Dynamic group key agreement protocols are cryptographic primitives to provide secure group communications in decentralized and dynamic networks. Such protocols provide additional operations to update the group key while adding new participants into the group and removing existing participants from the group without re-executing the protocol from the beginning. However, the lack of scalability emerges as one of the most significant issues of dynamic group key agreement protocols when the number of participants in the group increases. For instance, frequent participant join requests for large groups may cause an effect similar to a Distributed Denial of Service (DDoS) attack and violate the system availability due to the increase in group key update time. Therefore, analyzing the scalability of dynamic group key agreement protocols is crucial to detect conditions where the system becomes unavailable. In this paper, we propose an analytical performance model to evaluate the scalability of dynamic group key agreement protocols by using queueing models. We also extend our performance model for evaluating the scalability of secure file sharing systems that utilize group key agreement protocols. Moreover, we present a demonstrative use case to show the applicability of our performance model on an example group key agreement protocol and a secure file sharing system.
A method for detecting remote data exfiltration from computer networks is described, capable of recognizing patterns of exfiltration occurring over days to weeks. Normal traffic flow data, in the form of egress and ingress bytes over time, is used to train an ensemble of semi-supervised learners. The detection ensemble is modular, with individual classifiers trained on different traffic features thought to characterize malicious data transfers. We select features that model the egress to ingress byte balance over time, periodicity, and short time-scale irregularity of the traffic. The features are most efficiently modeled in the frequency domain, which has the added benefit that variable duration flows are transformed to a fixed-size feature vector, and by sampling the frequency space appropriately, arbitrarily-long flows can be tested. When trained on days- or weeks-worth of internet traffic from individual hosts, our ensemble achieves a 1% false positive rate. When tested on simulated exfiltration samples with a variety of different timing and data egress characteristics, the ensemble was generally successful at detecting exfiltration that is not simultaneously ingress-heavy, connection-sparse, and short duration---a combination that is not optimal for attackers seeking to transfer large amounts of data. The method is tested on a variety of systems, performing best on client workstations and worst, and not recommended for, outward facing servers. The modular ensemble can be customized to target exfiltration of different types or sophistication, or even different kinds of anomalous traffic.