Distinguishing Between Competing Crypto-Algorithms for the Known Ciphertext Case.

Date of Submission

December 2005

Date of Award

Winter 12-12-2006

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Master's Dissertation

Degree Name

Master of Technology

Subject Name

Computer Science

Department

Applied Statistics Unit (ASU-Kolkata)

Supervisor

Roy, Bimal Kumar (ASU-Kolkata; ISI)

Abstract (Summary of the Work)

In this report, we have described some statistical approaches for the classification of ciphertext in terms of the algorithm used to encrypt it. We have tried to do cryptanalysis using known ciphertext only attack. The above mentioned problem can be defined as followsgiven a finite set of possible algorithms and a ciphertext with an unknown origin, determine the algorithm which created this ciphertext by distinguishing among a finite set of possible algorithms.Since, there is no existing statistical literature to solve above mentioned problem, we tackled it from the very beginning. Our effort can be described in brief as follows.Formulation of Problem In the first step, we successfully formulated our problem . the unambiguous and simple manner(section 2).Design of Classification Strategy We viewed this problem as a classification prob- lem (section 2.1). Note that, in an area like cryptology where secrecy and confi- dentiality are the essence, misclassifying the origin of a ciphertext can lead to a potential disaster. It is desirable that our classifier must have very little probabil- ity of misclassification. Since it is very difficult to design a classifier with almost 0% misclassification probability, we slightly-modified commonly used classification approach to overcome such harsh restriction.Identification of Different ParametersIn general, parameters are terms associated to strategy, which is used to solve given problem. For example, in our Classifier design strategy, parameters are Choice of training ciphertext, Design choice for training set and similarity measure and so on. Till now, we identified 6 diffe parameters (section 2.2).Fixing parameters To handle a problem at hand, we must fix parameters. To fix it, we used observations obtained in experiment. We also applied our intuition and statistical known behavior of cryptosystems to fix it. For example, we al- lowed key mixing after getting observation, by experiment, that key may affect the performance of classifier (section 2.2).Analysis of behavior of cipher keys Intuitively, we felt that the key may had sig- nificant role in our process. We had analyzed its behavior and found that their were instances when the variation due to the key could hamper the distinguishing process (section 3.3).Design classifier using different learning schemes In this report, we used statisti- cal measures to design classifiers. In particular, we experimented using frequency based learning scheme. We analyzed the effectiveness of One character based frequency analysis (section 5.3) and two characters based frequency analysis (section 5.4). Also, see (section 2.2). We obtained encouraging result in this analysis (section 3.4 and 3.5).In this report, we presented four problems. First two problems were meant to analyze the behavior of frequency pattern, key etc. In last two problems, we applied knowledge gained from this analysis for classifier design.

Comments

ProQuest Collection ID: http://gateway.proquest.com/openurl?url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&res_dat=xri:pqm&rft_dat=xri:pqdiss:28843078

Control Number

ISI-DISS-2005-159

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

DOI

http://dspace.isical.ac.in:8080/jspui/handle/10263/6328

This document is currently not available here.

Share

COinS