The coordination between aerial and ground nodes has enhanced the versatility and quality of the traditional networks. The application of aerial systems in mission-critical operations, as well as civilian applications, brings in the context of safeguarding unmanned aerial systems (UAS) from malicious attackers. This study discusses the threats and attacks mounted on UAS, alongside the challenges introduced by the unmanned aerial vehicle (UAV) network structure itself. A framework for safeguarding UAS against malicious attackers and recovering the rogue UAVs is proposed in the study. The proposed framework enforces a dynamic conceptual grid-based layout over the actual geographical deployment. The dynamically shuffling grid ascertains the security of transmission channels, as every time the grid is shuffled periodically or based on abnormal behaviour, the safety paradigm is reinitiated. Public key cryptographic algorithms are deployed for securing the communication links. Neural networks-based predictions are used for detecting abnormality in behavioural, statistical, and mobility patterns. Principal component analysis based on multivariate statistical analysis is used for detecting outliers in the aerial network environment. The behaviour prediction and outlier detection algorithms significantly improve the overall performance of the network and provide immunity against the intruders with reduced false positives, high accuracy, and better detection rate.