Social media’s emergence has made it possible for people to share knowledge widely, but it has also created opportunities for malevolent influence efforts that jeopardise public confidence and cybersecurity. Because these coordinated tactics are large-scale, dynamic, and clandestine, it is still difficult to identify them early. The proposed work uses Graph neural networks (GNNs) along with multi-objective optimization to detect the malicious behavior of the users. Long Short-Term Memory (LSTM) networks are used to identify the dependency between user’s behaviour and time period. Experiments on real Twitter datasets show that the recommended approach performs better than traditional baselines in terms of influence campaign identification time and detection accuracy. The framework’s scalability and real-time capabilities enable proactive countermeasures that enhance cybersecurity and information integrity in social media ecosystems.