Cryptanalysis Using Deep Learning

This project explores the cryptanalysis of the Grain Cipher using deep learning methods. The Grain Cipher utilizes Linear Feedback Shift Registers (LFSRs) and Non-linear Feedback Shift Registers (NFSRs) for encryption. The primary focus is on generating keystreams corresponding to initialization vectors (IVs), a constant key, and plaintext inputs.

Key Highlights

• Updates:
  • Initial Approach: Encountered challenges in the initial code
  • Revised Strategy: Transitioned to C language, importing the Grain module from GitHub repositories (as it is not openly available).

Grain Cipher Overview

The Grain Cipher encrypts data by generating keystreams using two key components:

• Linear Feedback Shift Register (LFSR): Generates predictable sequences based on linear operations.
• Non-linear Feedback Shift Register (NFSR): Introduces non-linearity for added security.

The implementation aims to:

• Generate keystreams for a given key and randomly generated IVs.
• Store keystreams and corresponding IVs in a structured format for analysis.

Keystream Outline

Key Details

The key, represented by the key1 array in hexadecimal format, is as follows:

01 23 45 67 89 AB CD EF 12 34

Each element in the key1 array represents a byte (8 bits) of the key. The key remains constant across iterations.

IV and Keystream Generation

• IVs are randomly generated for each iteration.
• The keystreams are computed based on the constant key and generated IVs.
• A total of 10,000 unique IV-keystream pairs are stored in a CSV file (data.csv) using the transferDataToCSV function.

Hexadecimal Representation

In this implementation, hexadecimal values are used extensively. The representation format 0xXX indicates hexadecimal values in C programming.

Examples:

• 0x01: Hexadecimal value 01 (decimal: 1).
• 0x23: Hexadecimal value 23 (decimal: 35).
• 0x45: Hexadecimal value 45 (decimal: 69).