Improving resilience of energy nexus across demand sectors: Strengthening cybersecurity in water cycle and energy networks through data reconstruction using a novel hybrid reinforcement learning and bidirectional long short-term memory

• Presenting a novel cybersecurity measures in energy nexus. • Revealing the role of water demand data accuracy in reliability of energy nexus. • Underscoring the impact of precise demand data in cybersecurity. • Investigating false data injection (FDI) attacks in drinking energy nexus. • Restoring compromised demand data post-attack utilizing hybrid RL and BiLSTM. Smart farms, serving as models of smart energy hubs (SEHs), are evolving to integrate energy sources such as wind turbines (WTs) and gas network for electrical and thermal demands, alongside water desalination systems to ensure a sustainable freshwater supply. These systems play a crucial role in the anthropogenic water cycle, addressing treatment processes for water management while enhancing resource recovery. However, as critical nodes in the demand chain, they face increasing risks from sophisticated cyber threats that can compromise resource efficiency, stability, and environmental sustainability. This study explores the resilience of SEHs that incorporate electrical storage (ES), thermal storage (TS), and water storage (WS) facilities, highlighting their representation as energy hubs (EHs) for optimized resource management. A novel undetectable false data injection (FDI) attack is simulated, targeting thermal and water demands to exploit system vulnerabilities and destabilize the demand chain. To mitigate these risks, a hybrid approach combining reinforcement learning (RL) and bidirectional long short-term memory (BiLSTM) networks is proposed to reconstruct incorrect data. The results demonstrate that FDI attacks can cause significant deviations in thermal (up to 17.1 %) and water demands (up to 17.7 %), threatening the operational stability of the demand chain. Implementing the proposed correction method reduces the mean absolute percentage error (MAPE) to 2.3884 % for thermal demands and 5.0803 % for water demands, showcasing its efficacy in stabilizing the demand chain. This study highlights the importance of effective drinking water treatment and distribution within smart farming contexts. By utilizing cutting-edge computational intelligence methods and combining renewable energy solutions with in-depth data analysis, this study enhances environmental restoration initiatives and strengthens risk assessment approaches to address cyber threats in water resource management. Ultimately, it emphasizes the need for innovative solutions that enhance the sustainability of water resources while ensuring resilience against emerging threats in both agricultural and urban settings.