Spamming is the act of abusing an electronic messaging system by sending unsolicited bulk messages. Filtering of these messages is merely another line of defence and does not prevent spam messages from circulating in email systems. This problem causes users to distrust email systems, suspect even legitimate emails and leads to substantial investment in technologies to counter the spam problem. Spammers threaten users by abusing the lack of accountability and verification features of communicating entities. To contribute to the fight against spamming, a cloud-based system that analyses the email server logs and uses predictive analytics with machine learning to build trust identities that model the email messaging behavior of spamming and legitimate servers has been designed. The system constructs trust models for servers, updating them regularly to tune the models. This study proposed that this approach will not only minimize the circulation of spam in email messaging systems, but will also be a novel step in the direction of trust identities and accountability in email infrastructure.