Visible to the public Biblio

Filters: Author is Jain, S.  [Clear All Filters]
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z 
G
Gandhi, A., Jain, S..  2020.  Adversarial Perturbations Fool Deepfake Detectors. 2020 International Joint Conference on Neural Networks (IJCNN). :1—8.
This work uses adversarial perturbations to enhance deepfake images and fool common deepfake detectors. We created adversarial perturbations using the Fast Gradient Sign Method and the Carlini and Wagner L2 norm attack in both blackbox and whitebox settings. Detectors achieved over 95% accuracy on unperturbed deepfakes, but less than 27% accuracy on perturbed deepfakes. We also explore two improvements to deep-fake detectors: (i) Lipschitz regularization, and (ii) Deep Image Prior (DIP). Lipschitz regularization constrains the gradient of the detector with respect to the input in order to increase robustness to input perturbations. The DIP defense removes perturbations using generative convolutional neural networks in an unsupervised manner. Regularization improved the detection of perturbed deepfakes on average, including a 10% accuracy boost in the blackbox case. The DIP defense achieved 95% accuracy on perturbed deepfakes that fooled the original detector while retaining 98% accuracy in other cases on a 100 image subsample.
J
Jain, S., Sharma, S., Chandavarkar, B. R..  2020.  Mitigating Man-in-the-Middle Attack in Digital Signature. 2020 11th International Conference on Computing, Communication and Networking Technologies (ICCCNT). :1–5.
We all are living in the digital era, where the maximum of the information is available online. The digital world has made the transfer of information easy and provides the basic needs of security like authentication, integrity, nonrepudiation, etc. But, with the improvement in security, cyber-attacks have also increased. Security researchers have provided many techniques to prevent these cyber-attacks; one is a Digital Signature (DS). The digital signature uses cryptographic key pairs (public and private) to provide the message's integrity and verify the sender's identity. The private key used in the digital signature is confidential; if attackers find it by using various techniques, then this can result in an attack. This paper presents a brief introduction about the digital signature and how it is vulnerable to a man-in-the-middle attack. Further, it discusses a technique to prevent this attack in the digital signature.
K
Kaur, A., Jain, S., Goel, S..  2017.  A Support Vector Machine Based Approach for Code Smell Detection. 2017 International Conference on Machine Learning and Data Science (MLDS). :9–14.

Code smells may be introduced in software due to market rivalry, work pressure deadline, improper functioning, skills or inexperience of software developers. Code smells indicate problems in design or code which makes software hard to change and maintain. Detecting code smells could reduce the effort of developers, resources and cost of the software. Many researchers have proposed different techniques like DETEX for detecting code smells which have limited precision and recall. To overcome these limitations, a new technique named as SVMCSD has been proposed for the detection of code smells, based on support vector machine learning technique. Four code smells are specified namely God Class, Feature Envy, Data Class and Long Method and the proposed technique is validated on two open source systems namely ArgoUML and Xerces. The accuracy of SVMCSD is found to be better than DETEX in terms of two metrics, precision and recall, when applied on a subset of a system. While considering the entire system, SVMCSD detect more occurrences of code smells than DETEX.