Biblio

Fake news, frequently making use of tampered photos, has currently emerged as a global epidemic, mainly due to the widespread use of social media as a present alternative to traditional news outlets. This development is often due to the swiftly declining price of advanced cameras and phones, which prompts the simple making of computerized pictures. The accessibility and usability of picture-altering softwares make picture-altering or controlling processes significantly simple, regardless of whether it is for the blameless or malicious plan. Various investigations have been utilized around to distinguish this sort of controlled media to deal with this issue. This paper proposes an efficient technique of copy-move forgery detection using the deep learning method. Two deep learning models such as Buster Net and VGG with FPN are used here to detect copy move forgery in digital images. The two models' performance is evaluated using the CoMoFoD dataset. The experimental result shows that VGG with FPN outperforms the Buster Net model for detecting forgery in images with an accuracy of 99.8% whereas the accuracy for the Buster Net model is 96.9%.

The Software Defined Network (SDN) provides higher programmable functionality for network configuration and management dynamically. Moreover, SDN introduces a centralized management approach by dividing the network into control and data planes. In this paper, we introduce a deep learning enabled intrusion detection and prevention system (DL-IDPS) to prevent secure shell (SSH) brute-force attacks and distributed denial-of-service (DDoS) attacks in SDN. The packet length in SDN switch has been collected as a sequence for deep learning models to identify anomalous and malicious packets. Four deep learning models, including Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) and Stacked Auto-encoder (SAE), are implemented and compared for the proposed DL-IDPS. The experimental results show that the proposed MLP based DL-IDPS has the highest accuracy which can achieve nearly 99% and 100% accuracy to prevent SSH Brute-force and DDoS attacks, respectively.

The prevalence of photo processing apps suggests the demands of picture editing. As an implementation of the convolutional neural network, style transfer has been deep investigated and there are supported materials to realize it on PC platform. However, few approaches are mentioned to deploy a style transfer model on the mobile and meet the requirements of mobile users. The traditional style transfer model takes hours to proceed, therefore, based on a Perceptual Losses algorithm [1], we created a feedforward neural network for each style and the proceeding time was reduced to a few seconds. The training data were generated from a pre-trained convolutional neural network model, VGG-19. The algorithm took thousandth time and generated similar output as the original. Furthermore, we optimized the model and deployed the model with TensorFlow Mobile library. We froze the model and adopted a bitmap to scale the inputs to 720×720 and reverted back to the original resolution. The reverting process may create some blur but it can be regarded as a feature of art. The generated images have reliable quality and the waiting time is independent of the content and pattern of input images. The main factor that influences the proceeding time is the input resolution. The average waiting time of our model on the mobile phone, HUAWEI P20 Pro, is less than 2 seconds for 720p images and around 2.8 seconds for 1080p images, which are ten times slower than that on the PC GPU, Tesla T40. The performance difference depends on the architecture of the model.

Recent advances in technology have made the deep learning (DL) models available for use in a wide variety of novel applications; for example, generative adversarial network (GAN) models are capable of producing hyper-realistic images, speech, and even videos, such as the so-called “Deepfake” produced by GANs with manipulated audio and/or video clips, which are so realistic as to be indistinguishable from the real ones in human perception. Aside from innovative and legitimate applications, there are numerous nefarious or unlawful ways to use such counterfeit contents in propaganda, political campaigns, cybercrimes, extortion, etc. To meet the challenges posed by Deepfake multimedia, we propose a deep ensemble learning technique called DeepfakeStack for detecting such manipulated videos. The proposed technique combines a series of DL based state-of-art classification models and creates an improved composite classifier. Based on our experiments, it is shown that DeepfakeStack outperforms other classifiers by achieving an accuracy of 99.65% and AUROC of 1.0 score in detecting Deepfake. Therefore, our method provides a solid basis for building a Realtime Deepfake detector.

This paper proposes an approach that combines a deep learning-based method and a traditional machine learning-based method to efficiently detect malicious requests Web servers received. The first few layers of Convolutional Neural Network for Text Classification (TextCNN) are used to automatically extract powerful semantic features and in the meantime transferable statistical features are defined to boost the detection ability, specifically Web request parameter tampering. The semantic features from TextCNN and transferable statistical features from artificially-designing are grouped together to be fed into Support Vector Machine (SVM), replacing the last layer of TextCNN for classification. To facilitate the understanding of abstract features in form of numerical data in vectors extracted by TextCNN, this paper designs trace-back functions that map max-pooling outputs back to words in Web requests. After investigating the current available datasets for Web attack detection, HTTP Dataset CSIC 2010 is selected to test and verify the proposed approach. Compared with other deep learning models, the experimental results demonstrate that the approach proposed in this paper is competitive with the state-of-the-art.

Recent years, more and more testing criteria for deep learning systems has been proposed to ensure system robustness and reliability. These criteria were defined based on different perspectives of diversity. However, there lacks comprehensive investigation on what are the most essential diversities that should be considered by a testing criteria for deep learning systems. Therefore, in this paper, we conduct an empirical study to investigate the relation between test diversities and erroneous behaviors of deep learning models. We define five metrics to reflect diversities in neuron activities, and leverage metamorphic testing to detect erroneous behaviors. We investigate the correlation between metrics and erroneous behaviors. We also go further step to measure the quality of test suites under the guidance of defined metrics. Our results provided comprehensive insights on the essential diversities for testing criteria to exhibit good fault detection ability.

The exponential growth rate of malware causes significant security concern in this digital era to computer users, private and government organizations. Traditional malware detection methods employ static and dynamic analysis, which are ineffective in identifying unknown malware. Malware authors develop new malware by using polymorphic and evasion techniques on existing malware and escape detection. Newly arriving malware are variants of existing malware and their patterns can be analyzed using the vision-based method. Malware patterns are visualized as images and their features are characterized. The alternative generation of class vectors and feature vectors using ensemble forests in multiple sequential layers is performed for classifying malware. This paper proposes a hybrid stacked multilayered ensembling approach which is robust and efficient than deep learning models. The proposed model outperforms the machine learning and deep learning models with an accuracy of 98.91%. The proposed system works well for small-scale and large-scale data since its adaptive nature of setting parameters (number of sequential levels) automatically. It is computationally efficient in terms of resources and time. The method uses very fewer hyper-parameters compared to deep neural networks.

CAPTCHA is a web-based authentication method used by websites to distinguish between humans (valid users) and bots (attackers). Audio captcha is an accessible captcha meant for the visually disabled section of users such as color-blind, blind, near-sighted users. Firstly, this paper analyzes how secure current audio captchas are from attacks using machine learning (ML) and deep learning (DL) models. Each audio captcha is made up of five, seven or ten random digits[0-9] spoken one after the other along with varying background noise throughout the length of the audio. If the ML or DL model is able to correctly identify all spoken digits and in the correct order of occurance in a single audio captcha, we consider that captcha to be broken and the attack to be successful. Throughout the paper, accuracy refers to the attack model's success at breaking audio captchas. The higher the attack accuracy, the more unsecure the audio captchas are. In our baseline experiments, we found that attack models could break audio captchas that had no background noise or medium background noise with any number of spoken digits with nearly 99% to 100% accuracy. Whereas, audio captchas with high background noise were relatively more secure with attack accuracy of 85%. Secondly, we propose that the concepts of adversarial examples algorithms can be used to create a new kind of audio captcha that is more resilient towards attacks. We found that even after retraining the models on the new adversarial audio data, the attack accuracy remained as low as 25% to 36% only. Lastly, we explore the benefits of creating adversarial audio captcha through different algorithms such as Basic Iterative Method (BIM) and deepFool. We found that as long as the attacker has less than 45% sample from each kinds of adversarial audio datasets, the defense will be successful at preventing attacks.

Adversarial attacks can affect the performance of existing deep learning models. With the increased interest in graph based machine learning techniques, there have been investigations which suggest that these models are also vulnerable to attacks. In particular, corruptions of the graph topology can degrade the performance of graph based learning algorithms severely. This is due to the fact that the prediction capability of these algorithms relies mostly on the similarity structure imposed by the graph connectivity. Therefore, detecting the location of the corruption and correcting the induced errors becomes crucial. There has been some recent work which tackles the detection problem, however these methods do not address the effect of the attack on the downstream learning task. In this work, we propose an algorithm that uses node copying to mitigate the degradation in classification that is caused by adversarial attacks. The proposed methodology is applied only after the model for the downstream task is trained and the added computation cost scales well for large graphs. Experimental results show the effectiveness of our approach for several real world datasets.

Training deep neural networks is a computationally expensive task. Furthermore, models are often derived from proprietary datasets that have been carefully prepared and labelled. Hence, creators of deep learning models want to protect their models against intellectual property theft. However, this is not always possible, since the model may, e.g., be embedded in a mobile app for fast response times. As a countermeasure watermarks for deep neural networks have been developed that embed secret information into the model. This information can later be retrieved by the creator to prove ownership. Uchida et al. proposed the first such watermarking method. The advantage of their scheme is that it does not compromise the accuracy of the model prediction. However, in this paper we show that their technique modifies the statistical distribution of the model. Using this modification we can not only detect the presence of a watermark, but even derive its embedding length and use this information to remove the watermark by overwriting it. We show analytically that our detection algorithm follows consequentially from their embedding algorithm and propose a possible countermeasure. Our findings shall help to refine the definition of undetectability of watermarks for deep neural networks.

Recently in the vast advancement of Artificial Intelligence, Machine learning and Deep Neural Network (DNN) driven us to the robust applications. Such as Image processing, speech recognition, and natural language processing, DNN Algorithms has succeeded in many drawbacks; especially the trained DNN models have made easy to the researchers to produces state-of-art results. However, sharing these trained models are always a challenging task, i.e. security, and protection. We performed extensive experiments to present some analysis of watermark in DNN. We proposed a DNN model for Digital watermarking which investigate the intellectual property of Deep Neural Network, Embedding watermarks, and owner verification. This model can generate the watermarks to deal with possible attacks (fine tuning and train to embed). This approach is tested on the standard dataset. Hence this model is robust to above counter-watermark attacks. Our model accurately and instantly verifies the ownership of all the remotely expanded deep learning models without affecting the model accuracy for standard information data.

Deep learning has undergone tremendous advancements in computer vision studies. The training of deep learning neural networks depends on a considerable amount of ground truth datasets. However, labeling ground truth data is a labor-intensive task, particularly for large-volume video analytics applications such as video surveillance and vehicles detection for autonomous driving. This paper presents a rapid and accurate method for associative searching in big image data obtained from security monitoring systems. We developed a semi-automatic moving object annotation method for improving deep learning models. The proposed method comprises three stages, namely automatic foreground object extraction, object annotation in subsequent video frames, and dataset construction using human-in-the-loop quick selection. Furthermore, the proposed method expedites dataset collection and ground truth annotation processes. In contrast to data augmentation and data generative models, the proposed method produces a large amount of real data, which may facilitate training results and avoid adverse effects engendered by artifactual data. We applied the constructed annotation dataset to train a deep learning you-only-look-once (YOLO) model to perform vehicle detection on street intersection surveillance videos. Experimental results demonstrated that the accurate detection performance was improved from a mean average precision (mAP) of 83.99 to 88.03.

In recent years, developments in the field of computer vision have allowed deep learning-based techniques to surpass human-level performance. However, these advances have also culminated in the advent of adversarial machine learning techniques, capable of launching targeted image captioning attacks that easily fool deep learning models. Although attacks in the image domain are well studied, little work has been done in the video domain. In this paper, we show it is possible to extend prior attacks in the image domain to the video captioning task, without heavily affecting the video's playback quality. We demonstrate our attack against a state-of-the-art video captioning model, by extending a prior image captioning attack known as Show and Fool. To the best of our knowledge, this is the first successful method for targeted attacks against a video captioning model, which is able to inject 'subliminal' perturbations into the video stream, and force the model to output a chosen caption with up to 0.981 cosine similarity, achieving near-perfect similarity to chosen target captions.

Deep neural networks are susceptible to various inference attacks as they remember information about their training data. We design white-box inference attacks to perform a comprehensive privacy analysis of deep learning models. We measure the privacy leakage through parameters of fully trained models as well as the parameter updates of models during training. We design inference algorithms for both centralized and federated learning, with respect to passive and active inference attackers, and assuming different adversary prior knowledge. We evaluate our novel white-box membership inference attacks against deep learning algorithms to trace their training data records. We show that a straightforward extension of the known black-box attacks to the white-box setting (through analyzing the outputs of activation functions) is ineffective. We therefore design new algorithms tailored to the white-box setting by exploiting the privacy vulnerabilities of the stochastic gradient descent algorithm, which is the algorithm used to train deep neural networks. We investigate the reasons why deep learning models may leak information about their training data. We then show that even well-generalized models are significantly susceptible to white-box membership inference attacks, by analyzing state-of-the-art pre-trained and publicly available models for the CIFAR dataset. We also show how adversarial participants, in the federated learning setting, can successfully run active membership inference attacks against other participants, even when the global model achieves high prediction accuracies.

With the ever-growing occurrence of networking attacks, robust network security systems are essential to prevent and mitigate their harming effects. In recent years, machine learning-based systems have gain popularity for network security applications, usually considering the application of shallow models, where a set of expert handcrafted features are needed to pre-process the data before training. The main problem with this approach is that handcrafted features can fail to perform well given different kinds of scenarios and problems. Deep Learning models can solve this kind of issues using their ability to learn feature representations from input raw or basic, non-processed data. In this paper we explore the power of deep learning models on the specific problem of detection and classification of malware network traffic, using different representations for the input data. As a major advantage as compared to the state of the art, we consider raw measurements coming directly from the stream of monitored bytes as the input to the proposed models, and evaluate different raw-traffic feature representations, including packet and flow-level ones. Our results suggest that deep learning models can better capture the underlying statistics of malicious traffic as compared to classical, shallow-like models, even while operating in the dark, i.e., without any sort of expert handcrafted inputs.

Facial expression is the most powerful and natural non-verbal emotional communication method. Facial Expression Recognition(FER) has significance in machine learning tasks. Deep Learning models perform well in FER tasks, but it doesn't provide any justification for its decisions. Based on the hypothesis that facial expression is a combination of facial muscle movements, we find that Facial Action Coding Units(AUs) and Emotion label have a relationship in CK+ Dataset. In this paper, we propose a model which utilises AUs to explain Convolutional Neural Network(CNN) model's classification results. The CNN model is trained with CK+ Dataset and classifies emotion based on extracted features. Explanation model classifies the multiple AUs with the extracted features and emotion classes from the CNN model. Our experiment shows that with only features and emotion classes obtained from the CNN model, Explanation model generates AUs very well.

Modern industrial control systems (ICS) act as victims of cyber attacks more often in last years. These attacks are hard to detect and their consequences can be catastrophic. Cyber attacks can cause anomalies in the work of the ICS and its technological equipment. The presence of mutual interference and noises in this equipment significantly complicates anomaly detection. Moreover, the traditional means of protection, which used in corporate solutions, require updating with each change in the structure of the industrial process. An approach based on the machine learning for anomaly detection was used to overcome these problems. It complements traditional methods and allows one to detect signal correlations and use them for anomaly detection. Additional Tennessee Eastman Process Simulation Data for Anomaly Detection Evaluation dataset was analyzed as example of industrial process. In the course of the research, correlations between the signals of the sensors were detected and preliminary data processing was carried out. Algorithms from the most common techniques of machine learning (decision trees, linear algorithms, support vector machines) and deep learning models (neural networks) were investigated for industrial process anomaly detection task. It's shown that linear algorithms are least demanding on computational resources, but they don't achieve an acceptable result and allow a significant number of errors. Decision tree-based algorithms provided an acceptable accuracy, but the amount of RAM, required for their operations, relates polynomially with the training sample volume. The deep neural networks provided the greatest accuracy, but they require considerable computing power for internal calculations.

Malware has become sophisticated and organizations don't have a Plan B when standard lines of defense fail. These failures have devastating consequences for organizations, such as sensitive information being exfiltrated. A promising avenue for improving the effectiveness of behavioral-based malware detectors is to combine fast (usually not highly accurate) traditional machine learning (ML) detectors with high-accuracy, but time-consuming, deep learning (DL) models. The main idea is to place software receiving borderline classifications by traditional ML methods in an environment where uncertainty is added, while software is analyzed by time-consuming DL models. The goal of uncertainty is to rate-limit actions of potential malware during deep analysis. In this paper, we describe Chameleon, a Linux-based framework that implements this uncertain environment. Chameleon offers two environments for its OS processes: standard - for software identified as benign by traditional ML detectors - and uncertain - for software that received borderline classifications analyzed by ML methods. The uncertain environment will bring obstacles to software execution through random perturbations applied probabilistically on selected system calls. We evaluated Chameleon with 113 applications from common benchmarks and 100 malware samples for Linux. Our results show that at threshold 10%, intrusive and non-intrusive strategies caused approximately 65% of malware to fail accomplishing their tasks, while approximately 30% of the analyzed benign software to meet with various levels of disruption (crashed or hampered). We also found that I/O-bound software was three times more affected by uncertainty than CPU-bound software.