Intelligent DDoS packet filtering in high-speed networks

Currently high-speed networks have been attacked by successive waves of Distributed Denial of Service (DDoS) attacks. There are two major challenges on DDoS defense in the high-speed networks. One is to sensitively and accurately detect attack traffic, and the other is to filter out the attack traffic quickly, which mainly depends on high-speed packet classification. Unfortunately most current defense approaches can not efficiently detect and quickly filter out attack traffic. Our approach is to find the network anomalies by using neural network, deploy the system at distributed routers, identify the attack packets, and then filter them quickly by a Bloom filter-based classifier. The evaluation results show that this approach can be used to defend against both intensive and subtle DDoS attacks, and can catch DDoS attacks’ characteristic of starting from multiple sources to a single victim. The simple complexity, high classification speed and low storage requirements make it especially suitable for DDoS defense in high-speed networks.

Classifying DDoS packets in high-speed networks

Recently high-speed networks have been utilized by attackers as Distributed Denial of Service (DDoS) attack infrastructure. Services on high-speed networks also have been attacked by successive waves of the DDoS attacks. How to sensitively and accurately detect the attack traffic, and quickly filter out the attack packets are still the major challenges in DDoS defense. Unfortunately most current defense approaches can not efficiently fulfill these tasks. Our approach is to find the network anomalies by using neural network and classify DDoS packets by a Bloom filter-based classifier (BFC). BFC is a set of spaceefficient data structures and algorithms for packet classification. The evaluation results show that the simple complexity, high classification speed and accuracy and low storage requirements of this classifier make it not only suitable for DDoS filtering in high-speed networks, but also suitable for other applications such as string matching for intrusion detection systems and IP lookup for programmable routers.

High-precision five-axis machine for high-speed material processing using linear motors and parallel-serial kinematics

The paper describes some details of the mechanical and kinematics design of a five-axis mechanism. The design has been utilized to physically realize an industrial-scale five-axis milling machine that can carry a three KW spindle. However, the mechanism could be utilized in other material processing and factory automation applications. The mechanism has five rectilinear joints/axes. Two of these axes are arranged traditionally, i.e. in series, and the other three axes utilize the concept of parallel kinematics. This combination results in a design that allows three translational and two rotational two-mode degrees of freedom (DOFs). The design provides speed, accuracy and cost advantages over traditional five-axis machines. All axes are actuated using linear motors.

Dynamic simulation of collisions of heavy high-speed trucks with concrete barriers

Real vehicle collision experiments on full-scale road safety barriers are important to determine the outcome of a vehicle versus barrier impact accident. However, such experiments require large investment of time and money. Numerical simulation has therefore been imperative as an alternative method for testing concrete barriers. In this research, spring subgrade models were first developed to simulate the ground boundary of concrete barriers. Both heavy trucks and concrete barriers were modeled using finite element methods (FEM) to simulate dynamic collision performances. Comparison of the results generated from computer simulations and on-site full-scale experiments demonstrated that the developed models could be applied to simulate the collision of heavy trucks with concrete barriers to provide the data to design new road safety barriers and analyze existing ones.

Modelling factors of square tubes in high speed bending situations

Accurate finite element crash simulations of side impact depend upon a thorough understanding of dynamic tube bending. There is a need to understand the dynamic bending mode of square sections (equivalent of automotive structural parts) to obtain a greater confidence in CAE. This work varied strain rate and material definitions, such as Cowper-Symonds vs Zerilli-Armstrong, as well as initial velocity and yield strength. The results show that most of the plastic work is done between strains rates of 30 ¿ 300/s and strains up to 0.3. Peak strain rates were marginally above 1000/s with maximum strain greater than 1. When the strain rate definition and material model were modified, it was shown that a higher yield stress produced a higher reaction force. These results would suggest that the strain rate sensitivity needs to be carefully identified for accurate crash simulations.

Energy efficient, fully-connected mesh networks for high speed applications

Fully-connected mesh networks that can potentially be employed in a range of applications, are inherently associated with major deficiencies in interference management and network capacity improvement. The tree-connected (routing based) mesh networks used in today’s applications have major deficiencies in routing delays and reconfiguration delays in the implementation stage. This paper introduces a CDMA based fully-connected mesh network, which controls the transmission powers of the nodes in order to ensure that the communication channels remain interference-free and minimizes the energy consumption. Moreover, the bounds for the number of nodes and the spatial configuration are provided to ensures that the communication link satisfies the QoS (Quality of Service) requirements at all times.

High-speed pollen release in the white mulberry tree, Morus alba L

Anemophilous plants described as catapulting pollen explosively into the air have rarely attracted detailed examination. We investigated floral anthesis in a male mulberry tree with high-speed video and a force probe. The stamen was inflexed within the floral bud. Exposure to dry air initially resulted in a gradual movement of the stamen. This caused fine threads to tear at the stomium, ensuring dehiscence of the anther, and subsequently enabled the anther to slip off a restraining pistillode. The sudden release of stored elastic energy in the spring-like filament drove the stamen to straighten in less than 25 μs, and reflex the petals to velocities in excess of half the speed of sound. This is the fastest motion yet observed in biology, and approaches the theoretical physical limits for movements in plants.

Na+-doped zinc oxide nanofiber membrane for high speed humidity sensor

A novel sensitive humidity nanosensor based on Na1-doped ZnO nanofiber membrane has been prepared via electrospinning and calcination. The product was characterized by scanning electron microscopy and X-ray diffraction. During the whole relative humidity (11%–95%) measurement, the response and recovery time is about 3 and 6 s, respectively, with good linearity, and reproducibility. These remarkable and sensitive sensing performances make our product a good candidate in fabricating humidity sensors.