RESUMO
Malicious social bots pose a serious threat to social network security by spreading false information and guiding bad opinions in social networks. The singularity and scarcity of single organization data and the high cost of labeling social bots have given rise to the construction of federated models that combine federated learning with social bot detection. In this paper, we first combine the federated learning framework with the Relational Graph Convolutional Neural Network (RGCN) model to achieve federated social bot detection. A class-level cross entropy loss function is applied in the local model training to mitigate the effects of the class imbalance problem in local data. To address the data heterogeneity issue from multiple participants, we optimize the classical federated learning algorithm by applying knowledge distillation methods. Specifically, we adjust the client-side and server-side models separately: training a global generator to generate pseudo-samples based on the local data distribution knowledge to correct the optimization direction of client-side classification models, and integrating client-side classification models' knowledge on the server side to guide the training of the global classification model. We conduct extensive experiments on widely used datasets, and the results demonstrate the effectiveness of our approach in social bot detection in heterogeneous data scenarios. Compared to baseline methods, our approach achieves a nearly 3-10% improvement in detection accuracy when the data heterogeneity is larger. Additionally, our method achieves the specified accuracy with minimal communication rounds.
RESUMO
In order to improve user authentication accuracy based on keystroke dynamics and mouse dynamics in hybrid scenes and to consider the user operation changes in different scenes that aggravate user status changes and make it difficult to simulate user behaviors, we present a user authentication method entitled SIURUA. SIURUA uses scene-irrelated features and user-related features for user identification. First, features are extracted based on keystroke data and mouse movement data. Next, scene-irrelated features that have a low correlation with scenes are obtained. Finally, scene-irrelated features are fused with user-related features to ensure the integrity of the features. Experimental results show that the proposed method has the advantage of improving user authentication accuracy in hybrid scenes, with an accuracy of 84% obtained in the experiment.
Assuntos
Segurança Computacional , Computadores , Movimento , HumanosRESUMO
Phishing has become one of the biggest and most effective cyber threats, causing hundreds of millions of dollars in losses and millions of data breaches every year. Currently, anti-phishing techniques require experts to extract phishing sites features and use third-party services to detect phishing sites. These techniques have some limitations, one of which is that extracting phishing features requires expertise and is time-consuming. Second, the use of third-party services delays the detection of phishing sites. Hence, this paper proposes an integrated phishing website detection method based on convolutional neural networks (CNN) and random forest (RF). The method can predict the legitimacy of URLs without accessing the web content or using third-party services. The proposed technique uses character embedding techniques to convert URLs into fixed-size matrices, extract features at different levels using CNN models, classify multi-level features using multiple RF classifiers, and, finally, output prediction results using a winner-take-all approach. On our dataset, a 99.35% accuracy rate was achieved using the proposed model. An accuracy rate of 99.26% was achieved on the benchmark data, much higher than that of the existing extreme model.
Assuntos
Aprendizagem , Redes Neurais de Computação , Aprendizado de MáquinaRESUMO
Based on the openness and accessibility of user data, personality recognition is widely used in personalized recommendation, intelligent medicine, natural language processing, and so on. Existing approaches usually adopt a single deep learning mechanism to extract personality information from user data, which leads to semantic loss to some extent. In addition, researchers encode scattered user posts in a sequential or hierarchical manner, ignoring the connection between posts and the unequal value of different posts to classification tasks. We propose a hierarchical hybrid model based on a self-attention mechanism, namely HMAttn-ECBiL, to fully excavate deep semantic information horizontally and vertically. Multiple modules composed of convolutional neural network and bi-directional long short-term memory encode different types of personality representations in a hierarchical and partitioned manner, which pays attention to the contribution of different words in posts and different posts to personality information and captures the dependencies between scattered posts. Moreover, the addition of a word embedding module effectively makes up for the original semantics filtered by a deep neural network. We verified the hybrid model on the MyPersonality dataset. The experimental results showed that the classification performance of the hybrid model exceeds the different model architectures and baseline models, and the average accuracy reached 72.01%.
Assuntos
Processamento de Linguagem Natural , Redes Neurais de Computação , Personalidade , SemânticaRESUMO
Cross-domain relation extraction has become an essential approach when target domain lacking labeled data. Most existing works adapted relation extraction models from the source domain to target domain through aligning sequential features, but failed to transfer non-local and non-sequential features such as word co-occurrence which are also critical for cross-domain relation extraction. To address this issue, in this paper, we propose a novel tripartite graph architecture to adapt non-local features when there is no labeled data in the target domain. The graph uses domain words as nodes to model the co-occurrence relation between domain-specific words and domain-independent words. Through graph convolutions on the tripartite graph, the information of domain-specific words is propagated so that the word representation can be fine-tuned to align domain-specific features. In addition, unlike the traditional graph structure, the weights of edges innovatively combine fixed weight and dynamic weight, to capture the global non-local features and avoid introducing noise to word representation. Experiments on three domains of ACE2005 datasets show that our method outperforms the state-of-the-art models by a big margin.
RESUMO
Intrusion detection systems play an important role in preventing security threats and protecting networks from attacks. However, with the emergence of unknown attacks and imbalanced samples, traditional machine learning methods suffer from lower detection rates and higher false positive rates. We propose a novel intrusion detection model that combines an improved conditional variational AutoEncoder (ICVAE) with a deep neural network (DNN), namely ICVAE-DNN. ICVAE is used to learn and explore potential sparse representations between network data features and classes. The trained ICVAE decoder generates new attack samples according to the specified intrusion categories to balance the training data and increase the diversity of training samples, thereby improving the detection rate of the imbalanced attacks. The trained ICVAE encoder is not only used to automatically reduce data dimension, but also to initialize the weight of DNN hidden layers, so that DNN can easily achieve global optimization through back propagation and fine tuning. The NSL-KDD and UNSW-NB15 datasets are used to evaluate the performance of the ICVAE-DNN. The ICVAE-DNN is superior to the three well-known oversampling methods in data augmentation. Moreover, the ICVAE-DNN outperforms six well-known models in detection performance, and is more effective in detecting minority attacks and unknown attacks. In addition, the ICVAE-DNN also shows better overall accuracy, detection rate and false positive rate than the nine state-of-the-art intrusion detection methods.
RESUMO
Deep Learning algorithms have achieved state-of-the-art performance in various important tasks. However, recent studies have found that an elaborate perturbation may cause a network to misclassify, which is known as an adversarial attack. Based on current research, it is suggested that adversarial examples cannot be eliminated completely. Consequently, it is always possible to determine an attack that is effective against a defense model. We render existing adversarial examples invalid by altering the classification boundaries. Meanwhile, for valid adversarial examples generated against the defense model, the adversarial perturbations are increased so that they can be distinguished by the human eye. This paper proposes a method for implementing the abovementioned concepts through color space transformation. Experiments on CIFAR-10, CIFAR-100, and Mini-ImageNet demonstrate the effectiveness and versatility of our defense method. To the best of our knowledge, this is the first defense model based on the amplification of adversarial perturbations.
Assuntos
Algoritmos , Conhecimento , HumanosRESUMO
We present multiple-image encryption (MIE) based on compressive holography. In the encryption, a holographic technique is employed to record multiple images simultaneously to form a hologram. The two-dimensional Fourier data of the hologram are then compressed by nonuniform sampling, which gives rise to compressive encryption. Decryption of individual images is cast into a minimization problem. The minimization retains the sparsity of recovered images in the wavelet basis. Meanwhile, total variation regularization is used to preserve edges in the reconstruction. Experiments have been conducted using holograms acquired by optical scanning holography as an example. Computer simulations of multiple images are subsequently demonstrated to illustrate the feasibility of the MIE scheme.
RESUMO
Obscuring or otherwise minimizing the release of personality information from potential victims of social engineering attacks effectively interferes with an attacker's personality analysis and reduces the success rate of social engineering attacks. We propose a text transformation method named PerTransGAN using generative adversarial networks (GANs) to protect the personality privacy hidden in text data. Making use of reinforcement learning, we use the output of the discriminator as a reward signal to guide the training of the generator. Moreover, the model extracts text features from the discriminator network as additional semantic guidance signals. And the loss function of the generator adds a penalty item to reduce the weight of words that contribute more to personality information in the real text so as to hide the user's personality privacy. In addition, the semantic and personality modules are designed to calculate the semantic similarity and personality distribution distance between the real text and the generated text as a part of the objective function. Experiments show that the self-attention module and semantic module in the generator improved the content retention of the text by 0.11 compared with the baseline model and obtained the highest BLEU score. In addition, with the addition of penalty item and personality module, compared with the classification accuracy of the original data, the accuracy of the generated text in the personality classifier decreased by 20%. PerTransGAN model preserves users' personality privacy as found in user data by transforming the text and preserving semantic similarity while blocking privacy theft by attackers.
Assuntos
Redes Neurais de Computação , Privacidade , Personalidade , SemânticaRESUMO
While antiphishing techniques have evolved over the years, phishing remains one of the most threatening attacks on current network security. This is because phishing exploits one of the weakest links in a network system-people. The purpose of this research is to predict the possible phishing victims. In this study, we propose the multidimensional phishing susceptibility prediction model (MPSPM) to implement the prediction of user phishing susceptibility. We constructed two types of emails: legitimate emails and phishing emails. We gathered 1105 volunteers to join our experiment by recruiting volunteers. We sent these emails to volunteers and collected their demographic, personality, knowledge experience, security behavior, and cognitive processes by means of a questionnaire. We then applied 7 supervised learning methods to classify these volunteers into two categories using multidimensional features: susceptible and nonsusceptible. The experimental results indicated that some machine learning methods have high accuracy in predicting user phishing susceptibility, with a maximum accuracy rate of 89.04%. We conclude our study with a discussion of our findings and their future implications.