Data collection in Wireless Sensor Networks (WSN) and specifically in the Internet of Things (IoT) networks draws significant attention both by the industrial and academic communities. Numerous Medium Access Control (MAC) protocols for WSN have been suggested over the years, designed to cope with a variety of setups and objectives. However, most IoT devices are only required to exchange very little information (typically one out of several predetermined messages), and do so only sporadically. Furthermore, only a small subset (which is not necessarily known a priori) intends to transmit at any given time. Accordingly, a tailored protocol is much more suited than the existing general purpose WSN protocols. In many IoT applications securing the data transmitted and the identity of the transmitting devices is critical. However, security in such IoT networks is highly challenging since the devices are typically very simple, with highly constrained capabilities, e.g., limited memory and computational power or no sophisticated algorithmic capabilities, which make the utilization of complex cryptographic primitives unfeasible. Furthermore, note that in many such applications, securing the information transmitted is not sufficient, since knowing the transmitters identity conveys a lot of information (e.g., the identity of a hazard detector conveys the information that a threat was detected). In this paper, we design and analyze an efficient secure data collection protocol based on information theoretic principles, in which an eavesdropper observing only partial information sent on the channel cannot gain significant information on the transmitted messages or even on the identity of the devices that sent these messages. In the suggested protocol, the sink collects messages from upto K sensors simultaneously, out of a large population of sensors, without knowing in advance which sensors will transmit, and without requiring any synchronization, coordination or management overhead. In other words, neither the sink nor the other sensors need to know who are the actively transmitting sensors, and this data is decoded directly from the channel output. We provide a simple secure codebook construction with very efficient and simple encoding and decoding procedures.