As the Internet becomes increasingly integral to the Internet of Things trend, the risks of cyberattacks are rising in both frequency and severity. Traditional Intrusion Detection/Prevention Systems struggle to detect and prevent novel and constantly evolving attacks. Consequently, there is a need for an automated solution to enhance the ability to recognize and respond to these emerging threats. In this paper, we propose an approach for automatic rule generation for Intrusion Prevention Systems (IPS) based on deep learning models. This proposed approach enabled in-depth analysis of attack network flow data to generalize patterns into relevant Indicators of Compromise and automatically generate the corresponding attack prevention rules. Our experimental results with the Snort IPS demonstrate that the proposed approach provides prompt and effective responses to various types of real-world attacks