Glaucoma, an irreversible neurodegenerative disorder, can lead to vision loss and blindness. Visual field (VF) tests are crucial for quantifying functional damage in glaucoma, but the tests are time-consuming and the results have high variations, influenced by subjective behaviors and psychological states of patients. Consequently, predicting VF test results using objective, non-invasive, reproducible optical coherence tomography (OCT) data coupled with deep learning modeling is promising in improving clinical care. However, existing methods only focus on predicting single VF indicators, such as threshold sensitivities or deviation maps, and have poor performance in severe glaucoma. Since different VF indicators are correlated, developing a joint prediction model is beneficial. Furthermore, the availability of more VF test data than corresponding OCTs in most datasets poses a challenge in utilizing unpaired VF test data. This study proposes a multi-modal, multi-task learning framework based on OCT data for VF prediction. We introduce a dynamic weighted loss function to improve prediction performance for eyes with severe glaucoma. Additionally, we construct a novel PairMatcher model for augmenting unpaired VF data. Extensive experiments demonstrate that our framework outperforms existing methods, showcasing its potential for VF prediction in glaucoma.

Configuring a network is a tedious and error-prone task. In particular, configuring routing policies for a network is complex as it involves subtle dependencies in multiple routers across the network. Misconfigurations are common and certain misconfigurations can bring the Internet down. In 2005, a misconfigured router in AS 9121 blackholed traffic for tens of thousands of networks in the Internet. This paper describes NetPiler, a system that detects router misconfigurations. NetPiler consists of a routing policy configuration model and a misconfiguration detection algorithm. The model is applicable to routing policies configured on a single router as well as to network-wide configuration. Using the model, NetPiler detects configuration commands that do not influence the behavior of the network - we call these configurations ineffective commands. Although the ineffective commands could be benign, sometimes when the commands are mistakenly configured to be ineffective, they cause the network to misbehave deviating from the intended behavior. We have implemented NetPiler in approximately 128,000 lines of C++ code, and evaluated it on the configurations of four production networks. NetPiler discovers nearly a hundred ineffective commands. Some of these misconfigurations can result in loss of connectivity, access to protected networks, and financial implications by providing free transit services. We believe NetPiler can help networks to significantly reduce misconfigurations.

As a network evolves over time, multiple operators modify its configuration, without fully considering what has previously been done. Similar policies are defined more than once, and policies that become obsolete after a transition are left in the configuration. As a result, the network configuration becomes complicated and disorganized, escalating maintenance costs and operator faults. We present a method called NetPiler, which groups common policies by discovering a set of shared features and which uses the groupings for the configuration instead of using each individual policy. Such an approach removes redundancies and simplifies the configuration while preserving the intended behavior of the configuration. We apply NetPiler to the routing policy configurations from four different networks, and reduce more than 50% of BGP communities and the related commands. In addition, we show that the reduced community definitions are sufficient to satisfy changes as the network evolves over nearly two years.

Router configuration is a difficult and complex task. At the same time, it is a crucial task as it accounts for a network's profit, performance, and security. Routing policies are configured in low-level languages and the high- level intent is hard to decipher. In this paper, we propose a system, called NetPolis, which abstracts the high-level intents from low-level configuration of routing policies. The goal of NetPolis is to automatically generate the inter- domain routing policy configuration of a network. NetPolis takes the network's router configuration files and compares the import and export policies from various perspectives and granularities. The output is a multi-level model to represent neighbor networks with similar routing policy enforcements. We validate our approach by applying NetPolis to the router configuration files from a production network of a major ISP. The network operator confirms that NetPolis provides a compact summary of the network's routing policies and this summary helps to verify high-level intents, to identify misconfigurations, and to aid in policy modifications.

Configuring a network is a low-level, device-specific task. Many have compared it to writing a distributed program in assembly language, reserved only for highly experienced network operators. Automation has been proposed by researchers and industry as the solution to problems in network configuration. However, there is a certain amount of resistance from the operator community against automation. On the one hand, operators do desire a way for network-wide configuration. On the other hand, they still like to have access and control to details, to ensure flexibility and for debugging. In this paper, we attempt to answer the question "How should we automate network configuration" by studying where the complexity lies in network configuration. With an operational perspective, using data from three different types of production networks, we analyze the configuration files from these networks over the span of up to two years. Our analysis shows that the majority of changes to these files are a few lines each and made frequently. We found that routing, especially its policies, constitute a significant portion of the configuration files, as well as modifications to them. We then present complexity models to measure network-wide risk, impact and duplication of routing policies in network configuration. We show that risk and impact tend to grow over time, and the duplication factor is high. Based on the results of our analysis, we propose ways to automate the complex parts of network configuration.

Routing infrastructure plays a vital mle in the Internet, and attacks on routers can be damaging. Compromised routers can drop, modih, mislforward or reorder valid packets. Existing proposals for secure forwarding require substantial computational overhead and additional capabilities at routers. We propose Secure Split Assignment Trajectory Sampling (SATS), a system that detects malicious routers on the data plane. SATS locates a set of suspicious routers when packets do not follow their predicted paths. It works with a traffic measurement platform using packet sampling, has low overhead on routers and is applicable to high-speed networks. Different subsets ofpackets are sampled over dzyerent groups of routers to ensure that an attacker cannot completely evade detection. Our evaluation shows that SATS can signzjicantly limit a malicious router's harm to a small portion of traffic in a network.

Diagnosing inter-domain routing problems in the Internet is hard. BGP, the defacto inter-domain glue, is designed for routing, not diagnosis. It is extremely chatty - the most minor connectivity change produces hundreds of BGP messages and a major peering loss can generate millions - and making sense of the deluge of data remains challenging. We have developed statistical techniques to extract the large-scale structure of BGP events and visualization techniques to display that structure in operationally meaningful ways. These tools can be used to detect routing anomalies in real-time. We show case studies of routing instabilities at a Tier-1 ISP and a large institutional network, automatically diagnosed by our tools. We present drawbacks in using BGP events alone to understand inter-domain routing, and discuss how to solve them through the integration of additional data sources.

We propose a system that improves the performance of streaming media CDN by exploiting the path diversity provided by existing CDN infrastructure. Path diversity is provided by the different network paths that exist between a client and its nearby edge servers; and multiple description (MD) coding is coupled with this path diversity to provide resilience to losses. In our system, MD coding is used to code a media stream into multiple complementary descriptions, which are distributed across the edge servers in the CDN. When a client requests a media stream, it is directed to multiple nearby servers which host complementary descriptions. These servers simultaneously stream these complementary descriptions to the client over different network paths. This paper provides distortion models for MDC video and conventional video. We use these models to select the optimal pair of servers with complementary descriptions for each client while accounting for path lengths and path jointness and disjointness. We also use these models to evaluate the performance of MD streaming over CDN in a number of real and generated network topologies. Our results show that distortion reduction by about 20 to 40% can be realized even when the underlying CDN is not designed with MDC streaming in mind. Also, for certain topologies, MDC requires about 50% fewer CDN servers than conventional streaming techniques to achieve the same distortion at the clients.

Scalability of multicast forwarding state is likely to be a major issue facing inter-domain multicast deployment. We present a comprehensive analysis of the multicast forwarding state problem. Our goal is to understand the scaling trends of multicast forwarding state in the Internet, and to explore the intuitions that have motivated state reduction research. We conducted simulation experiments on both real and generated network topologies, with a range of parameters driven by multicast application characteristics. We found that the increase in peering among Internet backbone networks has led to more multicast forwarding state at a handful of core domains, but less state in the rest of the domains. We observed that scalability of multicast forwarding state with respect to session size follows a power law. Our findings show that distribution and concentration of multicast forwarding state in the Internet is significantly, impacted by the application characteristics. We investigated the proposals on non-branching multicast forwarding state elimination, and found substantial reduction is attainable even with very dense multicast sessions.

The efficiency of using multicast in multi-party applications is constrained by preference heterogeneity, where receivers range in their preferences for application data. We examine an approach in which approximately similar sources and receivers are clustered into multicast groups. The goal is to maximize preference overlap within each group while satisfying the constraint of limited network resources. This allows an application to control the number of multicast groups it uses and thus the number of connections it maintains. We present a clustering framework with a two-phase algorithm: a bootstrapping phase that groups new sources and receivers together, and an adaptation phase that re-groups them in reaction to changes. The framework is generic in that an application can customize the algorithm according to its requirements and data characteristics. We conducted detail simulation experiments to study various issues and tradeoffs in applying clustering to different preference patterns and application classes. We found that clustering successfully exploits preference similarity and utilizes network resources more efficiently than when it is not used. Also, application-level hints can be incorporated in our algorithm, which are instrumental in the creation of an effective grouping of sources and receivers. Our algorithm handles changes dynamically, and also limits multicast "join" and "leave" disruption to the application.