Krzysztof Rusek

Risk-awareness in resilient networks design: Value-at-risk is enough

The paper discusses business impact analysis in the context of resilient networks. The emphasis is put on risk evaluation using Value-at-Risk (VaR) and Conditional VaR measures. The concerns known for VaR are shown, especially the theoretical lack of subadditivity. The study shows that in practice the disadvantages do not appear in resilient networks design, and there is no need to apply the more complex CVaR.

Transient and stationary characteristics of a packet buffer modelled as an MAP/SM/1/b system

Abstract A packet buffer limited to a fixed number of packets (regardless of their lengths) is considered. The buffer is described as a finite FIFO queuing system fed by a Markovian Arrival Process (MAP) with service times forming a Semi-Markov (SM) process (MAP /SM /1/b in Kendall’s notation). Such assumptions allow us to obtain new analytical results for the queuing characteristics of the buffer. In the paper, the following are considered: the time to fill the buffer, the local loss intensity, the loss ratio, and the total number of losses in a given time interval. Predictions of the proposed model are much closer to the trace-driven simulation results compared with the prediction of the MAP /G/1/b model.

Risk mitigation in resilient networks

This article proposes shifting the perspective for the design of resilient networks from cost-focused to one suited for business purposes. Risk engineering is used as a basis to enable us to monetarily express not only the cost of recovery, but also the impact of failures affecting connections (expressed with use of penalties imposed on an operator), and then to find the tradeoff between the cost of the assigned recovery methods and the improved level of resilience. During risk assessment, monetary quantification of penalties is applied with compensation policies, and business relevant risk measures are used. Then, risk response selection is based on various risk mitigation strategies (involving profit maximization, total benefit coverage, cost balance, and risk minimization) proposed in the security risk management. Looking for the cost-risk trade-off related to the assumed mitigation strategy is a complex optimization problem that cannot be modeled with deterministic linear programming. Therefore, to be able to choose recovery options, we develop a genetic algorithm. The results show diversity of recovery procedures selected for various selected mitigation strategies.

Correct router interface modeling

The aim of this paper is to determine how to model router interface in order to accurately predict packet drops. There is an enormous amount of research on traffic models reported, however, a model of router interface has not gained proper consideration yet. Our experiments reveal that an incorrect model of the router interface can result in a significant disparity between drop probabilities measured on a physical interface and derived from a trace-driven simulation study. In this paper an accurate model of the router interface (Cisco IOS-based routers with a non-distributed architecture) was presented.

Two-stage neural network regression of eye location in face images

Automatic eye localization is a crucial part of many computer vision algorithms for processing face images. Some of the existing algorithms can be very accurate, albeit at the cost of computational complexity. In this paper, a new solution to the problem of automatic eye localization is proposed. Eye localization is posed as a nonlinear regression problem solved by two feed-forward multilayer perceptrons (MLP) working in a cascade. The input feature vector of the first network is constructed from coefficients of a two dimensional discrete cosine transform(DCT) of a face image. The second network generates corrections based on small image patches. Feature extraction and neural network prediction have known and efficient implementations, thus the entire procedure can be very fast. The paper hints at the neural network structure and the procedure for generating artificial training samples from a low number of face images. In terms of accuracy, the method is comparable to state-of-the-art techniques; however it is based on numerical procedures that could be highly optimized (fast Fourier transform and matrix multiplication).

LDA for Face Profile Detection

This paper presents a new approach to face profile classification. It was shown that the Linear Discriminant Analysis combined with the Principal Component Analysis can be used to construct an accurate face profile predictor. Proposed classifier achieves 85% accuracy in face profile prediction.

Effective Risk Assessment in Resilient Communication Networks

The paper discusses business impact analysis in the context of resilient communication networks. It is based on the total (aggregated) penalty that may be paid by an operator when the services (identified with transport demands) provided are interrupted due to network failures. The level of penalty is expressed as a commonly accepted business risk measure, Value-at-Risk (

Neural Network Regression of Eyes Location in Face Images

VaR