Internet of Things (IoT) device usage has grown and has been adopted in various daily use devices applied in healthcare, smart homes, smart grids, connected cars and the list goes on. IoT devices have security vulnerabilities and cannot completely guarantee data privacy. As is the case with any network, IoT devices are also prone to hacks and Hardware Intrinsic (HI) attacks such as Hardware Trojans (HT), Firmware Modification and Memory Manipulation. The manifestation of HI attack can lead to various types of security issues which includes data theft and denial of service. Traditional HT attack detection techniques are valid for integrated circuit level only, and considered to be very invasive for an IoT device. Therefore, in this paper we propose a non-invasive approach that investigates Hardware Intrinsic Attack Detection in IoT (HIADIoT) devices. This approach detects covert channel and power depletion attacks through the power profile of IoT devices in different modes of operation utilizing machine learning algorithm. The power profile behavior of different IoT devices was observed over a period of time and then preprocessed to serve as data points. These data points is then provided to Random Forest Algorithm which correctly classifies 95.5% of the data point and recognizes potential HI attacks.