Share
طريقة تحليلية لانظمة التشفير الصوتي البايومتري == Analytical Approach of Biometric Based Voice Encryption System
Author name:
علي كاظم مطر
Supervisor name:
ستار بدر سدخان المالكي | بهيجة خضر شكر
General topic:
Computer Science
Specific topic:
Computer Science
Degree:
Doctorate
University:
University of Babylon - Information Technology Collage - Department Of Software
Language:
English
University location:
Babylon
First pages:
28T780 - p.pdf
Abstract:
There is no absolute security for important systems can ever get because attackers have always the ability to broke and attack them through their disadvantages. These days, biometrics is used to raise protection rate compared with the traditional methods. Authentication systems and cryptosystems are some of the important aspects of the practical life that use biometrics. The biometric voice is one of these traits that can build suitable secure systems according to the varying in the biometric voice.In this dissertation, two proposed analyzers were introduced to analyze Authentication system of biometric voice and the encryption process of the cryptosystems. The Authentication analyzer was used Additive White Gaussian Noise (AWGN) to stimulate effect of background/ transmission channel noise, and analyze the behaviors of the authentication system using FAR and FRR biometric performance measures.Performing proposed Authentication analyzer found degradation in the accuracy about (9.6% to 19.2%) in Splashdata database, about (1.9% to 5.7%) in Texas Instrument of Massachusetts Institute and Technology (TIMIT) database, and 0% in Texas Instruments - Digits (TIDIGITS) database of all selected members cannot be rejected illegally even when AWGN was reached 20 dB. SNR. Also, performing proposed Authentication analyzer found that the highest security degradation was about 0.05769 in Splashdata database, 0.01923 in TIMIT database and 0.05769 in TIDIGITS database of all selected members even when AWGN was reached 20 dB. SNR.The second proposed analyzer (Encryption analyzer) was also tested on the same databases to analyze them according to three randomness tests from National Institute of Standards and Technology (NIST) packages (pre - encryption phase), then these databases were also tested using Cross - correlation and Chi tests (post - encryption phase). These two phases produced two results by implementing two Mamdani fuzzifiers to get the certainty of each result. The final output of proposed Encryption analyzer was produced by merging the two previous Mamdani fuzzifiers in another final Mamdani fuzzifier.The performance of the Encryption analyzer was proved by classifying good/bad Keystream using the (post - encryption tests) and also with the comparison of the average value of other different 5 randomness tests from NIST.Finally, the ANFIS structure was used to generalize the hidden relationships that trained from the three randomness tests (pre - encryption process). The generalization process made ANFIS having the ability to predict the values of unseen (untrained) patterns. ANFIS results were promising results according to the proposed Encryption analyzer.Chapter
