Search

Visual inspection systems (VISs) are one of the key technologies needed for mass production in the manufacturing industry. Fast and accurate algorithms are required for quality control of large quantities of products. If the manufactured products change over time inspection of the products becomes a challenging task. In this paper, Split-Brain Convolutional Autoencoder approach to detect and localize defects without defective samples is proposed. Two disjointed Convolutional Autoencoder networks are employed to predict the subchannel of the image from another subchannel. The reconstruction residual maps generated from each subchannel are combined and then thresholded to produce the segmented out image patches as the ultimate result. This approach greatly alters the reconstruction abilities of Convolutional Autoencoders. Experimental results achieved using a challenging dataset have shown that this approach can achieve good defect detection and defective region segmentation results.

The performance and operational characteristics of the Domain Name System (DNS) protocol are of deep interest to the research and network operations community. In this paper, we present measurement results from a unique dataset containing more than 26 billion DNS query-response pairs collected from more than 600 globally distributed recursive DNS resolvers. We use this dataset to reaffirm findings in published work and notice some significant differences that could be attributed both to the evolving nature of DNS traffic and to our differing perspective. For example, we find that although characteristics of DNS traffic vary greatly across networks, the resolvers within an organization tend to exhibit similar behavior. We further find that more than 50% of DNS queries issued to root servers do not return successful answers, and that the primary cause of lookup failures at root servers is malformed queries with invalid top-level domains (TLDs). Furthermore, we propose a novel approach that detects malicious domain groups using temporal correlation in DNS queries. Our approach requires no comprehensive labeled training set, which can be difficult to build in practice. Instead, it uses a known malicious domain as anchor and identifies the set of previously unknown malicious domains that are related to the anchor domain. Experimental results illustrate the viability of this approach, i.e., we attain a true positive rate of more than 96%, and each malicious anchor domain results in a malware domain group with more than 53 previously unknown malicious domains on average.

Self-supervised methods are interesting for remote sensing because there are not many human labeled datasets available, but there is practically unlimited amount of data that can be used for self-supervised learning. In this paper we analyze the use of split-brain autoencoders in the context of remote sensing image classification. Weinvestigate the importance of training set size, choice of color space and size of the model to the classification accuracy. We show that even with small amount of unlabeled training images, if we finetune the weights learned by the autoencoder, we can achieve almost state of the art results of 89.27% on AID dataset.

The Domain Name System (DNS) is a critical Internet service, which translates easily memorized domain names to numerical IP addresses for locating computer resources and services. In this paper, we try to explore the behaviors of DNS lookup by mining DNS logs from three primary DNS servers in a large university campus network in China. Our dataset is made up of two parts, namely DNS query logs and messages received or send by DNS servers. Firstly, through analyzing these DNS query logs, we are able to understand the overall trend of users’ surfing. For dealing with huge DNS dataset, we introduce an algorithm we call DNSReduce , which can be used to dig out top 10 client IP addresses and top 10 destination domain names efficiently. Moreover, we make comparative analysis of lookup behavior between wired and wireless users. Secondly, with messages received or send by DNS servers we can find these DNS servers׳ behaviors, i.e., TTLs, equivalent answers and are able to accurately identify domain names with dynamic IP addresses. We provide different and specific visualization techniques for presenting these analysis results and show these techniques are very useful for understanding user behaviors, analyzing security events and characterizing overall tendency in campus network management.

We propose split-brain autoencoders, a straightforward modification of the traditional autoencoder architecture, for unsupervised representation learning. The method adds a split to the network, resulting in two disjoint sub-networks. Each sub-network is trained to perform a difficult task -- predicting one subset of the data channels from another. Together, the sub-networks extract features from the entire input signal. By forcing the network to solve cross-channel prediction tasks, we induce a representation within the network which transfers well to other, unseen tasks. This method achieves state-of-the-art performance on several large-scale transfer learning benchmarks., Accepted to CVPR 2017

The rapid development of High Availability (HA) system has attracted growing attention from both industry and academia. Due to the network and hardware failure, a complete system would split into two or more separate partitions, which begin to compete for shared resources, resulting in system chaos and data corruption. In this paper, we propose a novel weight-based leader election approach for split brain in a distributed system. The leader of separate partitions would be elected by embedding an arbitration program in the client. Unlike the traditional mode, the leader election in our proposed approach is lightweight and fully automatic, meaning that no additional hardware resources or manual intervention are required. The approach presented in this paper has been validated to be valid through qualitative and quantitative experiments.

Name resolution using the Domain Name System (DNS) is integral to today's Internet. The resolution of a domain name is often dependent on namespace outside the control of the domain's owner. In this article we review the DNS protocol and several DNS server implementations. Based on our examination, we propose a formal model for analyzing the name dependencies inherent in DNS. Using our name dependency model we derive metrics to quantify the extent to which domain names affect other domain names. It is found that under certain conditions, more than half of the queries for a domain name are influenced by namespaces not expressly configured by administrators. This result serves to quantify the degree of vulnerability of DNS due to dependencies that administrators are unaware of. When we apply metrics from our model to production DNS data, we show that the set of domains whose resolution affects a given domain name is much smaller than previously thought. However, behaviors such as using cached addresses for querying authoritative servers and chaining domain name aliases increase the number and diversity of influential domains, thereby making the DNS infrastructure more vulnerable.

Current Content Delivery Networks (CDNs) primarily utilize the Domain Name System (DNS) to direct users towards optimal content replica servers. Typically, these CDNs will leverage specialized DNS servers to perform load balancing, by providing different users with different IP endpoints for a particular hostname. However, there typically exists some form of DNS cache between the user and the authoritative DNS for a particular web service (Examples include web browsers, operating systems, and home routers). While these DNS caches typically improve user experience and network performance by reducing the number of redundant DNS requests, there exist scenarios where the presence of a DNS cache can actually harm network performance and user experience. For example: if a server or datacentre were to suddenly become unavailable, end users may become `stuck' trying to access the old IP address contained in their DNS caches, even though the authoritative DNS server may have updated it's records accordingly. In order to address this issue, we propose a system that enables network operators to transparently redirect users towards optimal servers during times of network congestion or high server load. By leveraging SDN and NFV, we are able to implement this system over a variety of network infrastructures, and to various levels of scale. Our initial tests indicate that our methodology has the capability to vastly improve user experience by several orders of magnitude during times of high network load, over a variety of network conditions. Our system is application- layer agnostic, and is compatible with various networking protocols.

For a user to access any resource on the Internet, it is necessary to first locate a server hosting the requested resource. The Domain Name System service (DNS) represents the first step in this process, translating a human readable name, the resource host name, into an IP address. With the expansion of Content Distribution Networks (CDNs), the DNS service has seen its importance increase. In a CDN, objects are replicated on different servers to decrease the distance from the client to a server hosting the object that needs to be accessed. The DNS service should improve user experience by directing its demand to the optimal CDN server. While most of the Internet Service Providers (ISPs) offer a DNS service to their customers, it is now common to see clients using a public DNS service instead. This choice may have an impact on Web browsing performance. In this paper we study the impact of choosing one DNS service instead of another and we compare the performance of a large European ISP DNS service with the one of a public DNS service, Google DNS. We propose a causal approach to expose the structural dependencies of the different parameters impacted by the DNS service used and we show how to model these dependencies with a Bayesian network. This model allows us to explain and quantify the benefits obtained by clients using their ISP DNS service and to propose a solution to further improve their performance.

The Domain Name System (DNS) is an essential part of the Internet infrastructure and provides fundamental services, such as translating host names into IP addresses for Internet communication. The DNS is vulnerable to a number of potential faults and attacks. In particular, false routing announcements can deny access to the DNS service or redirect DNS queries to a malicious impostor. Due to the hierarchical DNS design, a single fault or attack against the routes to any of the top-level DNS servers can disrupt Internet services to millions of users. We propose a path-filtering approach to protect the routes to the critical top-level DNS servers. Our approach exploits both the high degree of redundancy in top-level DNS servers and the observation that popular destinations, including top-level DNS servers, are well-connected via stable routes. Our path-filter restricts the potential top-level DNS server route changes to be within a set of established paths. Heuristics derived from routing operations are used to adjust the potential routes over time. We tested our path-filtering design against BGP routing logs and the results show that the design can effectively ensure correct routes to top-level DNS servers without impacting DNS service availability.

The Domain Name System (DNS) is a critical infrastructure of any network, and, not surprisingly a common target of cybercrime. There are numerous works that analyse higher level DNS traffic to detect anomalies in the DNS or any other network service. By contrast, few efforts have been made to study and protect the recursive DNS level. In this paper, we introduce a novel abstraction of the recursive DNS traffic to detect a flooding attack, a kind of Distributed Denial of Service (DDoS). The crux of our abstraction lies on a simple observation: Recursive DNS queries, from IP addresses to domain names, form social groups; hence, a DDoS attack should result in drastic changes on DNS social structure. We have built an anomaly-based detection mechanism, which, given a time window of DNS usage, makes use of features that attempt to capture the DNS social structure, including a heuristic that estimates group composition. Our detection mechanism has been successfully validated (in a simulated and controlled setting) and with it the suitability of our abstraction to detect flooding attacks. To the best of our knowledge, this is the first time that work is successful in using this abstraction to detect these kinds of attacks at the recursive level. Before concluding the paper, we motivate further research directions considering this new abstraction, so we have designed and tested two additional experiments which exhibit promising results to detect other types of anomalies in recursive DNS servers.

Behind the growth of the amount of broadband users connections in the world, the wide use of mobile devices as cell phones and PDAs and other accesses from anywhere throughout Internet, always require some level of availability to support customer needs. This necessity made high availability architectures be used on large scale and has been migrated to service providers which offer solutions and, furthermore, they ensure short time to repair. This paper addresses “high availability”, one of the basic requirements of today systems on the Internet, presenting the main concepts, drawbacks and an automaton for these drawbacks' solution. Based on the two main protocols for high availability (VRRP and CARP), in particular, two problems are addressed, which leads to the environments do not respond to requests for services: split brain and no-brain. These protocols will be fast presented, as well as the study on the basis of the problems mentioned and observed in a production environment. The diagnostic for split-brain and no-brain was determiner for the solution problem. This work proposes a protocol extension, based on formal language methods (Petri Net), for high availability architectures regarding the mentioned drawbacks on protocols which reside in that kind of architecture.

In Service Oriented Architecture (SOA), web services may span several sites or logical tiers, each responsible for some part of the service. Most services need to be highly reliable and should allow no data corruption. A known problem in distributed systems that may lead to data corruption or inconsistency is the partition problem, also known as the split-brain phenomena. A split-brain occurs when a network, hardware, or software malfunction breaks a cluster of computer into several separate sub-clusters that reside side by side and are not aware of each other. When, during a session, two or more of these sub-clusters serve the same client, the data may become inconsistent or corrupted. ITRA - Inter Tier Relationship Architecture [1] enables web services to transparently recover from multiple failures in a multi-tier environment and to achieve continuous availability. However, the ITRA protocol does not handle partitions. In this paper we propose an extension to ITRA that supports continuous availability under partitions. Our unique approach, discussed in this paper, deals with partitions in multi-tier environments using the collaboration of neighboring tiers.

The domain name system (DNS) provides a critical service for the Internet -mapping of user-friendly domain names to their respective IP addresses. Yet, there is no standard set of metrics quantifying the quality of domain name service or QoDNS, let alone a thorough evaluation of it. This paper attempts to fill this gap from the perspective of a DNS proxy/cache, which is the bridge between clients and authoritative servers. We present an analytical model of DNS proxy operations that offers insights into the design tradeoffs of DNS infrastructure and the selection of critical DNS parameters. After validating our model against simulation results, we extend it to study the impact of DNS cache poisoning attacks and evaluate various DNS proposals with respect to the QoDNS metrics. In particular, we compare the performance of two newly proposed DNS security solutions: one based on cryptography and one using collaborative overlays.

This paper provides the mechanism for the autoconfiguration of recursive DNS server in mobile node and the optimization of DNS name resolution in the hierarchical mobile IPv6 network. Whenever the mobile node moves into a new MAP domain, the region managed by another MAP, in the hierarchical mobile IPv6 network, it detects the addresses of recursive DNS servers which are placed in the region and replaces the old ones with the new ones for DNS name resolution. This allows the time for DNS name resolution much reduced by using the nearest recursive DNS server which exists in the region. Therefore, through DNS autoconfiguration, the mechanism of this paper can optimize DNS name resolution.

The increasing deployment of overlay networks that rely on DNS tricks has led to added interest in examining DNS traffic. In this paper we report on a characterization of DNS traffic gathered over a period of several weeks at Internet Gateway Routers (IGRs) in the AT&T Common Backbone. The characterization is carried out using several novel techniques to identify clients, local DNS servers, and authoritative DNS servers. Our techniques include passive and active measurements, graph-based analysis, examination of outliers, and explicit checks against data obtained from several external sources. Our contribution is the reduction of a very large data set (over 1 terabyte of raw data) into a significantly smaller representation that is ideally suited for answering protocol-specific semantic queries quickly. After categorizing the addresses, we use the network aware clustering technique to group local DNS servers. By juxtaposing the DNS server clusters with clusters formed by Web clients obtained from a large portal Web site, we determine the distribution of identified DNS servers in busy clusters. A variety of applications are examined, ranging from identifying suspected zombies to helping Content Distribution Networks in mapping location of DNS servers.

Over a million computers implement the Internet's Domain Name System of DNS, making it the world's most distributed database and the Internet's most significant source of wide-area RPC-like traffic. Last year, over eight percent of the packets and four percent of the bytes that traversed the NSFnet were due to DNS. We estimate that a third of this wide-area DNS traffic was destined to seven root name servers. This paper explores the performance of DNS based on two 24-hour traces of traffic destined to one of these root name servers. It considers the effectiveness of name caching and retransmission timeout calculation, shows how algorithms to increase DNS's resiliency lead to disastrous behavior when servers fail or when certain implementation faults are triggered, explains the paradoxically high fraction of wide-area DNS packets, and evaluates the impact of flaws in various implementations of DNS. It shows that negative caching would improve DNS performance only marginally in an internet of correctly implemented name servers. It concludes by calling for a fundamental change in the way we specify and implement future name servers and distributed applications.

The Domain Name System (DNS) is an essential part of the Internet infrastructure and provides fundamental services, such as translating host names into IP addresses for Internet communication. The DNS is vulnerable to a number of potential faults and attacks. In particular, false routing announcements can deny access to the DNS service or redirect DNS queries to a malicious impostor Due to the hierarchical DNS design, a single fault or attack against the routes to any of the top level DNS servers can disrupt Internet services to millions of users. In this paper we propose a path-filtering approach to protect the routes to the critical top level DNS servers. Our approach exploits the high degree of redundancy in top level DNS servers and also exploits the observation that popular destinations, including top level DNS servers, are well connected via stable routes. Our path-filter restricts the potential top level DNS server route changes to be within a set of established paths. Heuristics derived from routing operations are used to adjust the potential routes overtime. We tested our path-filtering design against BGP routing logs and the results show that the design can effectively ensure correct routes to top level DNS servers without impacting DNS service availability.