An innovative framework for secure data transmission using machine learning based classification and ElGamal encryption with Ramanujan primes

The secure processing and transmission of sensitive data has arisen as a crucial concern in the modern digital world. This paper presents a conditional ElGamal framework in which the prime modulus is chosen from either conventional primes or the Ramanujan prime based on data sensitivity, while the encryption and decryption techniques remain identical to the standard ElGamal scheme. To classify data into normal and highly sensitive categories, several machine learning models are evaluated among which the Support Vector Machine model achieves highest mean accuracy under 5-fold cross-validation. The normal sensitive data is encrypted using the standard ElGamal encryption method whereas highly sensitive data is encrypted by using the proposed variant with Ramanujan prime-based key generation. The security analysis confirmed that the proposed framework preserves the security of standard ElGamal scheme under known-plain text attack and chosen-plain text attack. The choice of Ramanujan primes affects only the key-generation procedure and does not change or strengthen the theoretical security guarantees of ElGamal scheme. To ensure both data credibility and reliability, a hash-based message authentication code is appended to each message before transmission. The encryption-decryption time and the average data rate for highly sensitive data is less compared to the normal sensitive data which ensures the less CPU memory usage of the new framework. Hence, the proposed framework can be suitable for applications in cloud computing, healthcare, and e-governance environments.