Applications often show unique communication behavior. Knowledge about this behavior is beneficial in various use cases, such as anomaly or dependency detection. In this paper, we present network profiles that characterize typical application behavior. This requires a reliable and accurate association of processes and applications, which is challenging. We, therefore, introduce an eBPF-based matcher for this task that enables the creation of network profiles. In our evaluation we show that eBPF allows us to efficiently collect the relevant data to build application profiles, addressing issues of other data collection approaches. We further evaluate our work by using a network profile to identify emulated botnet activity masqueraded as a benign process.     «

Applications often show unique communication behavior. Knowledge about this behavior is beneficial in various use cases, such as anomaly or dependency detection. In this paper, we present network profiles that characterize typical application behavior. This requires a reliable and accurate association of processes and applications, which is challenging. We, therefore, introduce an eBPF-based matcher for this task that enables the creation of network profiles. In our evaluation we show that eBPF...     »