INTERNATIONAL JOURNAL OF SCIENTIFIC DEVELOPMENT AND RESEARCH International Peer Reviewed & Refereed Journals, Open Access Journal ISSN Approved Journal No: 2455-2631 | Impact factor: 8.15 | ESTD Year: 2016
open access , Peer-reviewed, and Refereed Journals, Impact factor 8.15
ADVERSARIAL DEFENSE FOR MNIST: INVESTIGATING ADVERSARIAL TRAINING AND FGSM
Authors Name:
Kommineni Srivathsav
, Sai Manas Rao Pulakonti , Kadali Narayana Anudeep , Kommineni Srinivas , Kommineni Sri Lakshmi Poojitha
Unique Id:
IJSDR2303175
Published In:
Volume 8 Issue 3, March-2023
Abstract:
This research looks at strategies for defending machine learning models from adversarial assaults, which are deliberate attempts to misclassify input fed to machine learning models in order to trick them. Machine learning systems' dependability and security are seriously threatened by adversarial assaults. The research paper focuses on adversarial training, a popular defense mechanism that involves augmenting the training data with adversarial examples to make the model more robust to adversarial attacks. In the study, a convolutional neural network trained on the MNIST dataset is used as an example to demonstrate how adversarial training might increase the model's performance on adversarial examples. The research paper concludes that adversarial training is an effective defense mechanism but has limitations and should be used in combination with other defense mechanisms. The outcomes show how crucial it is to protect machine learning models against adversarial attacks in order to ensure their dependability and robustness. To create protection systems that are more reliable and effective, further study is required.
Keywords:
Cite Article:
"ADVERSARIAL DEFENSE FOR MNIST: INVESTIGATING ADVERSARIAL TRAINING AND FGSM", International Journal of Science & Engineering Development Research (www.ijsdr.org), ISSN:2455-2631, Vol.8, Issue 3, page no.1068 - 1070, March-2023, Available :http://www.ijsdr.org/papers/IJSDR2303175.pdf
Downloads:
000337070
Publication Details:
Published Paper ID: IJSDR2303175
Registration ID:204792
Published In: Volume 8 Issue 3, March-2023
DOI (Digital Object Identifier):
Page No: 1068 - 1070
Publisher: IJSDR | www.ijsdr.org
ISSN Number: 2455-2631
Facebook Twitter Instagram LinkedIn