Multimodal Bio Cryptography for Securing Cloud Computing using Convolutional Neural Network

Cloud computing has grown in popularity as a storage platform that makes use of the internet to provide users with large amounts of data storage. The cloud's client base has grown throughout time, and access for consumers and businesses has expanded. At this time, security concerns develop among the cloud community. Different harmful assaults occur on well-known cloud services by exploiting weak password providers, causing them to lose confidentiality. Cryptography has evolved into an autonomous data encryption method, which allows users to manually cipher/decipher information. Secure keys can be generated using biometric key generation, eliminating the need for remembering. However, usability concerns persist since biometric credentials are required every time a file is encrypted or decrypted. This study focuses on multimodal biometrics, which will create a non-intrusive key for cryptography algorithms to encrypt/decrypt the file uploaded to the cloud by the client. To achieve this, this paper presents a convolutional neural networks approach which extract features from face and fingerprint biometrics and trains them to understand the patterns that are to be classified. Multi-label categorization features were employed as output labels, which were hashed to 512 bits from each of the multimodal biometrics and could be combined to yield 1024 bits for RSA cryptography encryption and decryption. False acceptance and rejection rates were utilized to assess the whole system. As a result, the face exhibits 2.13% FAR and the fingerprint 2.71% FAR on average.