Adversarial Machine Learning in Image Classification: A Survey Toward the Defender’s Perspective

Deep Learning algorithms have achieved state-of-the-art performance for Image Classification. For this reason, they have been used even in security-critical applications, such as biometric recognition systems and self-driving cars. However, recent works have shown those algorithms, which can even surpass human capabilities, are vulnerable to adversarial examples. In Computer Vision, adversarial examples are images containing subtle perturbations generated by malicious optimization algorithms to fool classifiers. As an attempt to mitigate these vulnerabilities, numerous countermeasures have been proposed recently in the literature. However, devising an efficient defense mechanism has proven to be a difficult task, since many approaches demonstrated to be ineffective against adaptive attackers. Thus, this article aims to provide all readerships with a review of the latest research progress on Adversarial Machine Learning in Image Classification, nevertheless, with a defender’s perspective. This article introduces novel taxonomies for categorizing adversarial attacks and defenses, as well as discuss possible reasons regarding the existence of adversarial examples. In addition, relevant guidance is also provided to assist researchers when devising and evaluating defenses. Finally, based on the reviewed literature, this article suggests some promising paths for future research.

Biometric Recognition of Finger Knuckle Print Based on the Fusion of Global Features and Local Features

Compared with the most traditional fingerprint identification, knuckle print and hand shape are more stable, not easy to abrase, forge, and pilfer; in aspect of image acquisition, the requirement of acquisition equipment and environment are not high; and the noncontact acquisition method also greatly improves the users’ satisfaction; therefore, finger knuckle print and hand shape of single-mode identification system have attracted extensive attention both at home and abroad. A large number of studies show that multibiometric fusion can greatly improve the recognition rate, antiattack, and robustness of the biometric recognition system. A method combining global features and local features was designed for the recognition of finger knuckle print images. On the one hand, principal component analysis (PCA) was used as the global feature for rapid recognition. On the other hand, the local binary pattern (LBP) operator was taken as the local feature in order to extract the texture features that can reflect details. A two-layer serial fusion strategy is proposed in the combination of global and local features. Firstly, the sample library scope was narrowed according to the global matching result. Secondly, the matching result was further determined by fine matching. By combining the fast speed of global coarse matching and the high accuracy of local refined matching, the designed method can improve the recognition rate and the recognition speed.

Best Regression for Eye Recognition

Human eye offers a number of opportunities for biometric recognition. The essential parts of the eye like cornea, iris, veins and retina can determine different characteristics. Systems using eyes’ features are widely deployed for identification in government requirement levels and laws; but also beginning to have more space in portable validation world. The first image was prepared to be used and monitored using CLAHE which means (Contrast Limited Adaptive Histogram Equalization) to improve the contrast of the image, after that the 3D surface plot was created for this image then different types of regression were used and the better one was chosen. The results showed that power regression is better, and fitter than other fitting methods (8th, 7th, 6th, 5th, 4th, 3rd, 2nd) degree polynomial, and straight line respectively, when depending on the sum of residual squared. The estimations of R-square demonstrated that (5th, 6th, 7th, 8th) have a great proportion of variance in the model followed by (power, 4th, 3rd, 2nd, straight line) respectively. The conclusion from these results is that the power regression has a better fitting than other types of fitting functions for this study and similar ones.

Political Effects of Digital Transformation of Urban Governance (On the Example of Moscow)

The paper is devoted to the study of particular political effects of digitalization of urban governance in the Russian Federation. Based on the concept of «surveillance capitalism» and research on the digital transformation of public administration, the authors analyzes the structure and logic of functioning of the «smart city» model using the example of Moscow. Based on the material of street protests, the political effects of the use of digital infrastructure by the city authorities, in particular, camera systems with face recognition technologies, are examined. The study of the Russian situation correlates with the latest decisions of the United Nations (UN) Human Rights Council and the European Union’s initiatives to control remote biometric recognition technologies.

A Biometric Recognition Method Using Deep CNN

Face is perhaps the first biometric trait of a person that catches one’s eye and it remains in memory for a long due to its uniqueness created by almighty. Recognizing a person using his/her face, is very natural to us and we do not need any special training for identification. But computers are programmed for analyzing things and making predictions almost in similar fashion that our brain does. Then, the recognition takes place by using some techniques and trainings. The recognition system which uses biometric properties is itself a secure and trusted technique but use of neural networks make it highly accurate and add more worth to it. A CNN model works in a fully supervised or guided environment and performs all the tasks in a robotic manner. The convolutional layer which lies in CNN model performs the complex calculation and extracts all the unique and useful features without any human involvement. I preferred to adopt Transfer learning in my work, by importing a pre-trained CNN model and I found 97.5% accuracy in recognition when I tested the model with my test samples. Keywords: Biometrics, Convolution, AlexNet, Feature Extraction, Transfer Learning