IJMTES – Paper Presented in: ‘2 day State Level workshop on Cyber Fest 17’, conducted by: ‘Department of Computer Engineering, Marathwada Mitra Mandal College of Engineering, Pune’ on 22-23 Feb 2017

Journal Title : International Journal of Modern Trends in Engineering and Science


Author’s Name : Ashvinee Kharat | Prof Dr B D Phulpagarunnamed

Paper Presented in : ‘2 day State Level workshop on Cyber Fest 17’, conducted by: ‘Department of Computer Engineering, Marathwada Mitra Mandal College of Engineering, Pune’ on 22-23 Feb 2017

Volume 04 Issue 05 2017

ISSN no:  2348-3121

Page no: 72-75

Abstract – Pattern attention and computing device learning approaches have been increasingly adopted in adversarial settings such as spam, intrusion, and malware detection, although their security against good-crafted assaults that aim to dodge detection with the aid of manipulating information at test time has now not yet been entirely assessed. Whilst previous work has been probably excited by devising adversary-aware classification algorithms to counter evasion makes an attempt, best few authors have considered the have an effect on of utilizing lowered feature units on classifier safety in opposition to the same assaults. A fascinating, preliminary influence is that classifier safety to evasion may be even worsened via the application of function decision. In this paper, we provide a more specific investigation of this facet, shedding some mild on the security residences of feature choice against evasion attacks. Stimulated via prior work on adversary-aware classifiers, we propose a novel adversary-aware characteristic resolution mannequin that can reinforce classifier safety in opposition to evasion assaults, with the aid of incorporating distinctive assumptions on the adversaries information manipulation approach. They focal point on an efficient, wrapper depends implementation of our technique, and experimentally validate its soundness on one of a kind software examples, together with unsolicited mail and malware detection. In this project AdaBoost classifier will be used. The basic concept behind AdaBoost is to engender a vigorous classifier by the conjuncture of many impuissant classifiers (hit rate barely better than 50

Key words— Adversarial Learning, Classifier Security, Evasion Attacks, Feature Selection, Malware Detection, Spam Filtering


  1. G. V. Cormack, TREC 2007 spam track overview in Proc. 16th Text Retrieval Conf. (TREC), 2007, pp. 123131.
  2. B. Biggio, Z. Akhtar, G. Fumera, G. L. Marcialis, and F. Roli, Security evaluation of biometric authentication systems under real spoofing attacks IET Biometrics, vol. 1, no. 1, pp. 1124, 2012.
  3. R. N. Rodrigues, L. L. Ling, and V. Govindaraju, Robustness of multimodal biometric fusion methods against spoof attacks J. Vis. Lang. Comput., vol. 20, no. 3, pp. 169179, 2009
  4. J. Newsome, B. Karp, and D. Song Paragraph: Thwarting signature learning by training maliciously in Recent Advances in Intrusion Detection (Lecture Notes in Computer Science). Berlin, Germany: Springer, 2006, pp. 81105.
  5. D. Moore, C. Shannon, D. J. Brown, G. M. Voelker, and S. Savage, Inferring Internet denial-of-service activity ACM Trans. Comput. Syst., vol. in Proc. 2nd Conf. Email Anti-Spam, Palo Alto, CA, USA, 2005
  6. J. Neumann, C. Schnrr, and G. Steidl, Combined SVM-based feature selection and classification Mach. Learn., vol. 61, nos. 1 3, pp. 129 150, 2005.
  7. J. Weston, A. Elisseeff, B. Schlkopf, and M. Tipping, Use of the zero norm with linear models and kernel methods J. Mach. Learn. Res., vol. 3, pp. 1439 1461, Mar. 2003.
  8. H. A. Le Thi, X. T. Vo, and T. P. Dinh, \ Robust feature selection for SVMs under uncertain data in Advances in Data Mining, Apps, and Theoretical Aspects. Berlin, Germany: Springer, 2013, pp. 151 165.
  9. B. Biggio, G. Fumera, and F. Roli, Multiple classifier systems under attack in Multiple Classifier Systems (LNCS 5997), N. E. Gayar, J. Kittler, and F. Roli, Eds. Berlin, Germany: Springer, 2010, pp. 74 83.
  10. M. Kolar and H. Liu, Feature selection in high-dimensional classification in Proc. 30th Int. Conf. Mach. Learn. (ICML Track), vol. 28. Atlanta, GA, USA, 2013, pp. 329 337.
Scroll Up