The rapid growth of mobile applications,the popularity of the Android system and its openness have attracted many hackers and even criminals,who are creating lots of Android malware.However,the current methods of Andr...The rapid growth of mobile applications,the popularity of the Android system and its openness have attracted many hackers and even criminals,who are creating lots of Android malware.However,the current methods of Android malware detection need a lot of time in the feature engineering phase.Furthermore,these models have the defects of low detection rate,high complexity,and poor practicability,etc.We analyze the Android malware samples,and the distribution of malware and benign software in application programming interface(API)calls,permissions,and other attributes.We classify the software’s threat levels based on the correlation of features.Then,we propose deep neural networks and convolutional neural networks with ensemble learning(DCEL),a new classifier fusion model for Android malware detection.First,DCEL preprocesses the malware data to remove redundant data,and converts the one-dimensional data into a two-dimensional gray image.Then,the ensemble learning approach is used to combine the deep neural network with the convolutional neural network,and the final classification results are obtained by voting on the prediction of each single classifier.Experiments based on the Drebin and Malgenome datasets show that compared with current state-of-art models,the proposed DCEL has a higher detection rate,higher recall rate,and lower computational cost.展开更多
A novel feature fusion method is proposed for the edge detection of color images. Except for the typical features used in edge detection, the color contrast similarity and the orientation consistency are also selected...A novel feature fusion method is proposed for the edge detection of color images. Except for the typical features used in edge detection, the color contrast similarity and the orientation consistency are also selected as the features. The four features are combined together as a parameter to detect the edges of color images. Experimental results show that the method can inhibit noisy edges and facilitate the detection for weak edges. It has a better performance than conventional methods in noisy environments.展开更多
In this paper,based on a bidirectional parallel multi-branch feature pyramid network(BPMFPN),a novel one-stage object detector called BPMFPN Det is proposed for real-time detection of ground multi-scale targets by swa...In this paper,based on a bidirectional parallel multi-branch feature pyramid network(BPMFPN),a novel one-stage object detector called BPMFPN Det is proposed for real-time detection of ground multi-scale targets by swarm unmanned aerial vehicles(UAVs).First,the bidirectional parallel multi-branch convolution modules are used to construct the feature pyramid to enhance the feature expression abilities of different scale feature layers.Next,the feature pyramid is integrated into the single-stage object detection framework to ensure real-time performance.In order to validate the effectiveness of the proposed algorithm,experiments are conducted on four datasets.For the PASCAL VOC dataset,the proposed algorithm achieves the mean average precision(mAP)of 85.4 on the VOC 2007 test set.With regard to the detection in optical remote sensing(DIOR)dataset,the proposed algorithm achieves 73.9 mAP.For vehicle detection in aerial imagery(VEDAI)dataset,the detection accuracy of small land vehicle(slv)targets reaches 97.4 mAP.For unmanned aerial vehicle detection and tracking(UAVDT)dataset,the proposed BPMFPN Det achieves the mAP of 48.75.Compared with the previous state-of-the-art methods,the results obtained by the proposed algorithm are more competitive.The experimental results demonstrate that the proposed algorithm can effectively solve the problem of real-time detection of ground multi-scale targets in aerial images of swarm UAVs.展开更多
图像异常检测旨在识别并定位图像中的异常区域,针对现有算法中不同层次特征信息利用不充分的问题,提出了基于多层次特征融合网络的图像异常检测算法。通过使用融合了异常先验知识的伪异常数据生成算法,对训练集进行了异常数据扩充,将异...图像异常检测旨在识别并定位图像中的异常区域,针对现有算法中不同层次特征信息利用不充分的问题,提出了基于多层次特征融合网络的图像异常检测算法。通过使用融合了异常先验知识的伪异常数据生成算法,对训练集进行了异常数据扩充,将异常检测任务转化为监督学习任务;构建了多层次特征融合网络,将神经网络中不同层次特征进行融合,丰富了特征中的低层纹理信息和高层语义信息,使得用于异常检测的特征更具区分性;训练时,设计了分数约束损失和一致性约束损失,并结合特征约束损失对整个网络模型进行训练。实验结果表明,MVTec数据集上图像级检测接收机工作特性曲线下面积(area under the receiver operating characteristic, AUROC)平均值为98.7%,像素级定位AUROC平均值为97.9%,每区域重叠率平均值为94.2%,均高于现有的异常检测算法。展开更多
随着社交网络平台的迅速发展,网络欺凌问题日益突出,文本与图片相结合的多样化网络表达形式提高了网络欺凌的检测和治理难度.构建了一个包含文本和图片的中文多模态网络欺凌数据集,将BERT(bidirectional encoder representations from t...随着社交网络平台的迅速发展,网络欺凌问题日益突出,文本与图片相结合的多样化网络表达形式提高了网络欺凌的检测和治理难度.构建了一个包含文本和图片的中文多模态网络欺凌数据集,将BERT(bidirectional encoder representations from transformers)模型与ResNet50模型相结合,分别提取文本和图片的单模态特征,并进行决策层融合,对融合后的特征进行检测,实现了对网络欺凌与非网络欺凌2个类别的文本和图片的准确识别.实验结果表明,提出的多模态网络欺凌检测模型能够有效识别出包含文本与图片的具有网络欺凌性质的社交网络帖子或者评论,提高了多模态形式网络欺凌检测的实用性、准确性和效率,为社交网络平台的网络欺凌检测和治理提供了一种新的思路和方法,有助于构建更加健康、文明的网络环境.展开更多
近年来,卷积神经网络在合成孔径雷达(synthetic aperture radar,SAR)图像船舶检测中取得突出成就,但小目标检测方面仍然存在较大不足。对此,提出一种基于YOLO(you only look once)v5的改进检测网络,结合空间感知通道注意力、自注意力机...近年来,卷积神经网络在合成孔径雷达(synthetic aperture radar,SAR)图像船舶检测中取得突出成就,但小目标检测方面仍然存在较大不足。对此,提出一种基于YOLO(you only look once)v5的改进检测网络,结合空间感知通道注意力、自注意力机制和上下文特征融合策略,以提高小型船舶的检测性能。首先,通道注意力机制抑制了背景信息并强调目标特征,显著提高检测精度。其次,在YOLOv5的骨干网络和检测层中引入自注意力模块,以捕获全局信息,增强定位能力。最后,通过融合浅层和深层特征,补充特征提取中丢失的小目标信息,进一步提高检测精度。基于大规模SAR船舶监测数据集(large-scale SAR ship detection dataset version 1.0 LSSSDDv1.0)数据集的实验结果表明,改进后的网络的全类平均精度(mean average precision,mAP)0.5指标达78.9%,显著优于现有方法。展开更多
基金supported by the National Natural Science Foundation of China(62072255)。
文摘The rapid growth of mobile applications,the popularity of the Android system and its openness have attracted many hackers and even criminals,who are creating lots of Android malware.However,the current methods of Android malware detection need a lot of time in the feature engineering phase.Furthermore,these models have the defects of low detection rate,high complexity,and poor practicability,etc.We analyze the Android malware samples,and the distribution of malware and benign software in application programming interface(API)calls,permissions,and other attributes.We classify the software’s threat levels based on the correlation of features.Then,we propose deep neural networks and convolutional neural networks with ensemble learning(DCEL),a new classifier fusion model for Android malware detection.First,DCEL preprocesses the malware data to remove redundant data,and converts the one-dimensional data into a two-dimensional gray image.Then,the ensemble learning approach is used to combine the deep neural network with the convolutional neural network,and the final classification results are obtained by voting on the prediction of each single classifier.Experiments based on the Drebin and Malgenome datasets show that compared with current state-of-art models,the proposed DCEL has a higher detection rate,higher recall rate,and lower computational cost.
基金supported partly by the National Basic Research Program of China (2005CB724303)the National Natural Science Foundation of China (60671062) Shanghai Leading Academic Discipline Project (B112).
文摘A novel feature fusion method is proposed for the edge detection of color images. Except for the typical features used in edge detection, the color contrast similarity and the orientation consistency are also selected as the features. The four features are combined together as a parameter to detect the edges of color images. Experimental results show that the method can inhibit noisy edges and facilitate the detection for weak edges. It has a better performance than conventional methods in noisy environments.
文摘In this paper,based on a bidirectional parallel multi-branch feature pyramid network(BPMFPN),a novel one-stage object detector called BPMFPN Det is proposed for real-time detection of ground multi-scale targets by swarm unmanned aerial vehicles(UAVs).First,the bidirectional parallel multi-branch convolution modules are used to construct the feature pyramid to enhance the feature expression abilities of different scale feature layers.Next,the feature pyramid is integrated into the single-stage object detection framework to ensure real-time performance.In order to validate the effectiveness of the proposed algorithm,experiments are conducted on four datasets.For the PASCAL VOC dataset,the proposed algorithm achieves the mean average precision(mAP)of 85.4 on the VOC 2007 test set.With regard to the detection in optical remote sensing(DIOR)dataset,the proposed algorithm achieves 73.9 mAP.For vehicle detection in aerial imagery(VEDAI)dataset,the detection accuracy of small land vehicle(slv)targets reaches 97.4 mAP.For unmanned aerial vehicle detection and tracking(UAVDT)dataset,the proposed BPMFPN Det achieves the mAP of 48.75.Compared with the previous state-of-the-art methods,the results obtained by the proposed algorithm are more competitive.The experimental results demonstrate that the proposed algorithm can effectively solve the problem of real-time detection of ground multi-scale targets in aerial images of swarm UAVs.
文摘图像异常检测旨在识别并定位图像中的异常区域,针对现有算法中不同层次特征信息利用不充分的问题,提出了基于多层次特征融合网络的图像异常检测算法。通过使用融合了异常先验知识的伪异常数据生成算法,对训练集进行了异常数据扩充,将异常检测任务转化为监督学习任务;构建了多层次特征融合网络,将神经网络中不同层次特征进行融合,丰富了特征中的低层纹理信息和高层语义信息,使得用于异常检测的特征更具区分性;训练时,设计了分数约束损失和一致性约束损失,并结合特征约束损失对整个网络模型进行训练。实验结果表明,MVTec数据集上图像级检测接收机工作特性曲线下面积(area under the receiver operating characteristic, AUROC)平均值为98.7%,像素级定位AUROC平均值为97.9%,每区域重叠率平均值为94.2%,均高于现有的异常检测算法。
文摘随着社交网络平台的迅速发展,网络欺凌问题日益突出,文本与图片相结合的多样化网络表达形式提高了网络欺凌的检测和治理难度.构建了一个包含文本和图片的中文多模态网络欺凌数据集,将BERT(bidirectional encoder representations from transformers)模型与ResNet50模型相结合,分别提取文本和图片的单模态特征,并进行决策层融合,对融合后的特征进行检测,实现了对网络欺凌与非网络欺凌2个类别的文本和图片的准确识别.实验结果表明,提出的多模态网络欺凌检测模型能够有效识别出包含文本与图片的具有网络欺凌性质的社交网络帖子或者评论,提高了多模态形式网络欺凌检测的实用性、准确性和效率,为社交网络平台的网络欺凌检测和治理提供了一种新的思路和方法,有助于构建更加健康、文明的网络环境.
文摘近年来,卷积神经网络在合成孔径雷达(synthetic aperture radar,SAR)图像船舶检测中取得突出成就,但小目标检测方面仍然存在较大不足。对此,提出一种基于YOLO(you only look once)v5的改进检测网络,结合空间感知通道注意力、自注意力机制和上下文特征融合策略,以提高小型船舶的检测性能。首先,通道注意力机制抑制了背景信息并强调目标特征,显著提高检测精度。其次,在YOLOv5的骨干网络和检测层中引入自注意力模块,以捕获全局信息,增强定位能力。最后,通过融合浅层和深层特征,补充特征提取中丢失的小目标信息,进一步提高检测精度。基于大规模SAR船舶监测数据集(large-scale SAR ship detection dataset version 1.0 LSSSDDv1.0)数据集的实验结果表明,改进后的网络的全类平均精度(mean average precision,mAP)0.5指标达78.9%,显著优于现有方法。
