ADAPTIVE ENERGY-ORIENTED HARDWARE AND SOFTWARE SYSTEM FOR CRYPTOGRAPHIC PROTECTION OF INTERNET OF THINGS FOG NODES BASED ON FUZZY LOGIC

Abstract

Purpose. The purpose of the study is to develop and experimentally test an energy-oriented hardware and software method for adaptive management of the protection of electronic edge devices (fog nodes) in the Internet of Things (IoT). The method is based on fuzzy logic and selects the configuration of hardware and software protection tools depending on the current battery charge level and the varying threat level. This approach ensures the energy sustainability of embedded systems in critical infrastructure while simultaneously maintaining data confidentiality and integrity.

Methodology. The research was conducted using a specialized hardware and software testbed implemented in a Hardware-in-the-Loop format based on a Raspberry Pi 5 single-board computer, which acted as an autonomous IoT hardware platform. The software architecture was deployed as interacting Docker containers: an API request processing service, an Adaptive Security Manager (ASM) agent, and a monitoring module. A Takagi-Sugeno zero-order fuzzy logic controller was used for decision-making. The input linguistic variables were the state of charge estimation of the power subsystem and the current threat level, while the output was one of four predefined security modes. To evaluate the operational costs, a scalar objective function was developed that considers processing latency, the power consumption of the electronic node, and the security level. The discharge dynamics of the power subsystem were reproduced using an empirical logical discharge model that links the CPU load to the rate of energy consumption.

Results. Experimental testing, conducted across four 120-second scenarios, confirmed the advantages of the adaptive approach over static security policies. Fixed policies were shown to be irrational: the application of strong algorithms (e.g., AES-256) leads to accelerated depletion of the power subsystem under heavy traffic, while the continuous use of simplified modes increases device vulnerability risks. The results demonstrated that in normal mode, the adaptive approach reduces the power consumption of the electronic node by approximately 7% compared to fixed maximum protection, while guaranteeing a transition to the highest security level during an attack. In a critical low-charge mode, the system implements a survival strategy by maintaining the most energy-efficient configuration of hardware and software protection tools. This reduces power consumption by 40% and extends the battery life of the hardware platform under load by 1.7 times. Sensitivity analysis proved the stability of the system's physical indicators regardless of changes in the weighting coefficients of the objective function.

Scientific Novelty. The evaluation criterion for the performance of IoT hardware platforms was improved by introducing an integral objective function that formalizes the trade-off between processing latency, energy costs, and protection configuration. The study is the first to propose and experimentally validate a fuzzy-logic-based model for managing hardware and software protection tools that implements the concept of predictable survivability (graceful degradation) for embedded IoT systems. Unlike existing solutions, this model dynamically aligns power subsystem constraints with the need for enhanced protection, prioritizing energy savings when the charge is critically low and maximizing protection when sufficient resources are present.

Practical Significance. The practical application of the proposed solution significantly increases the reliability and energy survivability of embedded systems and IoT hardware platforms operating under unstable power supply conditions. Extending the battery life in critical situations provides a crucial window of opportunity needed to transmit important alarm notifications or wait for power restoration. The developed architecture allows for flexible modification of management scenarios and can be easily implemented on real electronic edge devices.