Abstract
Smart speakers, such as Alexa and Google Home, support daily activities in smart home environments. Even though voice commands enable friction-less interactions, existing financial transaction authorization mechanisms hinder usability. A non-invasive authorization by leveraging presence and light sensors’ data is proposed in order to replace invasive procedure through smartphone notification. The Coloured Petri Net model was created for synthetic data generation, and one month data were collected in test bed with real users. Random Forest machine learning models were used for smart home behavior information retrieval. The LSTM prediction model was evaluated while using test bed data, and an open dataset from CASAS. The proposed authorization mechanism is based on Physical Unclonable Function usage as a random number generator seed in a Challenge Response protocol. The simulations indicate that the proposed scheme with specialized autonomous device could halve the total response time for low value financial transactions triggered by voice, from 7.3 to 3.5 s in a non-invasive manner, maintaining authorization security.
Highlights
Smart assistants are making natural interactions through voice commands that are accessible in smart home environments
Energy and computing power of constrained Internet of Things (IoT) devices limit asymmetric cryptography methods use in a smart home scenario, as these methods allow for secure mechanisms implementation, but at high computational expense [23]
The proposed implicit authentication mechanisms for IoT environments can be found in the literature: some are based only on smartphone data [22,53], while others rely on machine learning applied to WiFi signals [54,55] in order to authenticate IoT devices
Summary
Smart assistants are making natural interactions through voice commands that are accessible in smart home environments. Voice interactions provide pervasive services by integration with Internet of Things (IoT) devices tgar are present in smart homes. Existing authorization solutions for these transactions are based upon invasive user authentication through smartphone notification [11], which hinders usability by increased friction in these interactions. The research question is: how to authorize low-value (up to $50) financial transactions that are triggered by voice in smart home environments, in an non-invasive manner (i.e., not requiring additional user interaction other than voice command)?.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
