Stress is a complex physiological response to conditions perceived as perturbing the organism’s internal balance and involves the autonomic nervous system, resulting in a wide range of physiological responses, including cardiovascular and electrodermal changes. The aim of this work is to design and experimentally evaluate an adaptive neuromodulation system based on Binaural Beats (BB) for the modulation of cognitive stress. The study is organized into three phases: development of a protocol for stress induction, design of a system for real-time detection of the physiological state, and implementation of an algorithm for modulating the beat frequency according to the subject’s response. The stress detection system is based on a One-Class Support Vector Machine (OC-SVM) model, which uses features extracted from electrocardiogram (ECG) and electrodermal activity (EDA) signals, acquired using MotemaSens (OT Bioelettronica), to estimate physiological deviation from the resting state.