Campus networks have recently experienced a proliferation of devices ranging from personal use devices (e.g. smartphones, laptops, tablets), to special-purpose network equipment (e.g. firewalls, network address translation boxes, network caches, load balancers, virtual private network servers, and authentication servers), as well as special-purpose systems (badge readers, IP phones, cameras, location trackers, etc.). To establish directives and regulations regarding the ways in which these heterogeneous systems are allowed to interact with each other and the network infrastructure, organizations typically appoint policy writing committees (PWCs) to create acceptable use policy (AUP) documents describing the rules and behavioral guidelines that all campus network interactions must abide by. While users are the audience for AUP documents produced by an organization's PWC, network administrators are the responsible party enforcing the contents of such policies using low-level CLI instructions and configuration files that are typically difficult to understand and are almost impossible to show that they do, in fact, enforce the AUPs. In other words, mapping the contents of imprecise unstructured sentences into technical configurations is a challenging task that relies on the interpretation and expertise of the network operator carrying out the policy enforcement. Moreover, there are multiple places where policy enforcement can take place. For example, policies governing servers (e.g., web, mail, and file servers) are often encoded into the server's configuration files. However, from a security perspective, conflating policy enforcement with server configuration is a dangerous practice because minor server misconfigurations could open up avenues for security exploits. On the other hand, policies that are enforced in the network tend to rarely change over time and are often based on one-size-fits-all policies that can severely limit the fast-paced dynamics of emerging research workflows found in campus networks. This dissertation addresses the above problems by leveraging recent advances in Software-Defined Networking (SDN) to support systems that enable novel in-network approaches developed to support an organization's network security policies. Namely, we introduce PoLanCO, a human-readable yet technically-precise policy language that serves as a middle-ground between the imprecise statements found in AUPs and the technical low-level mechanisms used to implement them. Real-world examples show that PoLanCO is capable of implementing a wide range of policies found in campus networks. In addition, we also present the concept of Network Security Caps, an enforcement layer that separates server/device functionality from policy enforcement. A Network Security Cap intercepts packets coming from, and going to, servers and ensures policy compliance before allowing network devices to process packets using the traditional forwarding mechanisms. Lastly, we propose the on-demand security exceptions model to cope with the dynamics of emerging research workflows that are not suited for a one-size-fits-all security approach. In the proposed model, network users and providers establish trust relationships that can be used to temporarily bypass the policy compliance checks applied to general-purpose traffic -- typically by network appliances that perform Deep Packet Inspection, thereby creating network bottlenecks. We describe the components of a prototype exception system as well as experiments showing that through short-lived exceptions researchers can realize significant improvements for their special-purpose traffic.