Mental health disorders, such as stress, anxiety, and depression, are the frequent causes of disability all over the world, but they are not monitored in daily life due to the infrequent clinical visits, self-report bias, and the inability to provide continuous care. The recent progress in the Internet of Things (IoT) and wearable biosensors allows monitoring physiological and behavioral signals continuously and at real-world conditions that are associated with affective states and mental health course. The paper provides a literature-based review of an internet of things-based biosensors-based and passive sensing-based mental health monitoring. We do a review of the sensing modalities (EDA, ECG, PPG, respiration, temperature, accelerometry, EEG) combined modalities, multimodal fusion, stress and mood inference modules, and feasible designs of IoT pipeline across border edge-cloud systems. We examine benchmark data sets (e.g. WESAD, DEAP, AMIGOS) and also comment on real-life deployments. Particular attention is given to privacy, federated learning, user acceptability, and ethical governing because passive sensing is associated with distinct risks. Our proposal is a blueprint of end-to-end reference architecture and evaluation to support the work of researchers and developers in the construction of clinically meaningful, secure, and scalable systems. Lastly, we bring out open challenges including personalisation, domain shift, labeling scarcity, interpretability and fairness and we are offering a roadmap to future research.