Analysis on Security Evaluation of Pattern Classifiers under Attack

Abstract

Analysis on security evaluation of pattern classifiers under attack describes pattern classification systems that are security evaluation problems because of different attacks. Pattern Classification commonly used in adversarial applications, like biometric authentication, network intrusion detection, and spam filtering. In these applications data can be purposely manipulated by humans to undermine their operation. This adversarial scenario’s exploitation may sometimes affect their performance, systems may exhibit vulnerabilities and limit their practical utility. This adversarial scenario is not taken into account by classical design methods. These Applications have an intrinsic adversarial nature since the input data can be purposely manipulated by an intelligent and adaptive adversary to undermine classifier operation. This often gives rise to an arms race between the adversary and the classifier designer. The system evaluates at design phase the security of pattern classifiers, namely, the performance degradation under potential attacks they may incur during operation. A generalize framework is used for evaluation of classifier security that formalizes and generalizes the training and testing datasets, to discriminate between a “legitimate” and a “malicious” pattern class Training and Testing sets have been obtained from distribution using a classical reassembling technique like bootstrapping or cross validation. Security evaluation can be carried out by averaging the performance of the trained and tested data.

Authors and Affiliations

Rupali B. Navalkar, Prof. Rajeshri R. Shelke

Keywords

Related Articles

Line Following Pick and Place Robot for Green Planet

This project deals with the development and construction of a line follower robot capable to sense multiple colour combinations and placing buds in the given zone by picking it. These two robots to be prepared are as f...

A Guarantee and Service Provider Based GeTrust Model in Chord-Based P2p Networks

more and more users are attracted by p2p networks characterized by decentralization, autonomy and anonymity. However, users’ unconstrained behavior makes it necessary to use a trust model when establishing trust relatio...

Critical Study of the Disturbance in Meteorological Parameters Due to Global Warming

To study the relationship between global warming and change in meteorological parameters at regional level. Nagpur is central most cities in Indian domain and truly represents the effect of all directional weather chang...

Study of Characteristics of Nano Particles Reinforced With Metal Matrix Composites

Metal Matrix composites is well known for its wear resistance, fatigue, fracture toughness, stiffness .In this work SiC and Nano particles are reinforced in the aluminium metal matrix to improve its properties. This is...

A Survey of Literature on Impact of Silica Fume (SF) and Saw Dust Ash (SDA) On Expansive Soil

The quest to modify the engineering properties of weak or expansive soils seem to be unending as these expansive soils posse a huge challenge to construction of roads by Highway engineers due to their poor geotechnical...

Download PDF file
  • EP ID EP23132
  • DOI http://doi.org/10.22214/ijraset.2017.2043
  • Views 223
  • Downloads 6

How To Cite

Rupali B. Navalkar, Prof. Rajeshri R. Shelke (2017). Analysis on Security Evaluation of Pattern Classifiers under Attack. International Journal for Research in Applied Science and Engineering Technology (IJRASET), 5(2), -. https://europub.co.uk/articles/-A-23132