Defence methods for image adversarial attacks

In the previous post we reviewed some well-known methods for black-box decision-based adversarial attacks where the adversary has no knowledge about the victim model except for its discrete hard-label predictions, thus gradient-based methods become ineffective but simple random-walk based methods such as the Boundary Attack can still represent a threat even under these particular settings. Now that we have introduced both white and black-box attacks … Continue reading Defence methods for image adversarial attacks

Black-box decision-based attacks on images

In the previous post we reviewed a series of black-box score-based adversarial attacks where the adversary has to estimate the gradient by querying the target model and retrieving the labels’ confidence score. In this post we are going to explore the third category of black-box attacks, namely, black-box decision-based attacks. Under this settings, the only knowledge the attacker has about the model are only discrete … Continue reading Black-box decision-based attacks on images

Black-box score-based attacks on images

In the previous post we reviewed a series of black-box transfer-based adversarial attacks where the adversary has to  generate adversarial examples against a substitute model. In this post we are going to explore the second category of black-box attacks, namely, black-box score-based attacks. Under this setting, it is not possible to access to the white-box model’s gradients. The only knowledge about the attacked model are … Continue reading Black-box score-based attacks on images

Black-box transfer-based attacks on images

In the previous post we reviewed a series of white-box adversarial attacks where the adversary has full access and knowledge of the victim model. In this post we are going to explore the first category of black-box attacks, namely, black-box transfer-based attacks. Transfer-based attacks generate adversarial examples against a substitute model, possibly being as much similar as possible to the target model, which have a … Continue reading Black-box transfer-based attacks on images

White-box adversarial attacks on images

In the first post we introduced the concept of adversarial attacks and contextualized in the case of images. In this post we are going to explore the first category of attacks, namely, white-box attacks. Under this setting, the adversary has full access and knowledge of the model, that is, the architecture of the model, it’s parameters, gradients and loss respect to the input as well … Continue reading White-box adversarial attacks on images

Introduction to adversarial attacks on images

Nowadays, image classification deep learning models are always more present in our systems in order to create smarter applications or simply to replace human operators to automatically perform some repetitive tasks. Their increased utilization is due to their high accuracy such that recent models are now able to outperform humans’ brain in many object classification tasks. However, despite their good generalization, deep neural networks are … Continue reading Introduction to adversarial attacks on images

MovieSearch: a smart movie search engine

MovieSearch is a content specific search engine with the aim to retrieve movie information given the contents of a user’s query. The search engine relies on the OkapiBM25 algorithm and takes into consideration the text present in the overview, the title, the names of the cast, and the production companies of each movie. The backend has been developed with the framework Django while the front-end … Continue reading MovieSearch: a smart movie search engine

Sentences sentiment analysis with CNN

Opinion mining (sometimes known as sentiment analysis or emotion AI) refers to the use of natural language processing, text analysis, computational linguistics, and biometrics to systematically identify, extract, quantify, and study affective states and subjective information. Sentiment analysis is widely applied to the voice of the customer materials such as reviews and survey responses, online and social media, and healthcare materials for applications that range … Continue reading Sentences sentiment analysis with CNN

Word similarity and analogy with Skip-Gram

In this post, we are going to show words similarities and words analogies learned by 3 Skip-Gram models trained to learn words embedding from a 3GB corpus size taken scraping text from Wikipedia pages. Skip-Gram is unsupervised learning used to find the context words of given a target word. During its training process, Skip-Gram will learn a powerful vector representation for all of its vocabulary … Continue reading Word similarity and analogy with Skip-Gram

RNN: Recurrent Neural Networks

In normal feed-forward neural networks the activation flows only in one direction, from the input layer to the output layer, eventually passing through a set of hidden layers. Conversely, recurrent neural networks (RNN) have also connections pointing backward, thus allowing them to take also the temporal dimension into account. This novel architecture enables them to take as their input not just the current input xi … Continue reading RNN: Recurrent Neural Networks