range which accountsfor the long term effect of energy theft in an MTDC. Thecontribution of this paper can be summarized as follows._ An anomaly rate range based dynamic programmingalgorithm is proposed to insert DPRs into MTDCs forenergy theft detection while minimizing the deploymentcost. Our algorithm optimizes the DPR deploymentthrough exploring an innovative aggregatedanomaly rate range which accounts for the long termeffect of energy theft in an MTDC._ To the best of our knowledge, this is the first workaddressing the energy theft issue in multi-tenant datacenters._ Our algorithm is more practical than the one. It estimates the anomaly rate of each tenantfrom the historical record of energy usage, leveragingthe Minimum Covariance Determinant (MCD)basedanomaly identification technique. In addition,it evaluates the anomaly rate of the tenants via theanomaly rate range, which reflects the anomaly ratemore realistically._ Compared with a natural baseline algorithm, ouralgorithm inserts 18:9% less DPRs while requiring14:7% less tenants to be checked in order to discoverall energy theft cases in an MTDC, as verified by thelong-term simulation._ Our algorithm is highly scalable and is designed tohandle data centers of arbitrary size. It runs withinone second even for the largest test case with 2000tenants.

CONCLUSION AND FUTURE WORK

Energy theft in MTDCs is highly undesirable in the era ofsustainable computing as it encourages overuse of energyfor the cloud service providers. When energy theft occurs inan MTDC, the data center operator could have to examineeach smart meter in order to find the compromised ones.Due to the huge number of tenants, it is highly desirable tolocalize the energy theft detection by inserting DPRs into thedistribution network in the data center. Meanwhile, it is desirableto minimize the number of DPRs inserted due to itshigh cost. This paper proposes an anomaly range based dynamicprogramming algorithm for inserting the minimumnumber of DPRs into the data center while still limitingthe average number of tenants to be checked when energytheft occurs. Our algorithm optimizes the DPR deploymentthrough exploring an innovative aggregated anomaly raterange which accounts for the long term effect of energy theftin an MTDC. Compared with the baseline algorithm, ouralgorithm inserts 18:9% less DPRs with almost the samenumber of tenants to check on average, according to ourpreliminary results. To the best of our knowledge, this isthe first work addressing the energy theft issue in MTDCs.As the future work, the impact of different energy theftpatterns on our DPR insertion algorithm will be furtherstudied. Anomaly detection algorithms will be proposedto handle various specific energy theft patterns, such ascollusive energy theft, in which the malicious tenant attacksboth its own smart meter and the neighboring ones.

REFERENCES

[1] Data Center Knowledge. (2014) Four big advantagesof multi-tenant data centers. [Online]. Available:http://www.datacenterknowledge.com/archives/2014/08/07/four-big-advantages-multi-tenant-data-centers/

[2] M. A. Islam, H. Mahmud, S. Ren, and X. Wang, “Paying to save:Reducing cost of colocation data center via rewards,” in Proceedingsof 2015 IEEE 21st International Symposium on High PerformanceComputer Architecture (HPCA), Feb 2015, pp. 235–245.

[3] M. A. Islam, X. Ren, S. Ren, A. Wierman, and X. Wang, “A marketapproach for handling power emergencies in multi-tenant datacenter,” in Proceedings of 2016 IEEE International Symposium on HighPerformance Computer Architecture (HPCA), 2016.

[4] Natural Resources Defense Council. (2016) America’s datacenters consuming and wasting growing amounts ofenergy. [Online]. Available: http://www.nrdc.org/energy/data-center-efficiency-assessment.asp

[5] A. Qureshi, R. Weber, H. Balakrishnan, J. Guttag, and B. Maggs,“Cutting the electric bill for internet-scale systems,” in ACMSIGCOMM Computer Communication Review, vol. 39, no. 4. ACM,2009, pp. 123–134.

[6] N. H. Tran, D. H. Tran, S. Ren, Z. Han, E. N. Huh, and C. S. Hong,“How geo-distributed data centers do demand response: A gametheoreticapproach,” IEEE Transactions on Smart Grid, vol. 7, no. 2,pp. 937–947, March 2016.

[7] Y. Liu and S. Hu, “Cyberthreat analysis and detection for energytheft in social networking of smart homes,” IEEE Transactions onComputational Social Systems, vol. 2, no. 4, pp. 148–158, Dec 2015.

[8] J. Wurm, K. Hoang, O. Arias, A.-R. Sadeghi, and Y. Jin, “Securityanalysis on consumer and industrial iot devices,” in Proceedings of21st Asia and South Pacific Design Automation Conference (ASP-DAC2016), 2016.

[9] M. Qiu, W. Gao, M. Chen, J.-W. Niu, and L. Zhang, “Energyefficient security algorithm for power grid wide area monitoringsystem,” IEEE Transactions on Smart Grid, vol. 2, no. 4, pp. 715–723,2011.

[10] M. Qiu, H. Su, M. Chen, Z. Ming, and L. T. Yang, “Balance ofsecurity strength and energy for a pmu monitoring system insmart grid,” IEEE Communications Magazine, vol. 50, no. 5, pp. 142–149, May 2012.

[11] Y. Liu, S. Hu, and T. Y. Ho, “Leveraging strategic detection techniquesfor smart home pricing cyberattacks,” IEEE Transactions onDependable and Secure Computing, vol. 13, no. 2, pp. 220–235, March2016.