In this paper, we propose a highly efficient algorithm to model the human skin color. The underlying algorithm involves generating a discrete Cosine transform (DCT) at each pixel location, using the surrounding points...
详细信息
In this paper, we propose a highly efficient algorithm to model the human skin color. The underlying algorithm involves generating a discrete Cosine transform (DCT) at each pixel location, using the surrounding points. These DCT coefficients are assumed to follow a generalized Gaussian distribution (GGD). Next, the model parameters are estimated using the maximum-likelihood (ML) criterion applied to a set of training skin samples. Finally, each pixel is classified as skin or the opposite if its likelihood ratio is above some threshold. The experimental results show that our model avoids excessive false detection while still retaining a high degree of correct detection.
Intrusion detection systems (IDSs) can easily create thousands of alerts per day, up to 99% of which are false positives (i.e. alerts that are triggered incorrectly by benign events). This makes it extremely difficult...
详细信息
Intrusion detection systems (IDSs) can easily create thousands of alerts per day, up to 99% of which are false positives (i.e. alerts that are triggered incorrectly by benign events). This makes it extremely difficult for managers to analyze and react to attacks. This paper presents a novel method for handling IDS alerts more efficiently. It introduces outlier detection technique into this field, and designs a special outlier detection algorithm for identifying true alerts and reducing false positives. This algorithm uses frequent attribute values mined from historical alerts as the features of false positives, and then filters false alerts by the score calculated based on these features. We also proposed a two-phrase framework, which not only can filter newcome alerts in real time, but also can learn from these alerts and automatically adjust the filtering mechanism to new situations. Moreover our method needs no domain knowledge and little human assistance, so it is more practical than current ways. We have built a prototype implementation of our method. And the experiments on DARPA 2000 and real-world data have proved that this model has high performance.
An onset detection system based on linear prediction with scalable complexity is proposed in this work. One unique feature of the proposed onset detection algorithm is that it can offer a trade-off between complexity ...
详细信息
An onset detection system based on linear prediction with scalable complexity is proposed in this work. One unique feature of the proposed onset detection algorithm is that it can offer a trade-off between complexity and detection accuracy by adjusting its parameters. Consequently, it can be used in consumer electronics such as karaoke performance evaluation and automatic visual effect generation in portable media players.
In this paper, we propose a novel detection algorithm for the elementary signal estimator of an IDMA system considering channel estimation error. To develop the algorithm, we derive new probability density function of...
详细信息
In this paper, we propose a novel detection algorithm for the elementary signal estimator of an IDMA system considering channel estimation error. To develop the algorithm, we derive new probability density function of decision variable reflecting channel estimation error and modify the conventional algorithm based on it. Through computer simulations, it is shown that the proposed algorithm achieves lower bit error rate than the conventional one. This performance enhancement is provided with negligible increase of its computational complexity.
A spectral clustering intrusion detection approach is presented in this paper. The basic idea of the approach is to compute the similarities between the training data points, then to construct the affinity matrix, and...
详细信息
A spectral clustering intrusion detection approach is presented in this paper. The basic idea of the approach is to compute the similarities between the training data points, then to construct the affinity matrix, and to get the clusters according the main eigenvector of this affinity matrix. With the classified data instances, anomaly data clusters can be easily identified by normal cluster ratio. The benefits of the approach lie in that it is accurate in clustering and it needn 't labeled training data sets. Using the data sets of KDD99, the experiment result shows that this approach can detect intrusions efficiently in the real network connections.
The comprehension of the clustering result is a problem that hasn't yet resolve, which having important meaning to the usage of the cluster result and the evaluation of the cluster effect. We put forward the metho...
详细信息
The comprehension of the clustering result is a problem that hasn't yet resolve, which having important meaning to the usage of the cluster result and the evaluation of the cluster effect. We put forward the method discovering attribute feature cluster for any clustering result based on outlier detection technique, and put forward an outlier detection algorithm based on even distribution pattern. Through carrying on outlier analysis to all data cluster attribute descriptions, we discovered the feature attribute of each data cluster, and then carried out the comprehension to the clustering result. The remarkable point lies in the method doesn't only aim at a particular clustering algorithm, but also the analysis of any the clustering algorithm result. Experiment to the UCI data set indicated, the method submitted in this paper obtained better result.
It is not surprising that the process of change detection is fundamental to many machine vision applications. Most change detection algorithms assume that the illumination on a scene will remain constant. Unfortunatel...
详细信息
It is not surprising that the process of change detection is fundamental to many machine vision applications. Most change detection algorithms assume that the illumination on a scene will remain constant. Unfortunately, this assumption is not necessarily valid outside a well-controlled laboratory setting. The accuracy of existing algorithms decreases significantly when faced up with image sequences in which the illumination is allowed to vary. In this paper an unsupervised change detection algorithm has been proposed, it performs the change detection without any additional information besides the raw images considered. It performs well under time-varying illumination conditions, where other algorithms fail to perform.
In this paper, a method of passive steganalysis is proposed. We focus on detecting the existing of data hidden in audio files with spread spectrum (SS) data hiding. SS data hiding is considered as a process of adding ...
详细信息
In this paper, a method of passive steganalysis is proposed. We focus on detecting the existing of data hidden in audio files with spread spectrum (SS) data hiding. SS data hiding is considered as a process of adding noise. The technology of classifier and feature vector extraction are used to achieve the detection. First, we divide an audio signal into several frames. The wavelet coefficients before and after wavelet de-noise in each frame are calculated. Then, we pick some stat, of their difference as the feature vectors of the audio signal. Finally, according to the feature vectors of the audio signal, classifier will decide whether the audio signal have been processed by SS or not. In our experiment, support vector machines (SVM) play role of classifier, 600 audio files are used to be our experiment samples. After the feature vectors of all the samples are calculated, those feature vectors of samples are divided into two parts. One is testing part and the other is training part. The result of experiment shows that if the strength of data hiding is higher than 0.005, the rate of correct detection of training part is higher than 86.5% and the rate of correct detection of testing part is higher than 82.5%.
This paper investigates the effects of channel estimation errors on zero-forcing (ZF) vertical Bell laboratories layered space time (V-BLAST) detection. An analytical method is presented to derive the symbol error pro...
详细信息
This paper investigates the effects of channel estimation errors on zero-forcing (ZF) vertical Bell laboratories layered space time (V-BLAST) detection. An analytical method is presented to derive the symbol error probability (SEP) of the signals detected at each stage. The effects of imperfect channel estimation on the SEP performance of V-BLAST detection are then studied. It is shown that ZF-VBLAST detection is very sensitive to the channel estimation errors under high signal to noise ratio (SNR). It is also shown that when optimal ordering is adopted, the effects of channel estimation errors are more significant on the latter detection stages.
The rapid and accurate detection of anomalies in network traffic has always been a challenging task, and is absolutely critical to the efficient operation of the network. The availability of numerous different detecti...
详细信息
The rapid and accurate detection of anomalies in network traffic has always been a challenging task, and is absolutely critical to the efficient operation of the network. The availability of numerous different detection algorithms makes it difficult to choose a suitable configuration. An algorithm may have a high detection rate for high rate attacks, but might behave unfavorably when faced with attacks with gradually increasing rates. This paper proposes an online parallel anomaly detection system that implements multiple anomaly detection algorithms in parallel to detect anomalies in real-time. The main idea is to aggregate the detection data from multiple algorithms to come up with a single anomaly metric. We evaluate this system with realistic attacks on the DETER testbed. Our results show improved true positive and false negative rates for both high intensity and slow-rise ramped floods. Furthermore, the system is able to detect attacks separated by as little as 15 seconds with a high true positive rate.
暂无评论