The research aims to examine the viability and effectiveness of C2aaS in enhancing data security and cloud performance, therefore filling present research gaps. The method follows a novel path using Multi-Head Self-Attention and Bacterial Colony Optimisation (BCO) inside a cloud computing environment. By letting simultaneous focus on several parts of input sequences, Multi-Head Self-Attention maximises resource allocation depending on different work and data needs. Inspired by bacterial foraging activity, BCO constantly adapts to changing conditions, provides best distribution of computational resources, and solves system overload problems. The results show the double use of the model in optimal resource allocation and secured threat classification. The self-attention approach helps to combine threat classification, thereby enabling the identification and categorisation of security issues in line with C2aaS’s objectives. The results reveal how realistically the proposed approach could raise data security and cloud performance.