Interdependent systems confront rapidly growing cybersecurity threats. This paper delves into the realm of security decision-making within these complex interdependent systems. We design a resource allocation framework to improve the security of interdependent systems managed by a single rational (logical) defender. Our framework models these systems and their potential attack vulnerabilities using the notion of cyber attack graphs. We propose four defense mechanisms, incorporating a popular network analysis algorithm called PageRank which is used to identify the importance of different critical assets in the system. These mechanisms stem from existing graph theories widely used in graphical models (including Adjacent Nodes, In-degree Nodes, Min-Cut Edges, and Markov Blanket). We adopt two versions of the PageRank algorithm to extract useful information about the attack graphs we use. Our approaches show low sensitivity to the number of concurrent attacks launched over interdependent systems. We evaluate our decision-making framework via ten attack graphs, which include multiple real-world interdependent systems. We quantify the level of security improvement under our defense methods compared to four well-known resource allocation algorithms and other proposed approaches. Our proposed framework leads to better resource allocations compared to these algorithms in most test cases. According to our results and statistical tests, our defense resource allocation approach’s outcomes are superior. Our framework enhances security decision-making under various circumstances. Moreover, We release the full implementation of our framework for the research community to leverage it and build on it with new methods and datasets.