Despite the impressive performances reported by deep neural networks in different application domains, they remain largely vulnerable to adversarial examples, i.e., input samples that are carefully perturbed to cause misclassification at test time. In this work, we propose a deep neural rejection mechanism to detect adversarial examples, based on the idea of rejecting samples that exhibit anomalous feature representations at different network layers. With respect to competing approaches, our method does not require generating adversarial examples at training time, and it is less computationally demanding. To properly evaluate our method, we define an adaptive white-box attack that is aware of the defense mechanism and aims to bypass it. Under this worst-case setting, we empirically show that our approach outperforms previously-proposed methods that detect adversarial examples by only analyzing the feature representation provided by the output network layer.

Adversarial attacks on machine learning-based classifiers, along with defense mechanisms, have been widely studied in the context of single-label classification problems. In this paper, we shift the attention to multi-label classification, where the availability of domain knowledge on the relationships among the considered classes may offer a natural way to spot incoherent predictions, i.e., predictions associated to adversarial examples lying outside of the training data distribution. We explore this intuition in a framework in which first-order logic knowledge is converted into constraints and injected into a semi-supervised learning problem. Within this setting, the constrained classifier learns to fulfill the domain knowledge over the marginal distribution, and can naturally reject samples with incoherent predictions. Even though our method does not exploit any knowledge of attacks during training, our experimental analysis surprisingly unveils that domain-knowledge constraints can help detect adversarial examples effectively, especially if such constraints are not known to the attacker. We show how to implement an adaptive attack exploiting knowledge of the constraints and, in a specifically-designed setting, we provide experimental comparisons with popular state-of-the-art attacks. We believe that our approach may provide a significant step towards designing more robust multi-label classifiers.

No abstract

Android is targeted the most by malware coders as the number of Android users is increasing. Although there are many Android antimalware solutions available in the market, almost all of them are based on malware signatures, and more advanced solutions based on machine learning techniques are not deemed to be practical for the limited computational resources of mobile devices. In this paper we aim to show not only that the computational resources of consumer mobile devices allow deploying an efficient anti-malware solution based on machine learning techniques, but also that such a tool provides an effective defense against novel malware, for which signatures are not yet available. To this end, we first propose the extraction of a set of lightweight yet effective features from Android applications. Then, we embed these features in a vector space, and use a pre-trained machine learning model on the device for detecting malicious applications. We show that without resorting to any signatures, and relying only on a training phase involving a reasonable set of samples, the proposed system outperforms many commercial antimalware products, as well as providing slightly better performances than the most effective commercial products.

No abstract