Chapter 4 Evaluation and Discussion
4.2 Simulation results and discussion
The good and acceptable web page response time is less than 2 seconds [16]. Therefore, we used two evaluation metrics, deadline miss ratio and average response time, defined as follows, to evaluate simulation results. Besides, we analyzed the influences of the protocol
z Deadline miss ratio (DMR): the ratio of number of clients’ requests that miss the response time threshold (e.g., two seconds)
z Average response time: the average time consumed from issuing matching requests until receiving responses of all clients under a specific deadline miss ratio (5%) We generated scenarios to evaluate the deadline miss ratio for six FIS approaches, Liu [19], Wang [21], HFIS-HTTP with one-to-many server matching, HFIS-HTTP with one-to-one server matching, Chang [22], and HFIS-TCP with one-to-many server matching.
These scenarios are of different number of clients from 100 to 1500 with intervals of 100.
Figure 14 shows the deadline miss ratios of different FIS approaches.
0
100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 Number of clients
Deadline miss ratio (%)
Liu-HTTP-image (PC-based, 1-N) [19] HFIS-HTTP-feature (PC-based, 1-N) Wang-HTTP-image (PC-based, 1-1) [21] HFIS-HTTP-feature (PC-based, 1-1) Chang-TCP-image (Embedded-based, 1-N) [22] HFIS-TCP-feature (Embedded-based, 1-N)
Figure 14. Deadline miss ratios of different FIS approaches.
As the number of clients grows, the deadline miss ratio increases. The deadline miss ratio of the proposed HFIS increases more slowly than that of related work with the number of clients increased. This is because when the number of client requests exceeds the limitation of concurrent threads that Fingerprint Server can provide, longer queuing time for service is needed. Liu-HTTP-image (PC-based, 1-N) [19] has to wait for feature extraction and
fingerprint matching in Fingerprint Server. But due to the feature extraction offloading to all clients, HFIS-HTTP-feature (PC-based, 1-N) only needs to wait for the time of fingerprint matching. That is why the HFIS-HTTP-feature (PC-based, 1-N) can support more number of clients than Liu-HTTP-image (PC-based, 1-N) [19]. Both Wang-HTTP-image (PC-based, 1-1) [21] and HFIS-HTTP-feature (PC-based, 1-1) perform the one-to-one server matching, so their increasing rate of deadline miss is much smaller than the one-to-many server matching of Liu-HTTP-image (PC-based, 1-N) [19] and HFIS-HTTP-feature (PC-based, 1-N). For the same reason depicted above, HFIS-HTTP-feature (PC-based, 1-1) can support more number of clients than Wang-HTTP-image (PC-based, 1-1) [21] and HFIS-TCP-feature (Embedded-based, 1-N) can support more number of clients than Chang-TCP-image (Embedded-based, 1-N) [22]. Table 10 illustrates the extra numbers of clients supported of HFIS-HTTP and HFIS-TCP compared to related work.
Table 10. Extra number of clients supported.
Deadline miss ratio
We can get the number of clients supported and record the response time of each transaction under different deadline miss ratios for each approach. So, we calculated the average response time of different deadline miss ratio for each approach as shown in Figure 15. The average response time increases when deadline miss ratio increases. The proposed HFIS has shorter average response time than related work for all cases, except HFIS-TCP with Embedded-based Clients at lower deadline miss ratio. Compared to Chang-TCP-image (Embedded-based, 1-N) [22], HFIS-TCP-feature (Embedded-based, 1-N) has shorter average response time when the deadline miss ratio is not less than 2%; however, it has longer average response time at lower deadline miss ratios (0% and 1%). This is because our proposed approach assigns the task of feature extraction to each client, and the time consumed of feature extraction is highly related to the client capability. Nevertheless, it gets shorter average response time than Chang-TCP-image (embedded-based, 1-N) [22] with the deadline miss ratio increased.
0
Liu-HTTP-image (PC-based, 1-N) [19] HFIS-HTTP-feature (PC-based, 1-N) Wang-HTTP-image (PC-based, 1-1) [21] HFIS-HTTP-feature (PC-based, 1-1) Chang-TCP-image (Embedded-based, 1-N) [22] HFIS-TCP-feature (Embedded-based, 1-N)
Figure 15. Average response time under different deadline miss ratios.
Table 11 shows the improvements of the average response time for the proposed HFIS compared to related work. We focus on results that have average response time under 2 seconds and found our proposed HFIS has better improvement at different deadline miss ratio:
HFIS-HTTP-feature (PC-based, 1-N) at 2% DMR, HFIS-HTTP-feature (PC-based, 1-1) at 5%
DMR, and HFIS-TCP-feature (Embedded-based, 1-N) at 3% DMR.
Table 11. Improvement of the average response time.
Deadline miss ratio
6.67% 16.67% 23.66% 21.31% 24.67% 15.66%
Wang-HTTP-image
-25% -10.18% 6.85% 14.58% 14.14% 23.14%
Based on the assumption of 10 Mbps bandwidth, the data transmission time of feature transfer is decreased by 3.52% compared to image transfer. Compared to HFIS-HTTP, the traffic reduction ratio of HFIS-TCP is 52.08% and the protocol overhead greatly reduced from 55.56% (HFIS-HTTP) to 7.25% (HFIS-TCP). Less protocol overhead can prevent the waste of bandwidth in the network and increase the network efficiency. As to the offloading effect, it enhanced the system scalability with no doubt as our preceding discussion.
Chapter 5 Conclusion
5.1 Concluding remarks
In this thesis, we have presented a hybrid fingerprint identification system, HFIS, for both types of HTTP-based and TCP-based applications. The fingerprint feature is extracted at the client side and is then transferred to the server for matching. The transferred data is a feature instead of an image, so the feature itself presents as a kind of encryption and it can enhance data security. Compared to other image-based approaches, under 2 seconds response time threshold, simulation results have shown that the proposed HFIS-HTTP increases the number of clients supported by 41.92% (reduces average response time by 23.66%) compared to Liu’s at 2% deadline miss ratio and by 84.06% (reduces average response time by 28.89%) compared to Wang’s at 5% deadline miss ratio; HFIS-TCP increases the number of clients supported by 45.28% (reduces average response time by 14.58%) compared to Chang’s at 3%
deadline miss ratio. In summary, the proposed HFIS-HTTP and HFIS-TCP have shorter average response time than related work for all cases, except HFIS-TCP with Embedded-based Clients at lower deadline miss ratios (0% and 1%), and thus is very feasible for various enterprise applications which require different combinations of high security, low cost, low response time, and high scalability.
5.2 Future work
To support a large scale fingerprint identification system, we can adopt multiple servers and a proper load balancing policy to avoid a potential bottleneck in TCP/IP Server or Web
Server of the HFIS system. In addition, we may deploy the HFIS to a cloud computing environment with on-demand virtual servers to provide scalable fingerprint identification services according to the incoming request arrival rate. That is, designing a cloud-based fingerprint identification system is feasible for large enterprise applications that require high security and high scalability.
References
[1] K. Uchida, “Fingerprint Identification,” NEC Journal of Advanced Technology, vol. 2, No.1, pp. 19-27, January 2005.
[2] T. Putte and J. Keuning, “Biometrical Fingerprint Recognition: Don’t get your fingers burned,” IFIP TC8/WG8.8 Fourth Working Conference on Smart Card Research and Advanced Applications, pp. 289-303, September 2000.
[3] A.K. Jain, L. Hong, and R. Bolle, “On-Line Fingerprint Verification”, IEEE Transactions on PAMI, pp. 302-314, 1997.
[4] S. Ribaric, D. Ribaric, and N. Pavesic, “Multimodal biometric user-identification system for network-based applications,” in Proceedings of IEE Vision on Image and Signal Processing, vol. 150, pp. 409-416, Dec. 2003.
[5] P. Schaumont, D. Hwang, S. Yang, and I. Verbauwhede, “Multilevel Design Validation in a Secure Embedded System,” IEEE Transactions on Computers, vol. 55, pp. 1380-1390, November 2006.
[6] P. Schaumont, D. Hwang, and I. Verbauwhede, “Platform-Based Design for an Embedded-Fingerprint-Authentication Device,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 24, pp. 1929-1936, Dec. 2005.
[7] V.K. Sagar, C. Greening, W.Y. Tan, and C.S.A. Leung, “Hardware/Software Co-Design of a Fingerprint Recognition System,” IEE Colloquium on Partitioning in Hardware-Software Codesigns, pp. 1-5, Feb. 1995.
[8] D.D. Hwang, and I. Verbauwhede, “Design of Portable Biometric Authenticators – Energy, Performance, and Security Tradeoffs,” IEEE Transactions on Consumer Electronics, pp. 1222-1231, Nov. 2004.
[9] M. Faundez-Zanuy., “A Door-Opening System Using a Low-Cost Fingerprint Scanner and a PC,” IEEE on Aerospace and Electronic Systems Magazine, pp. 23-26, Aug. 2004.
[10] P.J. Phillips, A. Martin, C.L. Wilson, and M. Przybocki, “An Introduction Evaluating Biometric Systems,” IEEE Computer, pp. 56-63, Feb. 2000.
[11] P.Y. Chan, and J.D. Enderle, “Automatic Door Opener,” in Proceedings of the IEEE Bioengineering Conference, pp. 139-140, April 2000.
[12] P. Gupta, S. Ravi, A. Raghunathan, and N.K. Jha, “Efficient Fingerprint-based User Authentication for Embedded Systems,” in Proceedings of Design Automation Conference, pp. 244-247, June 2005.
[13] G.C. Chao, S.S. Lee, H.C. Lai, and S.J. Horng, “Embedded Fingerprint Verification System,” in Proceedings of 11th International Conference on Parallel and Distributed Systems, pp. 52-56, July 2005.
[14] S.J. Alotaibi, and D. Argles, “FingerID: A new security model based on fingerprint recognition for distributed systems,” in Proceedings of 2011 World Congress on Internet Security (WorldCIS), pp. 284-289, 2011.
[15] HP LoadRunner Software. Available:
http://www8.hp.com/us/en/software-solutions/software.html?compURI=1175451#.T97On lLdHhc.
[16] F. Fui and H. Nah, “A study on tolerable waiting time: how long are Web user willing to wait,” Behavior and Information Technology, vol. 23, issue 3, pp. 153-163, 2004.
[17] D. Maltoni, D. Maio, A.K. Jain, and S. Prabhakar, “Handbook of Fingerprint Recognition,” 2nd Edition, 2009.
[18] J. Goldie, “The Ways to Bulletproof RS-485 Interfaces,” National Semiconductor Application Note 1057, 1996.
[19] W. Liu, C. Zhou, and Z. Ye, “Fingerprint Based Identity Authentication for Online Examination System,” in Proceedings of 2010 Second International Workshop on Education Technology and Computer Science, vol. 3, pp. 307-310, 2010.
[20] B. Shi, F. Xu, and M. Dai, “Design and implementation of wireless fingerprint identity authentication based on GPRS,” in Proceedings of 2010 Second International Conference Communication Systems on Networks and Applications, vol. 1, pp. 286-289, 2010.
[21] X. Wang, H. Zhang, T. Zhang, and L. Tong, “Design and Implementation of a Fingerprint Authentication System under B/S Architecture,” in Proceedings of 2009 CiSE on Computational Intelligence and Software Engineering, pp. 1-4, 2009.
[22] B.R. Chang, H.F. Tsai, C.M. Chen, and C.F. Huang, “Access Control of Cloud Computing Using Rapid Face and Fingerprint Identification,” in Proceedings of 2011 Second International Conference on Innovations in Bio-inspired Computing and Applications, pp. 179-182, 2011.
[23] Q. Shafi, J. Khan, N. Munir, and N.K. Baloch, “Fingerprint verification over the network and its application in attendance management,” in Proceedings of 2010 International Conference on Electronics and Information Engineering (ICEIE), vol. 2, pp. 555-559, 2010.
[24] J. Li, X. Zhu, X. Li, Z. Zhang, and J. Sui, “Wireless Fingerprint Attendance System Based on ZigBee Technology,” in Proceedings of 2010 2nd International Workshop on Intelligent Systems and Applications (ISA), pp. 1-4, 2010.
[25] B. Fernandez-Saavedra, R. Alonso-Moreno, A. Mendaza-Ormaza, and R. Sanchez-Reillo,
“Usability Evaluation of Fingerprint Based Access Control Systems,” in Proceedings of 2010 6th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, pp. 333-336, 2010.