Publication Date

12-2019

Date of Final Oral Examination (Defense)

8-19-2019

Type of Culminating Activity

Thesis

Degree Title

Master of Science in Computer Science

Department

Computer Science

Supervisory Committee Chair

John Stubban, Ph.D.

Supervisory Committee Member

Marion Scheepers, Ph.D.

Supervisory Committee Member

Casey Kennington, Ph.D.

Abstract

Differential power analysis attacks are special kinds of side-channel attacks where power traces are considered as the side-channel information to launch the attack. These attacks are threatening and significant security issues for modern cryptographic devices such as smart cards, and Point of Sale (POS) machine; because after careful analysis of the power traces, the attacker can break any secured encryption algorithm and can steal sensitive information.

In our work, we study differential power analysis attack using two popular neural networks: Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN). Our work seeks to answer three research questions(RQs):

RQ1: Is it possible to predict the unknown cryptographic algorithm using neural network models from different datasets?

RQ2: Is it possible to map the key value for the specific plaintext-ciphertext pair with or without side-band information?

RQ3: Using similar hyper-parameters, can we evaluate the performance of two neural network models (CNN vs. RNN)?

In answering the questions, we have worked with two different datasets: one is a physical dataset (DPA contest v1 dataset), and the other one is simulated dataset (toggle count quantity) from Verilog HDL. We have evaluated the efficiency of CNN and RNN models in predicting the unknown cryptographic algorithms of the device under attack. We have mapped to 56 bits key for a specific plaintext-ciphertext pair with and without using side-band information. Finally, we have evaluated vi our neural network models using different metrics such as accuracy, loss, baselines, epochs, speed of operation, memory space consumed, and so on. We have shown the performance comparison between RNN and CNN on different datasets. We have done three experiments and shown our results on these three experiments. The first two experiments have shown the advantages of choosing CNN over RNN while working with side-channel datasets. In the third experiment, we have compared two RNN models on the same datasets but different dimensions of the datasets.

DOI

10.18122/td/1611/boisestate

Download

Share

COinS