With the development of low-cost computer chips and wireless networks, any object or thing can become a component of the Internet of Things (IoT). As a result, these disparate things will gain digital intelligence, enabling them to connect with real-time data and be used for a wide range of valuable and practical IoT applications. However, due to the broadcast nature of wireless communication and the presence of eavesdroppers, the IoT systems pose serious threats to privacy and message integrity, particularly with mobile sensors. Recently, Physical-Layer Security (PLS) has been proposed as a cost-effective solution that mitigates the impact of growing security threats. In this paper, we propose a PLS-based hybrid Non-Orthogonal Multiple Access (NOMA) / Power Division Multiplexing (PDM) IoT system that provides a high probability of secured transmission, a low computational complexity, and a low power consumption. 3-D stochastic geometry tools have been used to introduce and evaluate our proposed scheme in a variety of practical scenarios where users are randomly located in a 3-D space. Furthermore, we derive the new scheme’s Key Agreement Ratio (KAR) expression, the probability expressions of using NOMA and PDM modes. Moreover, we derive the associated Secrecy Outage Probability (SOP), and the Average Ergodic Capacity (AEC) expressions for the proposed scheme as well as for the pure NOMA and pure PDM schemes. In addition, we introduce and investigate the Security Power Efficiency (SPE) metric, that serves as a new key agreement optimization metric. Numerical results are performed to validate the obtained analytical expressions and to assess the proposed scheme performances in terms of KAR, SOP, and AEC, when compared to pure NOMA and pure PDM schemes.