Recent trends and clustered technology in the IT industries create usage of various hand-held computing devices like mobile phones, digitizer, modem, tablets etc. The Mobile phones are pervasive and inevitable computing devices used for different purposes such as making calls, chart application, games, Online shopping, banking and multimedia broadcasting. To improve its services, many applications are communicating with each other. Example, the banking application may read the OTP message from the messenger automatically leading to security issues. The authorized applications installed in the mobile phones are combined as a group and monitored as secure group communications. Each application in the mobile phone is having its Private Key which is called as Member Key and a common key is shared to all the members in the group which is called as group key. To enhance the security of the group communication, the messages are encrypted using group key and decrypted with member key. In dynamic conditions, the fresh group key is generated whenever there is a change in group strength, directly depending on the computation cost and communication cost because of number of multiplication and key sharing to all the members of the group. Though the Tree Based Key Management System has used to maintain lower computation and communication complexity, this research proposes the novel technique which combines the Tree Based Key Management and Machine Learning Techniques to reduce the key generation cost and correspondingly enrich the key security for information sharing.