Search

Machine learning algorithms based on deep neural networks have been widely used in many fields especially in computer vision, with impressive results. However, these models are vulnerable to different types of attacks like adversarial ones, which require attention to the model security and confidentiality. This study proposes a defense strategy to improve the insurance of white-box models while minimizing adversarial attacks against Fast Gradient Sign Method (FGSM)-based attacks and tackling the issue of adversarial training to improve their robustness. Mainly, we proposed a CNN model based on personalized activation functions using Beta function and its primitive. Then, the new resulted low degree polynomials are used to approximate the ReLU, Sigmoid and Tanh activation functions. Batch normalization was evoked to significantly improve the learning capacity of neural networks. The obtained results, using Mnist dataset prove the effectiveness of the proposed model.

A challenging engineering optimization problem in electrical power generation is the unit commitment problem (UCP). Determining the scheduling for the economic consumption of production assets over a specific period is the premier objective of UCP. This paper presents a take on solving UCP with Binary Slime Mould Algorithm (BSMA). SMA is a recently created optimization method that draws inspiration from nature and mimics the vegetative growth of slime mould. A binarized SMA with constraint handling is proposed and implemented to UCP to generate optimal scheduling for available power resources. To test BSMA as a UCP optimizer, IEEE standard generating systems ranging from 10 to 100 units along with IEEE 118-bus system are used, and the results are then compared with existing approaches. The comparison reveals the superiority of BSMA over all the classical and evolutionary approaches and most of the hybridized methods considered in this paper in terms of total cost and convergence characteristics.

Faults and system failure components are primarily two causes of unstable or deteriorating control performance of power system. In this study, we present a novel approach to the decentralized restoration of large DC microgrids using fault-tolerant control (FTC). The microgrid achieves decentralization by partitioning into several smaller grids. Each independent grid views the actions of the other grids as an external disturbance. The malfunction of the controller is represented in the input matrix as a norm-bounded uncertainty. The disturbance impact is diminished due to the proposed invariant-set approach. The proposed control can address simultaneous failures in actuators with random placement and degradation levels. In a passive FTC system, when the defect cannot be detected (or the fault may not have been clearly addressed), the proposed technique is utilized. After the fault has occurred, it can be viewed as an uncertainty in system dynamics. The controller that stabilizes the system is obtained by solving iteratively bilinear matrix inequalities as linear matrix inequalities. In addition, this study presents and discusses positive outcomes of applying this method to a system of six interconnected DC microgrids in the event of multiple fault types. The proposed control successfully stabilizes the severe case of simultaneous actuator faults.

This paper introduces a novel approach to addressing the security challenges of the Internet of Things (IoT) by presenting a secure Chaos-based lightweight cryptosystem. The proposed design incorporates a Pseudo-Chaotic Numbers Generator combined with the Speck64/128 lightweight block cipher, to meet the stringent requirements of security and lightweight characteristics. First, we subject the cryptosystem to a battery of rigorous tests, including various cryptanalytics such as brute-force, differential attacks, and statistical attacks. The results of these tests clearly demonstrate the cryptosystem’s exceptional security resilience and its capacity to effectively withstand these attacks. Then, the novel cryptosystem architecture, specifically designed for resource-limited applications like IoT, was implemented on a Xilinx PYNQ-Z2 XC7Z020 FPGA platform, which aligns perfectly with the constraints of IoT devices. The investigation comprises a thorough analysis and evaluation of the developed cryptosystem according to the algorithm complexity and the achieved precision, hardware area, maximum operational frequency, throughput, efficiency, and power consumption. The findings prove the effectiveness of our approach in terms of computational complexity, memory requirements, and power consumption. Eventually, after reviewing and comparing our results to the existing literature, our cryptosystem’s superiority becomes evident. The simulation results and performance analysis show that the proposed Chaos-Based Lightweight Cryptosystem (SCBLC) can be considered an ideal choice for securing communication in IoT devices with limited resources, making significant strides toward enhancing IoT network security.

Presents corrections to the paper, (Corrections to “Defending CNN Against FGSM Attacks Using Beta-Based Personalized Activation Functions and Adversarial Training”).