Malware remains a big threat to cyber security, calling for machine learning based malware detection. While promising, such detectors are known be vulnerable evasion attacks. Ensemble typically facilitates countermeasures, while attackers can leverage this technique improve attack effectiveness as well. This motivates us investigate which kind of robustness the ensemble defense or achieve, particularly when they combat with each other. We thus propose new approach, named mixture attacks, by...

Machine learning-based malware detection is known to be vulnerable adversarial evasion attacks. The state-of-the-art that there are no effective defenses against these As a response the classification challenge organized by MIT Lincoln Lab and associated with AAAI-19 Workshop on Artificial Intelligence for Cyber Security (AICS'2019), we propose six guiding principles enhance robustness of deep neural networks. Some have been scattered in literature, but others introduced this paper first...

Machine Learning (ML) techniques can facilitate the automation of malicious software (malware for short) detection, but suffer from evasion attacks.Many studies counter such attacks in heuristic manners, lacking theoretical guarantees and defense effectiveness.In this paper, we propose a new adversarial training framework, termed Principled Adversarial Malware Detection (PAD), which offers convergence robust optimization methods.PAD lays on learnable convex measurement that quantifies...

The Internet of Things (IoT) has grown rapidly due to artificial intelligence driven edge computing. While enabling many new functions, computing devices expand the vulnerability surface and have become target malware attacks. Moreover, attackers used advanced techniques evade defenses by transforming their into functionality-preserving variants. We systematically analyze such evasion attacks conduct a large-scale empirical study in this paper evaluate impact on security. More specifically,...

Adversarial machine learning in the context of image processing and related applications has received a large amount attention. However, adversarial learning, especially deep malware detection much less attention despite its apparent importance. In this paper, we present framework for enhancing robustness Deep Neural Networks (DNNs) against samples, dubbed Hashing Transformation Networks} (HashTran-DNN). The core idea is to use hash functions with certain locality-preserving property...

Malware continues to be a major cyber threat, despite the tremendous effort that has been made combat them. The number of malware in wild steadily increases over time, meaning we must resort automated defense techniques. This naturally calls for machine learning based detection. However, is known vulnerable adversarial evasion attacks manipulate small features make classifiers wrongly recognize sample as benign one. state-of-the-art there are no effective countermeasures against these...

The deep learning approach to detecting malicious software (malware) is promising but has yet tackle the problem of dataset shift, namely that joint distribution examples and their labels associated with test set different from training set. This causes degradation models without users' notice. In order alleviate problem, one let a classifier not only predict label on given example also present its uncertainty (or confidence) predicted label, whereby defender can decide whether use or not....

Abstract Hydrovoltaic electricity generators (HEGs), which can harvest clean energy from the ubiquitous evaporation of water, have recently attracted significant interest. The utilization renewable porous aerogels in development HEGs enhance their sustainability and performance. Herein, an efficient HEG based on ambient‐dried composite (ADAs) composed nanocellulose carbon nanotubes (CNTs) is presented. abundant carboxyl groups CNTs enable electrostatic complexation with metal ions. This not...

With the rapid growing of crowdsourcing systems, class labels for supervised learning can be easily obtained from platforms. To deal with problem that crowds are usually noisy due to imperfect reliability non-expert workers, we let multiple workers provide same object. Then, true labeled object estimated through ground truth inference algorithms. The inferred integrated expected high quality. In this paper, propose a novel algorithm based on EM algorithm, which not only infers instances but...

With the development of artificial intelligence (AI) in field drug design and discovery, learning informative representations molecules is becoming crucial for those AI-driven tasks. In recent years, graph neural networks (GNNs) have emerged as a preferred choice deep architecture been successfully applied to molecular representation (MRL). Up-to-date MRL methods directly apply message passing mechanism on atom-level attributes (i.e., atoms bonds) molecules. However, they neglect latent yet...

Discrete wavelet transform (DWT) is sensitive to the translation/shift of input signals, so its effectiveness could be negatively impacted when we encounter translation among signals. To deal with such drawbacks, this paper proposes redundant DWT(RDWT) based method achieve image registration, invariant feature extraction and face recognition. We select a representative from each person form reference set perform DWT on it. For test face, RDWT compare horizontal vertical details corresponding...

In this paper, we present a novel method, namely SimWalk, to learn latent representations of networks. SimWalk maps nodes continuous vector space which maximizes the likelihood node sequences. We design probability-guided random walk procedure based on relation similarity, encourages sequences preserve context-related neighborhoods. Different with previous work generates rigid sequences, believe that relations in social net­works, especially can guide generate more linguistic sequence....

The Industrial 5.0 Model integrates enabling technologies such as deep learning, digital twins, and the meta-universe with new development concepts. However, model data security may pose challenges for developing zero-defect production other industrial manufacturing industries. To address this issue, we generate adversarial examples using a one-pixel attack in machine which can fool defect detection classification model. traditional based on Differential Evolution (DE) algorithm has limited...

This aiming at improving the lossless compression ratio of hyperspectral image, a three-dimensional LMS (3DLMS) algorithm is first deduced and applied into field image compression. A novel adaptive prediction model based on 3DLMS for proposed optimized by local casual set mean subtraction method. Experimental results AVIRIS images show that can remove both spatial spectral redundancy achieve higher ratios than other state-of-the-art algorithms. The feasibility in signal processing also...

We propose a novel learning algorithm, called Bagging-Adaboost ensemble algorithm with genetic post optimization, for object detection that uses local shape-based feature. The feature is motivated by the scheme use chamfer distance as shape comparison measure. It can be calculated very quickly using look-up table. Random sampling boosting used to select discriminative edge features set from over-complete dictionary of and form an detector. Genetic optimization procedure remove based...

We present a learning model for object detection that uses novel local edge features. The features are motivated by the scheme use chamfer distance as shape comparison measure. can be calculated very quickly using look-up table. Adaboost algorithm is used to select discriminative set from an over-complete pool and combine them form detector. To demonstrate our method we trained system detect car in complex natural scenes single model. Experimental results show extremely rapidly objects...