Biblio

Deep learning (DL) models have been shown to be vulnerable to adversarial attacks. DL model security against adversarial attacks is critical to using DL-trained models in forward deployed systems, e.g. facial recognition, document characterization, or object detection. We provide results and lessons learned applying a moving target defense (MTD) strategy against iterative, gradient-based adversarial attacks. Our strategy involves (1) training a diverse ensemble of DL models, (2) applying randomized affine input transformations to inputs, and (3) randomizing output decisions. We report a primary lesson that this strategy is ineffective against a white-box adversary, which could completely circumvent output randomization using a deterministic surrogate. We reveal how our ensemble models lacked the diversity necessary for effective MTD. We also evaluate our MTD strategy against a black-box adversary employing an ensemble surrogate model. We conclude that an MTD strategy against black-box adversarial attacks crucially depends on lack of transferability between models.

Memory Denial of Service (DoS) attacks are easy-to-launch, hard to detect, and significantly impact their targets. In memory DoS, the attacker targets the memory of his Virtual Machine (VM) and, due to hardware isolation issues, the attack affects the co-resident VMs. Theoretically, we can deploy VM migration as Moving Target Defense (MTD) against memory DoS. However, the current literature lacks empirical evidence supporting this hypothesis. Moreover, there is a need to evaluate how the VM migration timing impacts the potential MTD protection. This practical experience report presents an experiment on VM migration-based MTD against memory DoS. We evaluate the impact of memory DoS attacks in the context of two applications running in co-hosted VMs: machine learning and OLTP. The results highlight that the memory DoS attacks lead to more than 70% reduction in the applications' performance. Nevertheless, timely VM migrations can significantly mitigate the attack effects in both considered applications.

Space systems provide many critical functions to the military, federal agencies, and infrastructure networks. Nation-state adversaries have shown the ability to disrupt critical infrastructure through cyber-attacks targeting systems of networked, embedded computers. Moving target defenses (MTDs) have been proposed as a means for defending various networks and systems against potential cyber-attacks. MTDs differ from many cyber resilience technologies in that they do not necessarily require detection of an attack to mitigate the threat. We devised a MTD algorithm and tested its application to a real-time network. We demonstrated MTD usage with a real-time protocol given constraints not typically found in best-effort networks. Second, we quantified the cyber resilience benefit of MTD given an exfiltration attack by an adversary. For our experiment, we employed MTD which resulted in a reduction of adversarial knowledge by 97%. Even when the adversary can detect when the address changes, there is still a reduction in adversarial knowledge when compared to static addressing schemes. Furthermore, we analyzed the core performance of the algorithm and characterized its unpredictability using nine different statistical metrics. The characterization highlighted the algorithm has good unpredictability characteristics with some opportunity for improvement to produce more randomness.

This paper proposes a machine-learning-based advanced online dynamic security assessment (DSA) method, which provides a detailed evaluation of the system stability after a disturbance by predicting impending loss of synchronism (LOS) of generators. Voltage angles at generator buses are used as the features of the different random forest (RF) classifiers which are trained to consecutively predict LOS of the generators as a contingency proceeds and updated measurements become available. A wide range of contingencies for various topologies and operating conditions of the IEEE 118-bus system has been studied in offline analysis using the GE positive sequence load flow analysis (PSLF) software to create a comprehensive dataset for training and testing the RF models. The performances of the trained models are evaluated in the presence of measurement errors using various metrics. The results reveal that the trained models are accurate, fast, and robust to measurement errors.

An interesting idea of integrated non-Foster self-oscillating radiating system has been introduced recently. The device consists of two identical antennas, a negative impedance converter (NIC) and a tuning circuit. Admittance of one of the antennas is negatively converted via NIC, and cancelled by the positive admittance of the second identical antenna. With the change of frequency, admittances of both antennas change in the exactly same manner. It makes a self-oscillating and perfectly matched pair of antennas, regardless of the operating frequency. The adjustment of the frequency of a self-oscillating signal is achieved by the additional tunable resonant circuit. This paper analyses dynamics of oscillations of such self-oscillating radiating system and compares it with a classical negative resistance oscillator. Moreover, a simple numerical tool for prediction of the frequency and amplitude of oscillations is proposed.

One of the most concerning threats for modern AI systems is data poisoning, where the attacker injects maliciously crafted training data to corrupt the system's behavior at test time. Availability poisoning is a particularly worrisome subset of poisoning attacks where the attacker aims to cause a Denial-of-Service (DoS) attack. However, the state-of-the-art algorithms are computationally expensive because they try to solve a complex bi-level optimization problem (the ``hammer''). We observed that in particular conditions, namely, where the target model is linear (the ``nut''), the usage of computationally costly procedures can be avoided. We propose a counter-intuitive but efficient heuristic that allows contaminating the training set such that the target system's performance is highly compromised. We further suggest a re-parameterization trick to decrease the number of variables to be optimized. Finally, we demonstrate that, under the considered settings, our framework achieves comparable, or even better, performances in terms of the attacker's objective while being significantly more computationally efficient.

One of the many methods for identifying malware is to disassemble the malware files and obtain the opcodes from them. Since malware have predominantly been found to contain specific opcode sequences in them, the presence of the same sequences in any incoming file or network content can be taken up as a possible malware identification scheme. Malware detection systems help us to understand more about ways on how malware attack a system and how it can be prevented. The proposed method analyses malware executable files with the help of opcode information by converting the incoming executable files to assembly language thereby extracting opcode information (opcode count) from the same. The opcode count is then converted into opcode frequency which is stored in a CSV file format. The CSV file is passed to various machine learning algorithms like Decision Tree Classifier, Random Forest Classifier and Naive Bayes Classifier. Random Forest Classifier produced the highest accuracy and hence the same model was used to predict whether an incoming file contains a potential malware or not.

Intrusion Detection System (IDS) is one of the most important security tool for many security issues that are prevailing in today's cyber world. Intrusion Detection System is designed to scan the system applications and network traffic to detect suspicious activities and issue an alert if it is discovered. So many techniques are available in machine learning for intrusion detection. The main objective of this project is to apply machine learning algorithms to the data set and to compare and evaluate their performances. The proposed application has used the SVM (Support Vector Machine) and ANN (Artificial Neural Networks) Algorithms to detect the intrusion rates. Each algorithm is used to detect whether the requested data is authorized or contains any anomalies. While IDS scans the requested data if it finds any malicious information it drops that request. These algorithms have used Correlation-Based and Chi-Squared Based feature selection algorithms to reduce the dataset by eliminating the useless data. The preprocessed dataset is trained and tested with the models to obtain the prominent results, which leads to increasing the prediction accuracy. The NSL KDD dataset has been used for the experimentation. Finally, an accuracy of about 48% has been achieved by the SVM algorithm and 97% has been achieved by ANN algorithm. Henceforth, ANN model is working better than the SVM on this dataset.

The work shows the RFID cards as e-document rather than a paper passport with embedded chip as the e-passport. This type of Technological advancement creates benefits like the information can be stored electronically. The aim behind this is to reduce or stop the uses of illegal document. This will assure the security and prevent illegal entry in particular country by fake documents it will also maintain the privacy of the owner. Here, this research work has proposed an e-file verification device by means of RFID. Henceforth, this research work attempts to develop a new generation for file verification by decreasing the human effort. The most important idea of this examine is to make it feasible to get admission to the info of proprietor of the file the usage of RFID generation. For this the man or woman is issued RFID card. This card incorporates circuit which is used to store procedure information via way of modulating and demodulating the radio frequency sign transmitted. Therefore, the facts saved in this card are referred to the file element of the man or woman. With the help of the hardware of the proposed research work RFID Based E-Document verification provides a tag to the holder which produces waves of electromagnetic signal and then access the data. The purpose is to make the verification of document easy, secured and with less human intervention. In the proposed work, the comparative analysis is done using RFID technology in which 100 documents are verified in 500 seconds as compared to manual work done in 3000 seconds proves the system to be 6 times more efficient as compared to conventional method.

With the rapid advancements in information technology, data security has become an indispensable component. Cryptography performs a significant role in establishing information security. Computational problems have been utilized extensively by cryptographers to construct digital signature schemes. Digital signature schemes offer security services such as confidentiality, authenticity, integrity, and non-repudiation of a message. This paper proposes a modification of the Dilithium signature scheme that is secure against unforgeability attack based on the hardness of lattice problems such as Learning With Errors and Short Integer Solution over lattices. Using the rejection sampling technique, data is sampled from a uniform distribution to generate keys that are expanded into a matrix. The keys are hashed and signed by the sender to generate a message, which is then accepted by the receiver upon verification. Finally, the security analysis for the proposed signature scheme is provided with a strong emphasis on the security of the secret key. We prove that the attacker cannot forge a signature on a message, and recommended parameters are proposed.

This work investigates the impact of mechanical strain on wake-up behavior of planar HfO2 ferroelectric capacitor-based memory. External in-plane strain was applied using a four-point bending tool and strain impact on remanent polarization and coercive voltage of the ferroelectric was monitored. It was established that compressive strain is beneficial for 2Pr improvement, while tensile strain leads to its degradation, with a sensitivity of -8.4 ± 0.5 % per 0.1 % of strain. Strain-induced polarization rotation is considered to be the most likely mechanism affecting 2Pr At the same time, no strain impact on Vcwas observed in the investigated strain range. The results seen here can be utilized to undertake stress engineering of ferroelectric memory in order to improve its performance.

In Magnetic Josephson Junctions (MJJs) based on Superconductor-Insulator-Superconductor-Ferromagnet-Superconductor (SIS’FS), we provide evidence of an unconventional magnetic field behavior of the critical current characterized by an inverted magnetic hysteresis, i.e., an inverted shift of the whole magnetic field pattern when sweeping the external field. By thermoremanence measurements of S/F/S trilayers, we have ruled out that this uncommon behavior could be related to the F-stray fields. In principle, this finding could have a crucial role in the design and proper functioning of scalable cryogenic memories.

Magnetic soft robots are increasingly popular as they provide many advantages such as miniaturization and tetherless control that are ideal for applications inside the human body or in previously inaccessible locations.While non-magnetic elastomers have been extensively characterized and modelled for optimizing the fabrication of soft robots, a systematic material characterization of their magnetic counterparts is still missing. In this paper, commonly employed magnetic materials made out of Ecoflex™ 00-30 and Dragon Skin™ 10 with different concentrations of NdFeB microparticles were mechanically and magnetically characterized. The magnetic materials were evaluated under uniaxial tensile testing and their behavior analyzed through linear and hyperelastic model comparison. To determine the corresponding magnetic properties, we present a method to determine the magnetization vector, and magnetic remanence, by means of a force and torque load cell and large reference permanent magnet; demonstrating a high level of accuracy. Furthermore, we study the influence of varied magnitude impulse magnetizing fields on the resultant magnetizations. In combination, by applying improved, material-specific mechanical and magnetic properties to a 2-segment discrete magnetic robot, we show the potential to reduce simulation errors from 8.5% to 5.4%.

Autonomous systems have the potential to accomplish missions more quickly and effectively, while reducing risks to human operators and costs. However, since the use of autonomous systems is still relatively new, there are still a lot of challenges associated with trusting these systems. Without operators in direct control of all actions, there are significant concerns associated with endangering human lives or damaging equipment. For this reason, NATO has issued a challenge seeking to identify ways to improve decision-maker and operator trust when deploying autonomous systems, and de-risk their adoption. This paper presents the proposal of the winning solution to this NATO challenge. It approaches trust as a multi-dimensional concept, by incorporating the four dimensions of human-agent trust establishment in a digital twin context.

Autonomous navigation of a robot is more challenging in an uncontrolled environment owing to the necessity of coordination among several activities. This includes, creating a map of the surrounding, localizing the robot inside the map, generating a motion plan consistent with the map, executing the plan with control and all other tasks involved concurrently. Moreover, autonomous navigation problems are significant for future robotics applications such as package delivery, security, cleaning, agriculture, surveillance, search and rescue, construction, and transportation which take place in uncontrolled environments. Therefore, an attempt has been made in this research to develop a robot which could function as a security agent for a house to address the aforesaid particulars. This robot has the capability to navigate autonomously in the prescribed map of the operating zone by the user. The desired map can be generated using a Light Detection and Ranging (LiDAR) sensor. For robot navigation, it requires to pick out the robot location accurately itself, otherwise robot will not move autonomously to a particular target. Therefore, Adaptive Monte Carlo Localization (AMCL) method was used to validate the accuracy of robot localization process. Moreover, additional sensors were placed around the building to sense the prevailing security threats from intruders with the aid of the robot.

End users consider the data available through web as unmodified. Even when the web is secured by HTTPS, the data can be tampered in numerous tactical ways reducing trust on the integrity of data at the clients' end. One of the ways in which the web pages can be modified is via client side browser extensions. The extensions can transparently modify the web pages at client's end and can include new data to the web pages with minimal permissions. Clever modifications can be addition of a fake news or a fake advertisement or a link to a phishing website. We have identified through experimentation that such attacks are possible and have potential for serious damages. To prevent and detect such modifications we present a novel domain expressiveness based approach that uses DNS (Domain Name System) TXT records to express the Hash of important web pages that gets verified by the browsers to detect/thwart any modifications to the contents that are launched via client side malicious browser extensions or via cross site scripting. Initial experimentation suggest that the technique has potential to be used and deployed.

Performance-influence models can help stakeholders understand how and where configuration options and their interactions influence the performance of a system. With this understanding, stakeholders can debug performance behavior and make deliberate configuration decisions. Current black-box techniques to build such models combine various sampling and learning strategies, resulting in tradeoffs between measurement effort, accuracy, and interpretability. We present Comprex, a white-box approach to build performance-influence models for configurable systems, combining insights of local measurements, dynamic taint analysis to track options in the implementation, compositionality, and compression of the configuration space, without relying on machine learning to extrapolate incomplete samples. Our evaluation on 4 widely-used, open-source projects demonstrates that Comprex builds similarly accurate performance-influence models to the most accurate and expensive black-box approach, but at a reduced cost and with additional benefits from interpretable and local models.

Modern vehicles become increasingly digitalized with advanced information technology-based solutions like advanced driving assistance systems and vehicle-to-x communications. These systems are complex and interconnected. Rising complexity and increasing outside exposure has created a steadily rising demand for more cyber-secure systems. Thus, also standardization bodies and regulators issued standards and regulations to prescribe more secure development processes. This security, however, also has to be validated and verified. In order to keep pace with the need for more thorough, quicker and comparable testing, today's generally manual testing processes have to be structured and optimized. Based on existing and emerging standards for cybersecurity engineering, this paper therefore outlines a structured testing process for verifying and validating automotive cybersecurity, for which there is no standardized method so far. Despite presenting a commonly structured framework, the process is flexible in order to allow implementers to utilize their own, accustomed toolsets.

Nowadays, an increasing trend to use autonomous Unmanned Aerial Vehicles (UAV) for applications like logistics as well as security and surveillance can be recorded. Autonomic UAVs require robust and precise navigation to ensure efficient and safe operation even in strong multipath environments and (intended) interference. The need for robust navigation on UAVs implies the necessary integration of low-cost, lightweight, and compact array antennas as well as structures for multipath mitigation into the UAV platform. This article investigates a miniaturized antenna array mounted on top of vertical choke rings for robust navigation purposes. The array employs four 3D printed elements based on dielectric resonators capable of operating in all GNSS bands while compact enough for mobile applications such as UAV.

Mathematical morphology has been explored in deep learning architectures, as a substitute to convolution, for problems like pattern recognition and object detection. One major advantage of using morphology in deep learning is the utility of morphological erosion and dilation. Specifically, these operations naturally embody interpretability due to their underlying connections to the analysis of geometric structures. While the use of these operations results in explainable learned filters, morphological deep learning lacks attribution modeling, i.e., a paradigm to specify what areas of the original observed image are important. Furthermore, convolution-based deep learning has achieved attribution modeling through a variety of neural eXplainable Artificial Intelligence (XAI) paradigms (e.g., saliency maps, integrated gradients, guided backpropagation, and gradient class activation mapping). Thus, a problem for morphology-based deep learning is that these XAI methods do not have a morphological interpretation due to the differences in the underlying mathematics. Herein, we extend the neural XAI paradigm of saliency maps to morphological deep learning, and by doing, so provide an example of morphological attribution modeling. Furthermore, our qualitative results highlight some advantages of using morphological attribution modeling.

Cybersecurity is becoming very crucial in the today's world where technology is now not limited to just computers, smartphones, etc. It is slowly entering into things that are used on daily basis like home appliances, automobiles, etc. Thus, opening a new door for people with wrong intent. With the increase in speed of technology dealing with such issues also requires quick response from security people. Thus, dealing with huge variety of devices quickly will require some extent of automation in this field. Generating threat intelligence automatically and also including those which are multilingual will also add plus point to prevent well known major attacks. Here we are proposing an AI based SOAR system in which the data from various sources like firewalls, IDS, etc. is collected with individual event profiling using a deep-learning detection method. For this the very first step is that the collected data from different sources will be converted into a standardized format i.e. to categorize the data collected from different sources. For standardized format Here our system finds out about the true positive alert for which the appropriate/ needful steps will be taken such as the generation of Indicators of Compromise report and the additional evidences with the help of Security Information and Event Management system. The security alerts will be notified to the security teams with the degree of threat.

This research presents a model for assessing information systems cybersecurity maturity level. The main purpose of the model is to provide comprehensive support for information security specialists and auditors in checking information systems security level, checking security policy implementation, and compliance with security standards. The model synthesized based on controls and practices present in ISO 27001 and ISO 27002 and the neural network of direct signal propagation. The methodology described in this paper can also be extended to synthesis a model for different security control sets and, consequently, to verify compliance with another security standard or policy. The resulting model describes a real non-automated process of assessing the maturity of an IS at an acceptable level and it can be recommended to be used in the process of real audit of Information Security Management Systems.

This study offers an alternative to current implementations of key exchange by utilizing NFC technologies within android mobile devices. Supporting key exchange protocols along with cryptographic algorithms are offered, which meet current security standards whilst maintaining a short key length for optimal transfer between devices. Peer-to-peer and Host Card Emulation operational modes are observed to determine the best suited approach for key exchange. The proposed model offers end to end encryption between Client-Client as opposed to the usual Client-Server encryption offered by most Instant Messaging applications.

Critical infrastructures have to withstand advanced and persistent threats, which can be addressed using Byzantine fault tolerant state-machine replication (BFT-SMR). In practice, unattended cyberdefense systems rely on threat level detectors that synchronously inform them of changing threat levels. However, to have a BFT-SMR protocol operate unattended, the state-of-the-art is still to configure them to withstand the highest possible number of faulty replicas \$f\$ they might encounter, which limits their performance, or to make the strong assumption that a trusted external reconfiguration service is available, which introduces a single point of failure. In this work, we present ThreatAdaptive the first BFT-SMR protocol that is automatically strengthened or optimized by its replicas in reaction to threat level changes. We first determine under which conditions replicas can safely reconfigure a BFT-SMR system, i.e., adapt the number of replicas \$n\$ and the fault threshold \$f\$ so as to outpace an adversary. Since replicas typically communicate with each other using an asynchronous network they cannot rely on consensus to decide how the system should be reconfigured. ThreatAdaptive avoids this pitfall by proactively preparing the reconfiguration that may be triggered by an increasing threat when it optimizes its performance. Our evaluation shows that ThreatAdaptive can meet the latency and throughput of BFT baselines configured statically for a particular level of threat, and adapt 30% faster than previous methods, which make stronger assumptions to provide safety.

Python libraries are widely used for machine learning and scientific computing tasks today. APIs in Python libraries are deprecated due to feature enhancements and bug fixes in the same way as in other languages. These deprecated APIs are discouraged from being used in further software development. Manually detecting and replacing deprecated APIs is a tedious and time-consuming task due to the large number of API calls used in the projects. Moreover, the lack of proper documentation for these deprecated APIs makes the task challenging. To address this challenge, we propose an algorithm and a tool APIScanner that automatically detects deprecated APIs in Python libraries. This algorithm parses the source code of the libraries using abstract syntax tree (ASTs) and identifies the deprecated APIs via decorator, hard-coded warning or comments. APIScanner is a Visual Studio Code Extension that highlights and warns the developer on the use of deprecated API elements while writing the source code. The tool can help developers to avoid using deprecated API elements without the execution of code. We tested our algorithm and tool on six popular Python libraries, which detected 838 of 871 deprecated API elements. Demo of APIScanner: https://youtu.be/1hy\_ugf-iek. Documentation, tool, and source code can be found here: https://rishitha957.github.io/APIScanner.