Abstract:
In modern times, security has become a key issue for resource-constrained devices like embedded devices, wireless sensors, Radio Frequency Identification (RFID) tags, and Internet of Things (IoT) devices, which are increasing rapidly. These devices are expected to generate a massive amount of sensitive data for controlling and monitoring purposes. However, their resources and capabili- ties are limited. They also work with our valuable private data, thus making the security of those devices of paramount importance. As a consequence, an efficient and safe cryptographic algorithm can keep user data secret from intruders. Traditional encryption ciphers, such as Rivest–Shamir–Adleman (RSA) or the advanced encryption standard (AES), are computationally expen- sive, require a huge amount of memory, and hence hinder the performance of resource-constrained devices. The entire security strength of a cryptographic algorithm depends on the keys that are needed to encrypt and decrypt the message. Recently, most ciphers provide enough security for small computing devices based on complex number theories. However, the theory of complex numbers maximizes computing power and consumes more memory. Two dis- tinct key generation techniques are presented in this thesis. One of these is the Neural Network (NN) cipher, which is designed on a multilayer feedforward neural network with the concept of a nonlinear activation function to satisfy the Shannon confusion properties. Another is the Modified Butterfly architecture of the Fast Fourier Transform (FFT) for Keys (MBFK) cipher.
In the proposed NN cipher, it is shown here that NN consisting of 4 input, 4 hidden, and 4 output neurons is the best in the key scheduling process. With this architecture, five unique keys are generated from a 64-bit cipher key. Nonlinear bit shuffling is applied to create enough diffusion. The NN approach generates secure keys with a higher avalanche effect to meet Shannon confusion and diffusion properties while using less power, resulting in better performance than reported existing techniques.
The MBFK cipher modifies the butterfly architecture of Fast Fourier Trans- formation (FFT) in the context of cryptography. With the butterfly architecture, the MBFK cipher has a higher avalanche effect, which ensures that the gener- ated keys are strong enough to protect information from different statistical
attacks. It also meets the standard of the Shannon properties of confusion and diffusion. The proposed MBFK cipher is computationally lightweight because it uses simple logical operations like XOR and XNOR to generate the keys. As a consequence, the proposed MBFK cipher provides adequate security while also being energy efficient.
The fair evaluation of lightweight cryptographic systems (FELICS) tool was used to evaluate the memory usage and execution cycle of both ciphers. The proposed ciphers were also implemented in MATLAB to test key sensitivity by plotting the histogram, correlation graph, and entropy of various encrypted images. In addition, a security analysis was performed in terms of the Number of Pixels Changing Rate (NPCR), Unified Averaged Changed Intensity (UACI), and correlation coefficients; the findings demonstrate that the suggested ci- phers offer the required scores to provide the intended security against attacks.
