TASA : A Virtual AI Assistant with Multilevel Authentication Using Face Recognition

Virtual assistants are becoming increasingly popular in our daily lives as they provide convenience and efficiency in completing various tasks. While virtual assistants offer numerous benefits, such as hands-free control and personalized assistance, they also present security risks that must be addressed to protect sensitive user information. TASA (Trusted Assistant with Secure Access) is an intelligent virtual assistant that validates users in real-time using face recognition before providing access to its key functionalities. The system can differentiate between the owner (admin), sub-admin, and guest users based on their access allowance and act accordingly. Existing virtual assistants lack user authentication, making them vulnerable to attacks that compromise sensitive information. TASA addresses this problem by providing full authority to its owner/admin, ensuring data safety, trust, and transparency among users. TASA employs a Convolutional Neural Network (CNN) that implements the Histogram of Oriented Gradients (HOG) algorithm as a feature descriptor. The model is trained on the user's images to register them with the system. TASA captures the video feed and extracts the user's face by capturing images from the video and eliminating unnecessary information to mitigate the risk of fake access using static images. Additionally, TASA provides an extra layer of security on top of the face recognition system by assigning a unique secret passphrase to each user to prevent unauthorized access. TASA showcases high accuracy with over 90% in user authentication, ensuring secure and convenient user interactions. TASA's hardware components exhibit reliability and efficiency, enabling fast and accurate processing of data inputs and outputs. Additionally, the software components, including the API for database operations, facilitate seamless storage, retrieval, and management of user information.