New framework for unsupervised universal steganalysis via SRISP-aided outlier detection

Formulating steganalysis as a binary classification problem has been highly successful. However, the existing detection algorithms are difficult to obtain high detection accuracy when applied in real-world circumstances. Because so-called model mismatch problem often occurs owing to unknown cover source and embedding parameters. To avoid the mess of model mismatch, we propose a new unsupervised universal steganalysis framework to detect individual stego images. First, cover images with statistical properties similar to those of the given test image are searched from a retrieval cover database to establish an aided cover sample set. Second, unsupervised outlier detection is performed on a test set composed of the given test image and its aided cover sample set to determine the type (cover or stego) of the given test image. Our proposed framework, called Similarity Retrieval of Image Statistical Properties (SRISP)-aided unsupervised outlier detection, requires no training, and thus it does not suffer from model mismatch. The framework employs standard steganalysis features and detects each test image individually. Experimental results illustrate that the framework substantially outperforms one-class support vector machine and the traditional unsupervised outlier detectors without considering SRISP; its detection performance is independent of the proportion of stego images in the test samples.     

基于K-means和MTLS-SVM算法的生理参数监测系统

在非医模式的生理参数监测系统中,对监测参数进行学习,可以提高诊断和预测精度.针对多任务时间序列中存在的信息挖掘不充分、预测精度低等问题,将机器学习中的监督和半监督学习方式结合起来对远程健康监护对象进行生理状况预测.该方法用K-means算法将相同类别的数据集群,并使用多任务最小二乘支持向量机(MTLS-SVM)来训练历史数据来进行趋势预测.为了评估该方法的有效性,将MTLS-SVM方法与K-means、MTLS-SVM方法比较,实验结果表明该方法具有较高的预测精度.     

融入环形振荡器木马特征的无监督硬件木马检测

机器学习用于集成电路硬件木马的检测可以有效提高检测率。无监督学习方法在特征选择上还存在不足,目前研究工作主要集中于有监督学习方法。文章引入环形振荡器木马的新特征,研究基于无监督机器学习的硬件木马检测方法。首先针对待测电路网表,提取每个节点的5维特征值,然后利用局部离群因子(LOF)算法计算各节点的LOF值,筛选出硬件木马节点。对Trust-HUB基准电路的仿真实验结果表明,该方法用于网表级电路硬件木马的检测,与现有基于无监督学习的检测方法相比,TPR(真阳性率)、P(精度)和F(度量)分别提升了16.19%、10.79%和15.56%。针对Trust-HUB基准电路的硬件木马检测的平均TPR、TNR和A,分别达到了58.61%、97.09%和95.60%。     

基于极限学习机的遥感影像阴影检测

为提高阴影检测精度,采用基于极限学习机的遥感影像阴影检测方法,将纹理特征的能量、熵、对比度和逆差距4个指数作为输入特征向量训练学习机模型,实现阴影区域的检测。实验结果表明,当纹理窗口大小为4×4,步长为1时,能够将阴影区域较好地检测出来。该方法能够克服阈值法的地物错检问题以及反复选取阈值的缺点,并解决了神经网络及支持向量机多参数选取困难的问题,有效地提高了阴影区域检测速度和精度,具有较好的鲁棒性和泛化性能,综合性能优于阈值法、神经网络和支持向量机。     

基于Deep Belief Nets方法的中文名实体分类研究

DBN是一种快速全局最优的神经网络分类方法,包含数层无监督学习网络和一层有监督学习网络。本文验证了DBN方法很好地适用于中文名实体分类任务。首先,采用多层RBM方法无监督地从字特征向量提取结构信息,得到更具有表征能力的特征;然后,利用BP方法微调网络参数并对提取后的特征向量进行分类,以此构成分类器进行名实体分类。通过对ACE 04的中文名实体进行的分类测试,准确率达到91.45%,明显高于支持向量机和反向传播神经网络等传统分类算法。     

An efficient deep sclera recognition framework with novel sclera segmentation,vessel extraction and gaze detection

Sclera recognition is a promising ocular biometric modality because of contact-less, gaze-independent image acquisition in visible light. Moreover, it is unaffected even if the subjects are wearing contact lenses in eyes. However, it is a difficult task because several steps are required, each of which must be performed accurately and efficiently. In this work, sclera recognition is performed in the following steps, namely, segmentation of sclera region, extraction of sclera vasculature pattern, detection of gaze direction and finally comparison of two vasculature patterns for matching and recognition. The proposed segmentation model DSeg is based on well-known deep learning model UNet and reduces model complexity by creating a Knowledge Base of sclera and non-sclera colors. DSeg is a lightweight and environment-friendly model, which outperforms UNet in terms of speed, efficiency and accuracy. Two rule-based unsupervised vessel extraction methods require prior sclera segmentation and exhibit competing recognition performance to a supervised deep model for vessel extraction, which does not require prior sclera segmentation. A novel deep recognition model is proposed which compares two vessel structures taking into account their affine-transformation, and produces a single Boolean output to decide whether the structures match or not. The model does not require post logic in the matching process. The model is further improved to detect errors in prediction. We achieve best recognition rates with low false-acceptance-rates for two sets of training and validation, using the publicly available dataset SBVPI and the best achieved AUC score is 0.98.     

Quantitative Comparison of Spot Detection Methods in Fluorescence Microscopy

Quantitative analysis of biological image data generally involves the detection of many subresolution spots. Especially in live cell imaging, for which fluorescence microscopy is often used, the signal-to-noise ratio (SNR) can be extremely low, making automated spot detection a very challenging task. In the past, many methods have been proposed to perform this task, but a thorough quantitative evaluation and comparison of these methods is lacking in the literature. In this paper, we evaluate the performance of the most frequently used detection methods for this purpose. These include seven unsupervised and two supervised methods. We perform experiments on synthetic images of three different types, for which the ground truth was available, as well as on real image data sets acquired for two different biological studies, for which we obtained expert manual annotations to compare with. The results from both types of experiments suggest that for very low SNRs $(approx 2)$, the supervised (machine learning) methods perform best overall. Of the unsupervised methods, the detectors based on the so-called $h$ -dome transform from mathematical morphology or the multiscale variance-stabilizing transform perform comparably, and have the advantage that they do not require a cumbersome learning stage. At high SNRs $(> 5)$, the difference in performance of all considered detectors becomes negligible.      

Ubiquitously Supervised Subspace Learning

In this paper, our contributions to the subspace learning problem are two-fold. We first justify that most popular subspace learning algorithms, unsupervised or supervised, can be unitedly explained as instances of a ubiquitously supervised prototype. They all essentially minimize the intraclass compactness and at the same time maximize the interclass separability, yet with specialized labeling approaches, such as ground truth, self-labeling, neighborhood propagation, and local subspace approximation. Then, enlightened by this ubiquitously supervised philosophy, we present two categories of novel algorithms for subspace learning, namely, misalignment-robust and semi-supervised subspace learning. The first category is tailored to computer vision applications for improving algorithmic robustness to image misalignments, including image translation, rotation and scaling. The second category naturally integrates the label information from both ground truth and other approaches for unsupervised algorithms. Extensive face recognition experiments on the CMU PIE and FRGC ver1.0 databases demonstrate that the misalignment-robust version algorithms consistently bring encouraging accuracy improvements over the counterparts without considering image misalignments, and also show the advantages of semi-supervised subspace learning over only supervised or unsupervised scheme.     

A framework for medical image retrieval using machine learning and statistical similarity matching techniques with relevance feedback.

A content-based image retrieval (CBIR) framework for diverse collection of medical images of different imaging modalities, anatomic regions with different orientations and biological systems is proposed. Organization of images in such a database (DB) is well defined with predefined semantic categories; hence, it can be useful for category-specific searching. The proposed framework consists of machine learning methods for image prefiltering, similarity matching using statistical distance measures, and a relevance feedback (RF) scheme. To narrow down the semantic gap and increase the retrieval efficiency, we investigate both supervised and unsupervised learning techniques to associate low-level global image features (e.g., color, texture, and edge) in the projected PCA-based eigenspace with their high-level semantic and visual categories. Specially, we explore the use of a probabilistic multiclass support vector machine (SVM) and fuzzy c-mean (FCM) clustering for categorization and prefiltering of images to reduce the search space. A category-specific statistical similarity matching is proposed in a finer level on the prefiltered images. To incorporate a better perception subjectivity, an RF mechanism is also added to update the query parameters dynamically and adjust the proposed matching functions. Experiments are based on a ground-truth DB consisting of 5000 diverse medical images of 20 predefined categories. Analysis of results based on cross-validation (CV) accuracy and precision-recall for image categorization and retrieval is reported. It demonstrates the improvement, effectiveness, and efficiency achieved by the proposed framework.     

基于生成对抗网络的无监督域适应分类模型

生成适应模型利用生成对抗网络实现模型结构,并在领域适应学习上取得了突破.但其部分网络结构缺少信息交互,且仅使用对抗学习不足以完全减小域间距离,从而使分类精度受到影响.为此,提出一种基于生成对抗网络的无监督域适应分类模型（Unsupervised Domain Adaptation classification model based on GAN,UDAG）.该模型通过联合使用生成对抗网络和多核最大均值差异度量准则优化域间差异,并充分利用无监督对抗训练及监督分类训练之间的信息传递以学习源域分布和目标域分布之间的共享特征.通过在四种域适应情况下的实验结果表明,UDAG模型学习到更优的共享特征嵌入并实现了域适应图像分类,且分类精度有明显提高.     

基于改进GMM和多特征融合的视频火焰检测算法

针对现有视频图像火焰检测算法前景提取不完整、准确率低和误检率高等问题,提出一种基于改进混合高斯模型(GMM)和多特征融合的视频火焰检测算法。首先针对背景建模,提出了自适应高斯分布数和学习率的改进GMM方法,以提高前景提取效果和算法实时性;然后利用火焰颜色特征筛选出疑似火焰区域,再通过融合改进局部二值模式纹理和边缘相似度特征用于火焰检测。基于支持向量机设计火焰融合特征分类器并进行对比实验,在公开数据集上的实验结果表明,所提算法有效提高了背景建模效果,火焰检测准确率可达到92.26%,误检率低至2.43%。     

Genetic Sequence Classification and its Application to Cross-Species Homology Detection

Although large-scale classification studies of genetic sequence data are in progress around the world, very few studies compare different classification approaches, e.g. unsupervised and supervised, in terms of objective criteria such as classification accuracy and computational complexity. In this paper, we study such criteria for both unsupervised and supervised classification of a relatively large sequence data set. The unsupervised approach involves use of different sequence alignment algorithms (e.g., Smith-Waterman, FASTA and BLAST) followed by clustering using the Maximin algorithm. The supervised approach uses a suitable numeric encoding (relative frequencies of tuples of nucleotides followed by principal component analysis) which is fed to a Multi-layer Backpropagation Neural Network. Classification experiments conducted on IBM-SP parallel computers show that FASTA with unsupervised Maximin leads to best trade-off between accuracy and speed among all methods, followed by supervised neural networks as the second best approach. Finally, the different classifiers are applied to the problem of cross-species homology detection.     

改进自训练模型在业务质差用户识别中的应用

质差用户识别是降低用户投诉率、提升用户满意度的重要环节。针对当前电信网络系统中业务感知相关的大量结构化及非结构化数据难以有效标注、质差用户标签不完备、现有监督学习模型训练样本不均衡而导致质差识别率低的问题,采用改进自训练半监督学习模型,利用少量满意度低分和投诉用户作为质差用户标签对网络数据进行标注,并通过标签迁移对大量未标注数据进行训练识别质差用户。实验表明,相比于识别准确率高但是训练成本高的全监督学习和识别准确率低的无监督学习,半监督学习可以充分利用无标签样本数据进行有效训练,保证较低训练成本的同时显著提升质差用户识别准确率。     

Clustering-based label estimation for network anomaly detection

A substantial body of work has been done to identify network anomalies using supervised and unsupervised learning techniques with their unique strengths and weaknesses. In this work, we propose a new approach that takes advantage of both worlds of unsupervised and supervised learnings. The main objective of the proposed approach is to enable supervised anomaly detection without the provision of the associated labels by users. To this end, we estimate the labels of each connection in the training phase using clustering. The “estimated” labels are then utilized to establish a supervised learning model for the subsequent classification of connections in the testing stage. We set up a new property that defines anomalies in the context of network anomaly detection to improve the quality of estimated labels. Through our extensive experiments with a public dataset (NSL-KDD), we will prove that the proposed method can achieve performance comparable to one with the “original” labels provided in the dataset. We also introduce two heuristic functions that minimize the impact of the randomness of clustering to improve the overall quality of the estimated labels.     

Pain detection using batch normalized discriminant restricted Boltzmann machine layers

A system for automatic pain detection whereby pain-related features are extracted from facial images using a four-layer Convolutional Deep Belief Network (CDBN) is proposed in this study. The CDBN is trained by greedy layer-wise procedure whereby each added layer is trained as a Convolutional Restricted Boltzmann Machine (CRBM) by contrastive divergence. Since conventional CRBM is trained in a purely unsupervised manner, there is no guarantee that learned features are appropriate for the supervised task at hand. A discriminative objective based on between-class and within-class distances is proposed to adapt CRBM to learn task-related features. When discriminative and generative objectives are appropriately combined, a competitive classification performance can be achieved. Moreover, we introduced batch normalization (BN) units in the structure of the CRBM model to smooth optimization landscape and speed up the learning process. BN units come right before sigmoid units. Extracted features are then used to train a linear SVM to classify each frame into pain or no-pain classes. Extensive experiments on UNBC-McMaster Shoulder Pain database demonstrate the effectiveness of the proposed method for automatic pain detection.     

主动学习半监督聚类入侵检测算法

针对聚类的入侵检测算法误报率高的问题,提出一种主动学习半监督聚类入侵检测算法.在半监督聚类过程中应用主动学习策略,主动查询网络中未标记数据与标记数据的约束关系,利用少量的标记数据生成正确的样本模型来指导大量的未标记数据聚类,对聚类后仍未能标记的数据采用改进的K-近邻法进一步确定未标记数据的类型,实现对新攻击类型的检测.实验结果表明了算法的可行性及有效性.     

基于特征解耦的无监督水下图像增强

水介质的吸收和散射特性致使水下图像存在不同类型的失真,严重影响后续处理的准确性和有效性。目前有监督学习的水下图像增强方法依靠合成的水下配对图像集进行训练,然而由于合成的数据可能无法准确地模拟水下成像的基本物理机制,所以监督学习的方法很难应用于实际的应用场景。该文提出一种基于特征解耦的无监督水下图像增强方法,一方面,考虑获取同一场景下的清晰-非清晰配对数据集难度大且成本高,提出采用循环生成对抗网络将水下图像增强问题转换成风格迁移问题,实现无监督学习;另一方面,结合特征解耦方法分别提取图像的风格特征和结构特征,保证增强前后图像的结构一致性。实验结果表明,该方法可以在非配对数据训练的情况下,能够有效恢复水下图像的颜色和纹理细节。     

Hadoop‐based analytic framework for cyber forensics

With an exponential increase in the data size and complexity of various documents to be investigated, existing methods of network forensics are found not much efficient with respect to accuracy and detection ratio. The existing techniques for network forensic analysis exhibit inherent limitations while processing a huge volume, variety, and velocity of data. It makes network forensic a time‐consuming and resource‐consuming task. To balance time taken and output delivered, these existing techniques put a limit on the amount of data under analysis, which results in a polynomial time complexity of these solutions. So to mitigate these issues, in this paper, we propose an effective framework to overcome the limitation to handle large volume, variety, and velocity of data. An architectural setup that consists of MapReduce framework on top of Hadoop Distributed File System environment is proposed in this paper. The proposed framework demonstrates its capability to handle issues of storage and processing of big data using cloud computing. Also, in the proposed framework, supervised machine learning (random forest‐based decision tree) algorithm has been implemented to demonstrate better sensitivity. To train and validate the model, online available data set from CAIDA is taken and university network traffic samples, with increasing size, has been taken for experiment. Results thus obtained confirm the superiority of the proposed framework in network forensics, with an average accuracy of 99.34% (malicious and nonmalicious traffic).     

Weakly supervised target detection in remote sensing images based on transferred deep features and negative bootstrapping

Target detection in remote sensing images (RSIs) is a fundamental yet challenging problem faced for remote sensing images analysis. More recently, weakly supervised learning, in which training sets require only binary labels indicating whether an image contains the object or not, has attracted considerable attention owing to its obvious advantages such as alleviating the tedious and time consuming work of human annotation. Inspired by its impressive success in computer vision field, in this paper, we propose a novel and effective framework for weakly supervised target detection in RSIs based on transferred deep features and negative bootstrapping. On one hand, to effectively mine information from RSIs and improve the performance of target detection, we develop a transferred deep model to extract high-level features from RSIs, which can be achieved by pre-training a convolutional neural network model on a large-scale annotated dataset (e.g. ImageNet) and then transferring it to our task by domain-specifically fine-tuning it on RSI datasets. On the other hand, we integrate negative bootstrapping scheme into detector training process to make the detector converge more stably and faster by exploiting the most discriminative training samples. Comprehensive evaluations on three RSI datasets and comparisons with state-of-the-art weakly supervised target detection approaches demonstrate the effectiveness and superiority of the proposed method.