Biblio

The Internet of Things (IoT) and mobile systems nowadays are required to perform more intensive computation, such as facial detection, image recognition and even remote gaming, etc. Due to the limited computation performance and power budget, it is sometimes impossible to perform these workloads locally. As high-performance GPUs become more common in the cloud, offloading the computation to the cloud becomes a possible choice. However, due to the fact that offloaded workloads from different devices (belonging to different users) are being computed in the same cloud, security concerns arise. Side channel attacks on GPU systems have been widely studied, where the threat model is the attacker and the victim are running on the same operating system. Recently, major GPU vendors have provided hardware and library support to virtualize GPUs for better isolation among users. This work studies the side channel attacks from one virtual machine to another where both share the same physical GPU. We show that it is possible to infer other user's activities in this setup and can further steal others deep learning model.

Convolutional Neural Networks (CNN) have shown astonishing results in the field of computer vision. This paper proposes a foreground segmentation algorithm based on CNN to tackle the practical challenges in the video surveillance system such as illumination changes, dynamic backgrounds, camouflage, and static foreground object, etc. The network is trained using the input of image sequences with respective ground-truth. The algorithm employs a CNN called VGG-16 to extract features from the input. The extracted feature maps are upsampled using a bilinear interpolation. The upsampled feature mask is passed through a sigmoid function and threshold to get the foreground mask. Binary cross entropy is used as the error function to compare the constructed foreground mask with the ground truth. The proposed algorithm was tested on two standard datasets and showed superior performance as compared to the top-ranked foreground segmentation methods.

Face detection is an application widely used in video surveillance systems and it is the first step for subsequent applications such as monitoring and recognition. For facial detection, there are a series of algorithms that allow the face to be extracted in a video image, among which are the Viola & Jones waterfall method and the method by geometric models using the Hausdorff distance. In this article, both algorithms are theoretically analyzed and the best one is determined by efficiency and resource optimization. Considering the most common problems in the detection of faces in a video surveillance system, such as the conditions of brightness and the angle of rotation of the face, tests have been carried out in 13 different scenarios with the best theoretically analyzed algorithm and its combination with another algorithm The images obtained, using a digital camera in the 13 scenarios, have been analyzed using Matlab code of the Viola & Jones and Viola & Jones algorithm combined with the Kanade-Lucas-Tomasi algorithm to add the feature of completing the tracking of a single object. This paper presents the detection percentages, false positives and false negatives for each image and for each simulation code, resulting in the scenarios with the most detection problems and the most accurate algorithm in face detection.

The usage of small drones/UAVs has significantly increased recently. Consequently, there is a rising potential of small drones being misused for illegal activities such as terrorism, smuggling of drugs, etc. posing high-security risks. Hence, tracking and surveillance of drones are essential to prevent security breaches. The similarity in the appearance of small drone and birds in complex background makes it challenging to detect drones in surveillance videos. This paper addresses the challenge of detecting small drones in surveillance videos using popular and advanced deep learning-based object detection methods. Different CNN-based architectures such as ResNet-101 and Inception with Faster-RCNN, as well as Single Shot Detector (SSD) model was used for experiments. Due to sparse data available for experiments, pre-trained models were used while training the CNNs using transfer learning. Best results were obtained from experiments using Faster-RCNN with the base architecture of ResNet-101. Experimental analysis on different CNN architectures is presented in the paper, along with the visual analysis of the test dataset.

Internet of Things (IoT) as per estimated will connect 50 billion devices by 2020. Since its evolution, IoT technology provides lots of flexibility to develop and implement any application. Most of the application improves the human living standard and also makes life easy to access and monitoring the things in real time. Though there exist some security and privacy issues in IoT system like authentication, computation, data modification, trust among users. In this paper, we have identified the IoT application like insurance, supply chain system, smart city and smart car where trust among associated users is an major issue. The current centralized system does not provide enough trust between users. Using Blockchain technology we have shown that trust issue among users can be managed in a decentralized way so that information can be traceable and identify/verify any time. Blockchain has properties like distributed, digitally share and immutable which enhance security. For Blockchain implementation, Ethereum platform is used.

This paper describes the goals of the ATMOSPHERE project, which is a multi-institutional research and development (R&D) effort aiming at designing and implementing a framework and platform to develop, build, deploy, measure and evolve trustworthy, cloud-enabled applications. The proposed system addresses the federation of geographically distributed cloud computing providers that rely on lightweight virtualization, and provide access to heterogeneous sets of resources. In addition, the system also considers both classic trustworthiness properties from the systems community, such as dependability and security, and from the machine learning community, such as fairness and transparency. We present the architecture that has been proposed to address these challenges and discuss some preliminary results.

Intelligent software applications are becoming ubiquitous and pervasive affecting various aspects of our lives and livelihoods. At the same time, the risks to which these systems expose the organizations and end users are growing dramatically. Trustworthiness of software applications is becoming a paramount necessity. Trust is to be regarded as a first-class citizen in the total product life cycle and should be addressed across all stages of software development. Trust can be looked at from two facets: one at an algorithmic level (e.g., bias-free, discrimination-aware, explainable and interpretable techniques) and the other at a process level by making development processes more transparent, auditable, and adhering to regulations and best practices. In this paper, we address the latter and propose a blockchain enabled governance framework for building trustworthy software. Our framework supports the recording, monitoring, and analysis of various activities throughout the application development life cycle thereby bringing in transparency and auditability. It facilitates the specification of regulations and best practices and verifies for its adherence raising alerts of non-compliance and prescribes remedial measures.

In this paper we propose ROS-Defender, a holistic approach to secure robotics systems, which integrates a Security Event Management System (SIEM), an intrusion prevention system (IPS) and a firewall for a robotic system. ROS-Defender combines anomaly detection systems at application (ROS) level and network level, with dynamic policy enforcement points using software defined networking (SDN) to provide protection against a large class of attacks. Although SIEMs, IPS, and firewall have been previously used to secure computer networks, ROSDefender is applying them for the specific use case of robotic systems, where security is in many cases an afterthought.

With WhatsApp's adoption of the Signal Protocol as its default, end-to-end encryption by the masses happened almost overnight. Unlike iMessage, WhatsApp notifies users that encryption is enabled, explicitly informing users about improved privacy. This rare feature gives us an opportunity to study people's understandings and perceptions of secure messaging pre-and post-mass messenger encryption (pre/post-MME). To study changes in perceptions, we compared the results of two mental models studies: one conducted in 2015 pre-MME and one in 2017 post-MME. Our primary finding is that users do not trust encryption as currently offered. When asked about encryption in the study, most stated that they had heard of encryption, but only a few understood the implications, even on a high level. Their consensus view was that no technical solution to stop skilled attackers from getting their data exists. Even with a major development, such as WhatsApp rolling out end-to-end encryption, people still do not feel well protected by their technology. Surprisingly, despite WhatsApp's end-to-end security info messages and the high media attention, the majority of the participants were not even aware of encryption. Most participants had an almost correct threat model, but don't believe that there is a technical solution to stop knowledgeable attackers to read their messages. Using technology made them feel vulnerable.

Industry 4.0 or the fourth industrial revolution encapsulates future industry development trends to achieve more intelligent manufacturing processes, including reliance on Cyber Physical Systems (CPS). The increase in online access and control given by the incorporation of CPSs introduces a new challenge securing the operations of the CPS in that they are not supported by standard security protocols. This paper describes a process used to effectively protect the operations of an IIoT system by implementing security protocols on the CPS within the IIoT. A series of predefined boundary conditions of the safety critical parameters for which a heating and cooling CPS can safely operate within were established. If the CPS is commended to operate outside of these boundaries, it will disconnect from all external communication network and default to some pre-defined safe-operation mode until the system has been evaluated locally by an administrator and released from the safe-mode. This method was tested and validated by establishing a sample IIoT and CPS testbed setup which monitor and control the temperature of a target environment. An attack was initiated to force the target environment outside of the determined safety-critical parameters. The system responded by disabling all network ports and defaulted to the safe-operation mode established previously.

The integration of Internet-of-Things and pervasive computing in medical devices have made the modern healthcare system “smart.” Today, the function of the healthcare system is not limited to treat the patients only. With the help of implantable medical devices and wearables, Smart Healthcare System (SHS) can continuously monitor different vital signs of a patient and automatically detect and prevent critical medical conditions. However, these increasing functionalities of SHS raise several security concerns and attackers can exploit the SHS in numerous ways: they can impede normal function of the SHS, inject false data to change vital signs, and tamper a medical device to change the outcome of a medical emergency. In this paper, we propose HealthGuard, a novel machine learning-based security framework to detect malicious activities in a SHS. HealthGuard observes the vital signs of different connected devices of a SHS and correlates the vitals to understand the changes in body functions of the patient to distinguish benign and malicious activities. HealthGuard utilizes four different machine learning-based detection techniques (Artificial Neural Network, Decision Tree, Random Forest, k-Nearest Neighbor) to detect malicious activities in a SHS. We trained HealthGuard with data collected for eight different smart medical devices for twelve benign events including seven normal user activities and five disease-affected events. Furthermore, we evaluated the performance of HealthGuard against three different malicious threats. Our extensive evaluation shows that HealthGuard is an effective security framework for SHS with an accuracy of 91 % and an F1 score of 90 %.

Nowadays, phishing is one of the most usual web threats with regards to the significant growth of the World Wide Web in volume over time. Phishing attackers always use new (zero-day) and sophisticated techniques to deceive online customers. Hence, it is necessary that the anti-phishing system be real-time and fast and also leverages from an intelligent phishing detection solution. Here, we develop a reliable detection system which can adaptively match the changing environment and phishing websites. Our method is an online and feature-rich machine learning technique to discriminate the phishing and legitimate websites. Since the proposed approach extracts different types of discriminative features from URLs and webpages source code, it is an entirely client-side solution and does not require any service from the third-party. The experimental results highlight the robustness and competitiveness of our anti-phishing system to distinguish the phishing and legitimate websites.

In this paper, a framework for capitalizing on the potential benefits of physical layer security in an amplify-and-forward cooperative spectrum sensing (AF-CSS) in a cognitive radio network (CRN) using a stochastic geometry is proposed. In the CRN network the sensing data from secondary users (SUs) are collected by a fusion center (FC) with the help of access points (AP) as relays, and when malicious eavesdropping secondary users (SUs) are listening. We focus on the secure transmission of active SUs transmitting their sensing data to the AP. Closed expressions for the average secrecy rate are presented. Numerical results corroborate our analysis and show that multiple antennas at the APs can enhance the security of the AF-CSS-CRN. The obtained numerical results show that average secrecy rate between the AP and its correlated FC decreases when the number of AP is increased. Nevertheless, we find that an increase in the number of AP initially increases the overall average secrecy rate, with a perilous value at which the overall average secrecy rate then decreases. While increasing the number of active SUs, there is a decrease in the secrecy rate between the sensor and its correlated AP.

Optical CDMA (OCMDA) technology directly encrypts optical transmission links at the physical layer, which can improve the security of communication system against fibre-optic eavesdropping attacks. System parameters will affect the performances of OCDMA systems, based on the wiretap channel model of OCDMA systems, "secrecy capacity" is employed as an indicator to estimate the effects of system parameters (the type of code words, the length of code words) on the security of the systems. Simulation results demonstrate that system parameters play an important role and choosing the code words with better cross-correlation characteristics can improve the security of OCDMA systems.

Wireless communication network (WCN) performance is primarily depends on physical layer security which is critical among all other layers of OSI network model. It is typically prone to anomaly/malicious user's attacks owing to openness of wireless channels. Cognitive radio networking (CRN) is a recently emerged wireless technology that is having numerous security challenges because of its unlicensed access of wireless channels. In CRNs, the security issues occur mainly during spectrum sensing and is more pronounced during distributed spectrum sensing. In recent past, various anomaly effects are modelled and developed detectors by applying advanced statistical techniques. Nevertheless, many of these detectors have been developed based on sensing data of one variable (energy measurement) and degrades their performance drastically when the data is contaminated with multiple anomaly nodes, that attack the network cooperatively. Hence, one has to develop an efficient multiple anomaly detection algorithm to eliminate all possible cooperative attacks. To achieve this, in this work, the impact of anomaly on detection probability is verified beforehand in developing an efficient algorithm using bivariate data to detect possible attacks with mahalanobis distance measure. Result discloses that detection error of cooperative attacks by anomaly has significant impact on eigenvalue-based sensing.

In this work we present a novel technique for physical layer security in the Internet-of-Things (IoT) networks. In the proposed architecture, each IoT node generates a phase-modulated random key/data and transmits it to a master node in the presence of an eavesdropper, referred to as Eve. The master node, simultaneously, broadcasts a high power signal using an omni-directional antenna, which is received as interference by Eve. This interference masks the generated key by the IoT node and will result in a higher bit-error rate in the data received by Eve. The two legitimate intended nodes communicate in a full-duplex manner and, consequently, subtract their transmitted signals, as a known reference, from the received signal (self-interference cancellation). We compare our proposed method with a conventional approach to physical layer security based on directional antennas. In particular, we show, using theoretical and measurement results, that our proposed approach provides significantly better security measures, in terms bit error rate (BER) at Eve's location. Also, it is proven that in our novel system, the possible eavesdropping region, defined by the region with BER \textbackslashtextless; 10-1, is always smaller than the reliable communication region with BER \textbackslashtextless; 10-3.

Failure to detect malware at its very inception leaves room for it to post significant threat and cost to cyber security for not only individuals, organizations but also the society and nation. However, the rapid growth in volume and diversity of malware renders conventional detection techniques that utilize feature extraction and comparison insufficient, making it very difficult for well-trained network administrators to identify malware, not to mention regular users of internet. Challenges in malware detection is exacerbated since complexity in the type and structure also increase dramatically in these years to include source code, binary file, shell script, Perl script, instructions, settings and others. Such increased complexity offers a premium on misjudgment. In order to increase malware detection efficiency and accuracy under large volume and multiple types of malware, this research adopts Convolutional Neural Networks (CNN), one of the most successful deep learning techniques. The experiment shows an accuracy rate of over 90% in identifying malicious and benign codes. The experiment also presents that CNN is effective with detecting source code and binary code, it can further identify malware that is embedded into benign code, leaving malware no place to hide. This research proposes a feasible solution for network administrators to efficiently identify malware at the very inception in the severe network environment nowadays, so that information technology personnel can take protective actions in a timely manner and make preparations for potential follow-up cyber-attacks.

This paper introduces complex network into software clone detection and proposes a clone code detection method based on software complex network feature matching. This method has the following properties. It builds a software network model with many added features and codes written with different languages can be detected by a single method. It reduces the space of code comparison, and it searches similar subnetworks to detect clones without knowing any clone codes information. This method can be used in detecting open source code which has been reused in software for security analysis.

The network coding optimization based on niche genetic algorithm can observably reduce the network overhead of encoding technology, however, security issues haven't been considered in the coding operation. In order to solve this problem, we propose a network coding optimization scheme for niche algorithm based on security performance (SNGA). It is on the basis of multi-target niche genetic algorithm(NGA)to construct a fitness function which with k-secure network coding mechanism, and to ensure the realization of information security and achieve the maximum transmission of the network. The simulation results show that SNGA can effectively improve the security of network coding, and ensure the running time and convergence speed of the optimal solution.

Currently, the most commonly used scheme for identity authentication on the Internet is based on asymmetric cryptography and the use of a centralized model. The centralized model needs a Certificate Authority (CA) as a trusted third party and a trust chain of CA. However, CA-based PKI is weak in the single point of failure and certificate transparency. Our system, called SS-DPKI, propose a public and decentralized PKI system model. We describe a detailed scheme as well as application to use decentralized PKI based secure communication. Our proposal prevents storage overhead on the data size of transactions and provide reasonable certificate verification time.

Public key infrastructures (PKIs) are one of the main building blocks for securing communications over the Internet. Currently, PKIs are under the control of centralized authorities, which is problematic as evidenced by numerous incidents where they have been compromised. The distributed, fault tolerant log of transactions provided by blockchains and more recently, smart contract platforms, constitutes a powerful tool for the decentralization of PKIs. To verify the validity of identity records, blockchain-based identity systems store on chain either all identity records, or, a small (or even constant) sized amount of data for verifying identity records stored off chain. However, as most of these systems have never been implemented, there is little information regarding the practical implications of each design's tradeoffs. In this work, we first implement and evaluate the only provably secure, smart contract based PKI of Patsonakis et al. on top of Ethereum. This construction incurs constant-sized storage at the expense of computational complexity. To explore this tradeoff, we propose and implement a second construction which, eliminates the need for trusted setup, preserves the security properties of Patsonakis et al. and, as illustrated through our evaluation, is the only version with constant-sized state that can be deployed on the live chain of Ethereum. Furthermore, we compare these two systems with the simple approach of most prior works, e.g., the Ethereum Name Service, where all identity records are stored on the smart contract's state, to illustrate several shortcomings of Ethereum and its cost model. We propose several modifications for fine tuning the model, which would be useful to be considered for any smart contract platform like Ethereum so that it reaches its full potential to support arbitrary distributed applications.

Due to the diversity and mobility of blockchain network nodes and the decentralized nature of blockchain networks, traditional trust value evaluation indicators cannot be directly used. In order to obtain trusted nodes, a trustworthiness calculation method based on trust blockchain nodes is proposed. Different from the traditional P2P network trust value calculation, the trust blockchain not only acquires the working state of the node, but also collects the special behavior information of the node, and calculates the joining time by synthesizing the trust value generated by the node transaction and the trust value generated by the node behavior. After the attenuation factor is comprehensively evaluated, the trusted nodes are selected to effectively ensure the security of the blockchain network environment, while reducing the average transaction delay and increasing the block rate.

Logic encryption is a powerful hardware protection technique that uses extra key inputs to lock a circuit from piracy or unauthorized use. The recent discovery of the SAT-based attack with Distinguishing Input Pattern (DIP) generation has rendered all traditional logic encryptions vulnerable, and thus the creation of new encryption methods. However, a critical question for any new encryption method is whether security against the DIP-generation attack means security against all other attacks. In this paper, a new high-level SAT-based attack called SigAttack has been discovered and thoroughly investigated. It is based on extracting a key-revealing signature in the encryption. A majority of all known SAT-resilient encryptions are shown to be vulnerable to SigAttack. By formulating the condition under which SigAttack is effective, the paper also provides guidance for the future logic encryption design.

Today, Smartphone is a necessary device for people connected to the Internet world. But user privacy and security are still playing important roles in the usage of mobile devices. The user was asked to enter related characters, numbers or drawing a simple graphic on the touch screen as passwords for unlocking the screensaver. Although it could provide the user with a simple and convenient security authentication mechanism, the process is hard to protect against the privacy information leakage under the strict security policy. Nowadays, various keypad lock screen Apps usually provides different type of schemes in unlocking the mobile device screen, such as simple-customized pattern, swipe-to-unlock with a static image and so on. But the vulnerability could provide a chance to hijacker to find out the leakage of graphic pattern information that influences in user information privacy and security.This paper proposes a new graphic pattern authentication mechanism to enhance the strength of that in the keypad lock screen Apps. It integrates random digital graphics and handwriting graphic input track recognition technologies to provide better and more diverse privacy protection and reduce the risk of vulnerability. The proposed mechanism is based on two factor identification scheme. First of all, it randomly changes digital graphic position based on unique passwords every time to increase the difficulty of the stealer's recording. Second, the input track of handwriting graphics is another identification factor for enhancing the complex strength of user authentication as well.

To manage and handle the issues of physical security in the modern world, there is a dire need for a multilevel security system to ensure the safety of precious belongings that could be money, military equipment or medical life-saving drugs. Security locker solution is proposed which is a multiple layer security system consisting of various levels of authentication. In most cases, only relevant persons should have access to their precious belongings. The unlocking of the box is only possible when all of the security levels are successfully cleared. The five levels of security include entering of password on interactive GUI, thumbprint, facial recognition, speech pattern recognition, and vein pattern recognition. This project is unique and effective in a sense that it incorporates five levels of security in a single prototype with the use of cost-effective equipment. Assessing our security system, it is seen that security is increased many a fold as it is near to impossible to breach all these five levels of security. The Raspberry Pi microcomputers, handling all the traits efficiently and smartly makes it easy for performing all the verification tasks. The traits used involves checking, training and verifying processes with application of machine learning operations.