Biblio

Recently, the field of adversarial machine learning has been garnering attention by showing that state-of-the-art deep neural networks are vulnerable to adversarial examples, stemming from small perturbations being added to the input image. Adversarial examples are generated by a malicious adversary by obtaining access to the model parameters, such as gradient information, to alter the input or by attacking a substitute model and transferring those malicious examples over to attack the victim model. Specifically, one of these attack algorithms, Robust Physical Perturbations (RP2), generates adversarial images of stop signs with black and white stickers to achieve high targeted misclassification rates against standard-architecture traffic sign classifiers. In this paper, we propose BlurNet, a defense against the RP2 attack. First, we motivate the defense with a frequency analysis of the first layer feature maps of the network on the LISA dataset, which shows that high frequency noise is introduced into the input image by the RP2 algorithm. To remove the high frequency noise, we introduce a depthwise convolution layer of standard blur kernels after the first layer. We perform a blackbox transfer attack to show that low-pass filtering the feature maps is more beneficial than filtering the input. We then present various regularization schemes to incorporate this lowpass filtering behavior into the training regime of the network and perform white-box attacks. We conclude with an adaptive attack evaluation to show that the success rate of the attack drops from 90% to 20% with total variation regularization, one of the proposed defenses.

With the increase in signal's bandwidth, the conventional analog to digital converters (ADCs), operating on the basis of Shannon/Nyquist theorem, are forced to work at very high rates leading to low dynamic range and high power consumptions. This paper here tells about one Analog to Information converter developed based on compressive sensing techniques. The high sampling rates, which is the main drawback for ADCs, is being successfully reduced to 4 times lower than the conventional rates. The system is also accompanied with the advantage of low power dissipation.

This paper presents a model calibration algorithm for the modulated wideband converter (MWC) with non-ideal analog lowpass filter (LPF). The presented technique uses a test signal to estimate the finite impulse response (FIR) of the practical non-ideal LPF, and then a digital compensation filter is designed to calibrate the approximated FIR filter in the digital domain. At the cost of a moderate oversampling rate, the calibrated filter performs as an ideal LPF. The calibrated model uses the MWC system with non-ideal LPF to capture the samples of underlying signal, and then the samples are filtered by the digital compensation filter. Experimental results indicate that, without making any changes to the architecture of MWC, the proposed algorithm can obtain the samples as that of standard MWC with ideal LPF, and the signal can be reconstructed with overwhelming probability.