The rapid advancement of the Internet of Things (IoT) is distinguished by heterogeneous technologies that provide cutting-edge services across a range of application domains. However, by eavesdropping on encrypted WiFi network traflc, attackers can infer private information such as the types and working status of IoT devices in a business or residential home. Moreover, since attackers do not need to join a WiFi network, such a privacy attack is very easy for attackers to conduct while at the same time invisible and leaving no trace to the network owner. In this paper, we extend our preliminary work originally presented at the CCNC'22 conference by using a new set of time series monitored WiFi data frames with extended machine learning algorithms. We instrument a testbed of 10 IoT devices and conduct a detailed evaluation using multiple machine learning techniques for fingerprinting, achieving high accuracy up to 95% in identifying what IoT devices exist and their working status. Compared with our previous work in , the new approach could achieve IoT device profiling much quicker while maintaining the same level of classification accuracy. Moreover, the experimental results show that outside intruders can significantly harm the IoT devices without joining a WiFi network and can launch the attack within a minimum time without leaving any detectable footprints.