MLMihai Lazarescu

Mihai Lazarescu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Mihai Lazarescu has a total of 64 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Green Multistage Upgrade of Switches and Controllers for Bundled-Links SDNs

    by Hiryanto Lely, Sieteng Soh, Kwan-Wu Chin, Duc-Son Pham, Mihai M. Lazarescu

    The switches in legacy networks perform both control and data forwarding operations. In contrast, in a software defined network (SDN), a controller is responsible for network control and each SDN-switch (s-switch) only forwards data packets. Advantageously, controllers can be upgraded easily to run state-of-the-art network configuration and management solutions. In this context, this paper addresses the novel problem of upgrading a legacy network into an SDN over multiple stages that span months or years. More specifically, it aims to minimize energy consumption by optimizing switch upgrades, controller placement, and traffic routing subject to an operator’s budget, traffic delay tolerance, and controller capacity. We formulate the problem as a mixed integer program (MIP), and develop a heuristic solution that ensures a single path is used between switches and up to two link-disjoint paths are used between an s-switch and its controller. Our simulation results show that increasing an operator’s budget and the number of upgrade stages reduce the energy consumption of tested networks by 68.42%. Further, our heuristic solution yields energy saving that is within 5% away from the optimal value. In addition, deploying controllers at strategic locations saves more energy than placing them at arbitrary locations. Lastly, our heuristic solution guarantees the delay requirement of traffic demands and runs up to 307 times faster than a prior solution.

  • Green Multi-Stage Upgrade for Bundled-Link SDNs With Budget and Delay Constraints

    by Lely Hiryanto, Sieteng Soh, Kwan-Wu Chin, Mihai Lazarescu

    Upgrading a legacy network to a Software Defined Network in stages, and minimizing the energy consumption of a network are now of great interest to operators. To this end, this paper addresses a novel problem: minimize the energy consumption of a network by upgrading switches over multiple stages subject to the available monetary budget at each stage. Our problem considers (i) bundled links that can be powered-off individually, (ii) decreasing upgrade cost and increasing traffic demands over multiple stages, and (iii) rerouting demands to an alternative path with a delay that is within a predefined limit. Additionally, the link load on the path is no larger than a given maximum utilization. We formulate the problem as an Integer Linear Program and propose a greedy heuristic called Green Multi-Stage Switch Upgrade (GMSU). Experiment results on five actual network topologies show that: 1) our approaches reduce the energy consumption by up to 73.98%; 2) rerouting traffic demand via longer paths improves the energy saving by up to 9.6%; 3) GMSU produces results that are at most 4.83% from the optimal result; and 4) increasing the budget and the number of stages affect the total energy saved and the number of upgraded switches.


  • An Efficient Feature Extraction Method for the Detection of Material Rings in Rotary Kilns

    by Duc-Son Pham, John Ridley, Mihai Lazarescu

    Rotary kilns are important industrial plants to process material. We address the problem of detecting material rings formed in alumina rotary kilns, which lead to high waste and substantial economic loss. This necessitates the use of artificial intelligence to help kiln operators detect material rings promptly. We describe a recently developed system that extracts useful features and detects the presence of material rings efficiently and accurately in real time. Our major contribution is a novel feature extraction method based on geometric transformation and peak localization exploiting the domain knowledge. We also contribute a large data set covering many thousand labelled frames for the evaluation of ring detection. On this data set, we demonstrate that our novel feature extraction method outperforms other alternatives in the literature. Our method is also demonstrated as favourable over deep learning approaches. Our system produces an overall accuracy of nearly 96%, which is acceptable for deployment.


  • Open Access

    USING DISTRIBUTION ANALYSIS FOR PARAMETER SELECTION IN REPSTREAM

    by Ross Callister, Perth Kent Street, Duc-Son Pham, Mihai Lazarescu

    One of the most significant challenges in data clustering is the evolution of the data distributions over time. Many clustering algorithms have been introduced to deal specifically with streaming data, but common amongst them is that they require users to set input parameters. These inform the algorithm about the criteria under which data points may be clustered together. Setting the initial parameters for a clustering algorithm is itself a non-trivial task, but the evolution of the data distribution over time could mean even optimally set parameters could become non-optimal as the stream evolves. In this paper we extend the RepStream algorithm, a combination graph and density-based clustering algorithm, in a way which allows the primary input parameter, the K value, to be automatically adjusted over time. We introduce a feature called the edge distribution score which we compute for data in memory, as well as introducing an incremental method for adjusting the K parameter over time based on this score. We evaluate our methods against RepStream itself, and other contemporary stream clustering algorithms, and show how our method of automatically adjusting the K value over time leads to higher quality clustering output even when the initial parameters are set poorly.

  • Link Scheduling in Rechargeable Wireless Sensor Networks with Harvesting Time and Battery Capacity Constraints

    by Tony, Sieteng Soh, Mihai Lazarescu, Kwan-Wu Chin

    A link scheduler ensures the transmissions in rechargeable Wireless Sensor Networks (rWSNs) are collision-free. Hence, it plays a critical role in ensuring high network capacity and the energy used for transmission/reception is not wasted due to collisions. This paper proposes a scheduler that generates a Time Division Multiple Access link schedule for use in a rWSN. Different from most prior works, our scheduler considers the time required by each node to harvest sufficient energy to transmit/receive a packet. Further, it utilizes the more efficient Harvest-Use-Store (HUS) model and considers sensor nodes with finite battery capacity. We present a greedy heuristic that activates links according to the earliest time in which their end nodes have sufficient energy to transmit/receive a packet. Our simulation results show that the time to recharge a sensor node significantly increases the link schedule or superframe lengths; i.e., by up to 563.9% as compared to the case where sensor nodes have no energy constraint. Further, in comparison to the Harvest-Store-Use (HSU) model, using HUS can reduce superframe lengths by up to 45.3%. Our experiments also show that increasing battery capacity does not effect the superframe length significantly; i.e., it reduces the length only by up to 2.5%. Finally, our proposed heuristic can generate superframe lengths that are on average 23% longer as compared to the lower bound on the superframe length when nodes have energy constraint.


  • An Investigation of Power Law Probability Distributions for Network Anomaly Detection

    by Stefan Prandl, Mihai Lazarescu, Duc Son Pham, Sie Teng Soh, Subhash Kak

    It has been previously determined that SYN packet inter arrival times are conformant with Benford's law, which predicts the frequency of the leading digits in naturally occurring collections of numbers, and suggested that conformity or non-conformity to Benford's law could be used to detect network anomalies. This paper expands upon that suggestion by making three contributions. First, we verify that conformity to Benford's law of inter arrival times is also true for certain types of both TCP and UDP packets. Second, we discover that packet length could also be another alternative to inter arrival times, with the advantage that it follows both Benford's and Zipf's laws, implying its reliability in detecting network traffic anomaly. Finally, we explore the potential application of power laws in the specific detection of denial-of-service (DoS) attacks using both inter arrival times and packet length. Extensive experiments on the MAWI benchmark dataset and two additional datasets support our claims and demonstrate that whilst Benfordian analysis of inter arrival times can identify DoS attacks, the combination of Benfordian and Zipfian analysis of packet length gives more reliable detection.


  • Open Access

    Graph-based clustering with DRepStream

    by Ross Callister, Mihai Lazarescu, Duc-Son Pham

    Finding and setting input parameters for clustering algorithms is a challenging thing due to the unsupervised nature of clustering. The accuracy of clustering algorithms can be affected greatly by setting parameters appropriately for the dataset, however without ground truth labels and external validation it can be impossible to know when the parameters are set well. In this paper we propose the DRepStream algorithm, which extends the RepStream algorithm. DRepStream uses a graph-based approach, and unlike its predecessor does not require the primary K parameter used in K-nearest neighbour graphs. Our algorithm automatically computes the number of outgoing edges for each vertex in the graph using a computed metric known as the anomalous edge score. We evaluate the performance of our algorithm on other previous stream clustering algorithms on real world benchmark datasets.


  • A Study of Web Application Firewall Solutions

    by Stefan Prandl, Mihai Lazarescu, Duc-Son Pham