Biblio

Storage area networking is driving commodity data center switches to support lossless Ethernet (DCB). Unfortunately, to enable DCB for all traffic on arbitrary network topologies, we must address several problems that can arise in lossless networks, e.g., large buffering delays, unfairness, head of line blocking, and deadlock. We propose TCP-Bolt, a TCP variant that not only addresses the first three problems but reduces flow completion times by as much as 70%. We also introduce a simple, practical deadlock-free routing scheme that eliminates deadlock while achieving aggregate network throughput within 15% of ECMP routing. This small compromise in potential routing capacity is well worth the gains in flow completion time. We note that our results on deadlock-free routing are also of independent interest to the storage area networking community. Further, as our hardware testbed illustrates, these gains are achievable today, without hardware changes to switches or NICs.

Tor is a popular low-latency anonymous communication system. However, it is currently abused in various ways. Tor exit routers are frequently troubled by administrative and legal complaints. To gain an insight into such abuse, we design and implement a novel system, TorWard, for the discovery and systematic study of malicious traffic over Tor. The system can avoid legal and administrative complaints and allows the investigation to be performed in a sensitive environment such as a university campus. An IDS (Intrusion Detection System) is used to discover and classify malicious traffic. We performed comprehensive analysis and extensive real-world experiments to validate the feasibility and effectiveness of TorWard. Our data shows that around 10% Tor traffic can trigger IDS alerts. Malicious traffic includes P2P traffic, malware traffic (e.g., botnet traffic), DoS (Denial-of-Service) attack traffic, spam, and others. Around 200 known malware have been identified. To the best of our knowledge, we are the first to perform malicious traffic categorization over Tor.
 

The Domain Name System (DNS) is widely seen as a vital protocol of the modern Internet. For example, popular services like load balancers and Content Delivery Networks heavily rely on DNS. Because of its important role, DNS is also a desirable target for malicious activities such as spamming, phishing, and botnets. To protect networks against these attacks, a number of DNS-based security approaches have been proposed. The key insight of our study is to measure the effectiveness of security approaches that rely on DNS in large-scale networks. For this purpose, we answer the following questions, How often is DNS used? Are most of the Internet flows established after contacting DNS? In this study, we collected data from the University of Auckland campus network with more than 33,000 Internet users and processed it to find out how DNS is being used. Moreover, we studied the flows that were established with and without contacting DNS. Our results show that less than 5 percent of the observed flows use DNS. Therefore, we argue that those security approaches that solely depend on DNS are not sufficient to protect large-scale networks.

Botnets are the most common vehicle of cyber-criminal activity. They are used for spamming, phishing, denial-of-service attacks, brute-force cracking, stealing private information, and cyber warfare. Botnets carry out network scans for several reasons, including searching for vulnerable machines to infect and recruit into the botnet, probing networks for enumeration or penetration, etc. We present the measurement and analysis of a horizontal scan of the entire IPv4 address space conducted by the Sality botnet in February 2011. This 12-day scan originated from approximately 3 million distinct IP addresses and used a heavily coordinated and unusually covert scanning strategy to try to discover and compromise VoIP-related (SIP server) infrastructure. We observed this event through the UCSD Network Telescope, a /8 darknet continuously receiving large amounts of unsolicited traffic, and we correlate this traffic data with other public sources of data to validate our inferences. Sality is one of the largest botnets ever identified by researchers. Its behavior represents ominous advances in the evolution of modern malware: the use of more sophisticated stealth scanning strategies by millions of coordinated bots, targeting critical voice communications infrastructure. This paper offers a detailed dissection of the botnet's scanning behavior, including general methods to correlate, visualize, and extrapolate botnet behavior across the global Internet.
 

Routers in the Content-Centric Networking (CCN) architecture maintain state for all pending content requests, so as to be able to later return the corresponding content. By employing stateful forwarding, CCN supports native multicast, enhances security and enables adaptive forwarding, at the cost of excessive forwarding state that raises scalability concerns. We propose a semi-stateless forwarding scheme in which, instead of tracking each request at every on-path router, requests are tracked at every d hops. At intermediate hops, requests gather reverse path information, which is later used to deliver responses between routers using Bloom filter-based stateless forwarding. Our approach effectively reduces forwarding state, while preserving the advantages of CCN forwarding. Evaluation results over realistic ISP topologies show that our approach reduces forwarding state by 54%-70% in unicast delivery, without any bandwidth penalties, while in multicast delivery it reduces forwarding state by 34%-55% at the expense of 6%-13% in bandwidth overhead.
 

In this paper, an algorithm is proposed to automatically produce hierarchical graph-based representations of maritime shipping lanes extrapolated from historical vessel positioning data. Each shipping lane is generated based on the detection of the vessel behavioural changes and represented in a compact synthetic route composed of the network nodes and route segments. The outcome of the knowledge discovery process is a geographical maritime network that can be used in Maritime Situational Awareness (MSA) applications such as track reconstruction from missing information, situation/destination prediction, and detection of anomalous behaviour. Experimental results are presented, testing the algorithm in a specific scenario of interest, the Dover Strait.
 

Software-Defined Networking (SDN) allows network capabilities and services to be managed through a central control point. Moving Target Defense (MTD) on the other hand, introduces a constantly adapting environment in order to delay or prevent attacks on a system. MTD is a use case where SDN can be leveraged in order to provide attack surface obfuscation. In this paper, we investigate how SDN can be used in some network-based MTD techniques. We first describe the advantages and disadvantages of these techniques, the potential countermeasures attackers could take to circumvent them, and the overhead of implementing MTD using SDN. Subsequently, we study the performance of the SDN-based MTD methods using Cisco's One Platform Kit and we show that they significantly increase the attacker's overheads.

Port hopping is a typical moving target defense, which constantly changes service port number to thwart reconnaissance attack. It is effective in hiding service identities and confusing potential attackers, but it is still unknown how effective port hopping is and under what circumstances it is a viable proactive defense because the existed works are limited and they usually discuss only a few parameters and give some empirical studies. This paper introduces urn model and quantifies the likelihood of attacker success in terms of the port pool size, number of probes, number of vulnerable services, and hopping frequency. Theoretical analysis shows that port hopping is an effective and promising proactive defense technology in thwarting network attacks.
 

Network virtualization sits firmly on the Internet evolutionary path allowing researchers to experiment with novel clean-slate designs over the production network and practitioners to manage multi-tenants infrastructures in a flexible and scalable manner. In such scenarios, isolation between virtual networks is often intended as purely logical: this is the case of address space isolation or flow space isolation. This approach neglects the effect that network virtualization has on resource allocation network-wide. In this work we investigate the price paid by a purely logical approach in terms of performance degradation. This performance loss is paid by the actual users of a multi-tenants datacenter network. We propose a solution to this problem leveraging on a new network virtualization primitive, namely an online link utilization feedback mechanism. It provides each tenant with the necessary information to make efficient use of network resources. We evaluate our solution trough a real implementation exploiting the OpenFlow protocol. Empirical results confirm that the proposed scheme is able to support tenants in exploiting virtualized network resources effectively.
 

It is expected that clean-slate network designs will be implemented for wide-area network applications. Multi-tenancy in OpenFlow networks is an effective method to supporting a clean-slate network design, because the cost-effectiveness is improved by the sharing of substrate networks. To guarantee the programmability of OpenFlow for tenants, a complete flow space (i.e., header values of the data packets) virtualization is necessary. Wide-area substrate networks typically have multiple administrators. We therefore need to implement a flow space virtualization over multiple administration networks. In existing techniques, a third party is solely responsible for managing the mapping of header values for flow space virtualization for substrate network administrators and tenants, despite the severity of a third party failure. In this paper, we propose an AutoVFlow mechanism that allows flow space virtualization in a wide-area networks without the need for a third party. Substrate network administrators implement a flow space virtualization autonomously. They are responsible for virtualizing a flow space involving switches in their own substrate networks. Using a prototype of AutoVFlow, we measured the virtualization overhead, the results of which show a negligible amount of overhead.
 

This paper presents a unified approach for the detection of network anomalies. Current state of the art methods are often able to detect one class of anomalies at the cost of others. Our approach is based on using a Linear Dynamical System (LDS) to model network traffic. An LDS is equivalent to Hidden Markov Model (HMM) for continuous-valued data and can be computed using incremental methods to manage high-throughput (volume) and velocity that characterizes Big Data. Detailed experiments on synthetic and real network traces shows a significant improvement in detection capability over competing approaches. In the process we also address the issue of robustness of network anomaly detection systems in a principled fashion.
 

This paper presents FlowNAC, a Flow-based Network Access Control solution that allows to grant users the rights to access the network depending on the target service requested. Each service, defined univocally as a set of flows, can be independently requested and multiple services can be authorized simultaneously. Building this proposal over SDN principles has several benefits: SDN adds the appropriate granularity (fine-or coarse-grained) depending on the target scenario and flexibility to dynamically identify the services at data plane as a set of flows to enforce the adequate policy. FlowNAC uses a modified version of IEEE 802.1X (novel EAPoL-in-EAPoL encapsulation) to authenticate the users (without the need of a captive portal) and service level access control based on proactive deployment of flows (instead of reactive). Explicit service request avoids misidentifying the target service, as it could happen by analyzing the traffic (e.g. private services). The proposal is evaluated in a challenging scenario (concurrent authentication and authorization processes) with promising results.
 

Cyber scanning refers to the task of probing enterprise networks or Internet wide services, searching for vulnerabilities or ways to infiltrate IT assets. This misdemeanor is often the primarily methodology that is adopted by attackers prior to launching a targeted cyber attack. Hence, it is of paramount importance to research and adopt methods for the detection and attribution of cyber scanning. Nevertheless, with the surge of complex offered services from one side and the proliferation of hackers' refined, advanced, and sophisticated techniques from the other side, the task of containing cyber scanning poses serious issues and challenges. Furthermore recently, there has been a flourishing of a cyber phenomenon dubbed as cyber scanning campaigns - scanning techniques that are highly distributed, possess composite stealth capabilities and high coordination - rendering almost all current detection techniques unfeasible. This paper presents a comprehensive survey of the entire cyber scanning topic. It categorizes cyber scanning by elaborating on its nature, strategies and approaches. It also provides the reader with a classification and an exhaustive review of its techniques. Moreover, it offers a taxonomy of the current literature by focusing on distributed cyber scanning detection methods. To tackle cyber scanning campaigns, this paper uniquely reports on the analysis of two recent cyber scanning incidents. Finally, several concluding remarks are discussed.
 

Detection of high risk network flows and high risk hosts is becoming ever more important and more challenging. In order to selectively apply deep packet inspection (DPI) one has to isolate in real time high risk network activities within a huge number of monitored network flows. To help address this problem, we propose an iterative methodology for a simultaneous assessment of risk scores for both hosts and network flows. The proposed approach measures the risk scores of hosts and flows in an interdependent manner; thus, the risk score of a flow influences the risk score of its source and destination hosts, and also the risk score of a host is evaluated by taking into account the risk scores of flows initiated by or terminated at the host. Our experimental results show that such an approach not only effective in detecting high risk hosts and flows but, when deployed in high throughput networks, is also more efficient than PageRank based algorithms.

Denial-of-Service (DoS) and probe attacks are growing more modern and sophisticated in order to evade detection by Intrusion Detection Systems (IDSs) and to increase the potent threat to the availability of network services. Detecting these attacks is quite tough for network operators using misuse-based IDSs because they need to see through attackers and upgrade their IDSs by adding new accurate attack signatures. In this paper, we proposed a novel signal and image processing-based method for detecting network probe and DoS attacks in which prior knowledge of attacks is not required. The method uses a time-frequency representation technique called S-transform, which is an extension of Wavelet Transform, to reveal abnormal frequency components caused by attacks in a traffic signal (e.g., a time-series of the number of packets). Firstly, S-Transform converts the traffic signal to a two-dimensional image which describes time-frequency behavior of the traffic signal. The frequencies that behave abnormally are discovered as abnormal regions in the image. Secondly, Otsu's method is used to detect the abnormal regions and identify time that attacks occur. We evaluated the effectiveness of the proposed method with several network probe and DoS attacks such as port scans, packet flooding attacks, and a low-intensity DoS attack. The results clearly indicated that the method is effective for detecting the probe and DoS attack streams which were generated to real-world Internet.

Intrusion Detection Systems (IDS) have become a necessity in computer security systems because of the increase in unauthorized accesses and attacks. Intrusion Detection is a major component in computer security systems that can be classified as Host-based Intrusion Detection System (HIDS), which protects a certain host or system and Network-based Intrusion detection system (NIDS), which protects a network of hosts and systems. This paper addresses Probes attacks or reconnaissance attacks, which try to collect any possible relevant information in the network. Network probe attacks have two types: Host Sweep and Port Scan attacks. Host Sweep attacks determine the hosts that exist in the network, while port scan attacks determine the available services that exist in the network. This paper uses an intelligent system to maximize the recognition rate of network attacks by embedding the temporal behavior of the attacks into a TDNN neural network structure. The proposed system consists of five modules: packet capture engine, preprocessor, pattern recognition, classification, and monitoring and alert module. We have tested the system in a real environment where it has shown good capability in detecting attacks. In addition, the system has been tested using DARPA 1998 dataset with 100% recognition rate. In fact, our system can recognize attacks in a constant time.