With the rapid expansion of cybercrime, attackers are exploiting vulnerabilities in cloud computing and network infrastructures, posing significant security threats. Traditional Intrusion Detection Systems (IDS) struggle to cope with the dynamic and sophisticated nature of cyber-attacks, necessitating the development of intelligent and adaptive security techniques. Machine learning (ML) has emerged as a powerful tool in cybersecurity, offering improved detection rates, reduced false alarms, and lower computational costs. ML techniques have been applied to various cybersecurity domains, including intrusion detection, malware classification, spam filtering, and phishing detection. While ML cannot fully automate cybersecurity systems, it enhances threat detection efficiency, alleviating the burden on security analysts. This study proposes an intelligent network attack detection framework utilizing deep learning models. The Cyber-Physical System (CPS) is represented as a coordinated network of agents, with one agent acting as a leader, guiding the others. The attack detection phase employs deep neural networks to identify threats in their early stages, ensuring a proactive defense mechanism. To further enhance security, robust control algorithms are integrated to isolate compromised agents using a reputation-based mechanism. Experimental results demonstrate that deep learning techniques significantly outperform traditional IDS methods in detecting and mitigating network attacks. This approach improves cybersecurity by making threat detection more efficient, proactive, and cost-effective, addressing the limitations of conventional security mechanisms.