Online signature analysis plays a vital role in today’s digital landscape, where the number of digital transactions and the need for identity verification are constantly increasing. The signature encompasses various aspects of an individual’s unique characteristics, including both structural and behavioural elements. Researchers have been intrigued by the intricacies surrounding signature verification and generation for quite some time. The main objective of this study is to present a fresh approach for creating online signatures that improves security in digital transactions and signature verification systems. This is accomplished by utilising the proposed Sign-Diffusion framework. Traditional approaches to signature recognition often lack the required resilience, as they are trained on datasets that include fraudulent attempts to replicate the original signature. To overcome the challenges and limitations generated because of manual mimicking capability, we propose a solution that has conditional diffusion as a building block aided with a state space model to capture long-term structural forecasting for online signature generation. This method tackles the task of extracting detailed features that capture the intricate spatial and temporal characteristics of signature dynamics. It also ensures adaptability to various signing styles. The approach we have developed introduces a foundational model that is capable of generating near-equal user dependent online signatures. This will also help to identify deepfakes in the area of online signature generation. The above mentioned facts highlights the prospects for further exploration in the area of online signature verification methods, specifically for detection of system-generated forgeries that are more advanced and sophisticated than manually created forgeries.