Additive Manufacturing (AM) has revolutionized industries by enabling the production of complex, customized products with unparalleled efficiency. However, the increasing reliance on AM in critical sectors such as aerospace, healthcare, and defense has exposed it to significant cybersecurity and reliability challenges, including intellectual property theft, process sabotage, and data tampering. These vulnerabilities as well as reliability issues can compromise product integrity, safety, and operational continuity, posing severe risks to both industry and national security. In this work, we propose a novel methodology for modeling the AM process chain as a Cyber-Physical System (CPS) using multi-modal data structured in a graph format. Our methodology leverages Graph Neural Networks (GNNs) to detect and localize anomalies across diverse data modalities, enabling precise identification of both the nature and source of attack/fault. By integrating data fusion, advanced anomaly classification, and localization techniques, our solution provides a robust methodology for enhancing the security and reliability of AM processes, ensuring their safe deployment in critical applications. Furthermore, the proposed technique is adaptable to other industrial systems, underscoring its potential for broader impact in securing critical infrastructure.

The growing complexity of global supply chains has made hardware Trojans a significant threat in sensor-based power electronics. Traditional Trojan designs depend on digital triggers or fixed threshold conditions that can be detected during standard testing. In contrast, we introduce Environ-mental Rate Manipulation (ERM), a novel Trojan triggering mechanism that activates by monitoring the rate of change in environmental parameters rather than their absolute values. This approach allows the Trojan to remain inactive under normal conditions and evade redundancy and sensor-fusion defenses. We implement a compact 14 µm2circuit that mea-sures capacitor charging rates in standard sensor front-ends and disrupts inverter pulse-width modulation PWM signals when a rapid change is induced. Experiments on a commercial Texas Instruments solar inverter demonstrate that ERM can trigger catastrophic driver chip failure. Furthermore, ETAP simulations indicate that a single compromised 100 kW in-verter may initiate cascading grid instabilities. The attack's significance extends beyond individual sensors to entire classes of environmental sensing systems common in power electronics, demonstrating fundamental challenges for hardware security.

Function name prediction is crucial for understanding stripped binaries in software reverse engineering, a key step for enabling subsequent vulnerability analysis and patching. However, existing approaches often struggle with architecture-specific limitations, data scarcity, and diverse naming conventions. We present AGNOMIN, a novel architecture-agnostic approach for multi-label function name prediction in stripped binaries. AGNOMIN builds Feature-Enriched Hierarchical Graphs (FEHGs), combining Control Flow Graphs, Function Call Graphs, and dynamically learned PCode features. A hierarchical graph neural network processes this enriched structure to generate consistent function representations across architectures, vital for scalable security assessments. For function name prediction, AGNOMIN employs a Renée-inspired decoder, enhanced with an attention-based head layer and algorithmic improvements. We evaluate AGNOMIN on a comprehensive dataset of 9,000 ELF executable binaries across three architectures, demonstrating its superior performance compared to state-of-the-art approaches, with improvements of up to 27.17% in precision and 55.86% in recall across the testing dataset. Moreover, AGNOMIN generalizes well to unseen architectures, achieving 5.89% higher recall than the closest baseline. AGNOMIN's practical utility has been validated through security hackathons, where it successfully aided reverse engineers in analyzing and patching vulnerable binaries across different architectures.

Modern optical mouse sensors, with their advanced precision and high responsiveness, possess an often overlooked vulnerability: they can be exploited for side-channel attacks. This paper introduces Mic-E-Mouse, the first-ever side-channel attack that targets high-performance optical mouse sensors to covertly eavesdrop on users. We demonstrate that audio signals can induce subtle surface vibrations detectable by a mouse's optical sensor. Remarkably, user-space software on popular operating systems can collect and broadcast this sensitive side channel, granting attackers access to raw mouse data without requiring direct system-level permissions. Initially, the vibration signals extracted from mouse data are of poor quality due to non-uniform sampling, a non-linear frequency response, and significant quantization. To overcome these limitations, Mic-E-Mouse employs a sophisticated end-to-end data filtering pipeline that combines Wiener filtering, resampling corrections, and an innovative encoder-only spectrogram neural filtering technique. We evaluate the attack's efficacy across diverse conditions, including speaking volume, mouse polling rate and DPI, surface materials, speaker languages, and environmental noise. In controlled environments, Mic-E-Mouse improves the signal-to-noise ratio (SNR) by up to +19 dB for speech reconstruction. Furthermore, our results demonstrate a speech recognition accuracy of roughly 42% to 61% on the AudioMNIST and VCTK datasets. All our code and datasets are publicly accessible on Mic-E-Mouse website11https://sites.google.com/view/mic-e-mouse.

Uncertainty Quantification (UQ) is crucial for ensuring the reliability of machine learning models deployed in real-world autonomous systems. However, existing approaches typically quantify task-level output prediction uncertainty without considering epistemic uncertainty at the multimodal feature fusion level, leading to sub-optimal outcomes. Additionally, popular uncertainty quantification methods, e.g., Bayesian approximations, remain challenging to deploy in practice due to high computational costs in training and inference. In this paper, we propose HyperDUM, a novel deterministic uncertainty method (DUM) that efficiently quantifies feature-level epistemic uncertainty by leveraging hyper-dimensional computing. Our method captures the channel and spatial uncertainties through channel and patch -wise projection and bundling techniques respectively. Multimodal sensor features are then adaptively weighted to mitigate uncertainty propagation and improve feature fusion. Our evaluations show that HyperDUM on average outperforms the state-of-the-art (SOTA) algorithms by up to 2.01%/1.27% in 3D Object Detection and up to 1.29% improvement over baselines in semantic segmentation tasks under various types of uncertainties. Notably, HyperDUM requires 2.36× less Floating Point Operations and up to 38.30× less parameters than SOTA methods, providing an efficient solution for real-world autonomous systems.

We study the application of emerging chiplet-based Neural Processing Units to accelerate vehicular AI perception workloads in constrained automotive settings. The motivation stems from how chiplets technology is becoming integral to emerging vehicular architectures, providing a cost-effective tradeoff between performance, modularity, and customization; and from perception models being the most computationally demanding workloads in a autonomous driving system. Using the Tesla Autopilot perception pipeline as a case study, we first breakdown its constituent models and profile their performance on different chiplet accelerators. From the insights, we propose a novel scheduling strategy to efficiently deploy perception workloads on multi-chip AI accelerators. Our experiments using a standard DNN performance simulator, MAESTRO, show our approach realizes 82% and 2.8 × increase in throughput and processing engines utilization compared to monolithic accelerator designs.

Software vulnerabilities remain pervasive, even with the rise of AI-powered code assistants, advanced static analysis tools, and comprehensive testing frameworks. It’s clear that we must move beyond merely preventing these bugs; we need to eliminate them swiftly and efficiently. However, manual code intervention is slow, expensive, and can often introduce new security flaws, especially in legacy codebases. The advent of highly advanced Large Language Models (LLMs) presents a significant opportunity for automated software defect patching. We introduce LLM4CVE, an LLM-based iterative pipeline designed for robust and accurate repair of vulnerable functions in real-world code. We evaluate our pipeline using State-of-the-Art LLMs, including GPT-3.5, GPT-4o, Llama 3 8B, and Llama 3 70B. Our results demonstrate a human-verified quality score of 8.51/10 and a 20% increase in ground-truth code similarity with Llama 3 70B. To foster further research in LLM-based vulnerability repair, we release our evaluation framework, fine-tuned model weights, and experimental results on our website: https://sites.google.com/view/llm4cve

Deep learning has been widely adopted for human activity recognition (HAR) while generalizing a trained model across diverse users and scenarios remains challenging due to distribution shifts (DS). The inherent low-resource challenge in HAR, i.e., collecting and labeling adequate human-involved data can be prohibitively costly, further raising the difficulty of tackling DS. We propose TACO, a novel transformer-based contrastive meta-learning approach for generalizable HAR. TACO addresses DS by synthesizing virtual target domains in training with explicit consideration of model generalizability. Additionally, we extract expressive feature with the attention mechanism of Transformer and incorporate the supervised contrastive loss function within our meta-optimization to enhance representation learning. Our evaluation demonstrates that TACO achieves notably better performance across various low-resource DS scenarios.

Industrial Control Systems (ICS) rely heavily on ProgrammableLogic Controllers (PLCs) to manage critical infrastructure, yet analyzing PLC executables remains challenging due to diverse proprietary compilers and limited access to source code.To bridge this gap, we introduce PLC-BEAD, a comprehensive dataset containing 2431 compiled binaries from 700+ PLC programs across four major industrial compilers (CoDeSys, GEB, OpenPLC-V2, OpenPLC-V3).This novel dataset uniquely pairs each binary with its original Structured Text source code and standardized functionality labels, enabling both binary-level and source-level analysis.We demonstrate the dataset's utility through PLCEmbed, a transformer-based framework for binary code analysis that achieves 93% accuracy in compiler provenance identification and 42% accuracy in finegrained functionality classification across 22 industrial control categories.Through comprehensive ablation studies, we analyze how compiler optimization levels, code patterns, and class distributions influence model performance.We provide detailed documentation of the dataset creation process, labeling taxonomy, and benchmark protocols to ensure reproducibility.Both PLC-BEAD and PLCEmbed are released as open-source resources to foster research in PLC security, reverse engineering, and ICS forensics, establishing new baselines for data-driven approaches to industrial cybersecurity.

Advances in Internet of Things (IoT) wearable sensors and edge-artificial intelligence (Edge-AI) have enabled practical realizations of machine learning (ML)-enabled mobile sensing applications like human activity recognition (HAR). The effective deployment of these data-driven models necessitates learning robust representations capable of handling prevalent distribution shifts (DS), including new users, device positions, rotations, and more. In that respect, contrastive learning (CL) has shown promise in learning transformation-invariant features, outperforming traditional HAR methods. However, recent findings reveal that the contrastive loss induces shrinkage and expansion of the feature space which may limit the generalization capacity of the model. To address this, we propose DisCovHAR, a contrastive attention method to selectively apply the contrastive loss to a subset of the feature space through the transformer encoder attention mechanism. Extensive experiments on three HAR datasets (DSADS, PAMAP2, and USCHAD) demonstrate its superiority over state-of-the-art methods. Specifically, our approach yields up to 4.47% and 7.82% average accuracy improvements in subject-wise and position-wise generalization settings. Furthermore, DisCovHAR demonstrates up to 5.07% increased robustness compared to prior methods under multivariate distribution shift scenarios.

The globalization of the Integrated Circuit (IC) supply chain has moved most of the design, fabrication, and testing process from a single trusted entity to various untrusted third party entities around the world. The risk of using untrusted third-Party Intellectual Property (3PIP) is the possibility for adversaries to insert malicious modifications known as Hardware Trojans (HTs). These HTs can compromise the integrity, deteriorate the performance, and deny the functionality of the intended design. Various HT detection methods have been proposed in the literature; however, many fall short due to their reliance on a golden reference circuit, a limited detection scope, the need for manual code review, or the inability to scale with large modern designs. We propose a novel golden reference-free HT detection method for both Register Transfer Level (RTL) and gate-level netlists by leveraging Graph Neural Networks (GNNs) to learn the behavior of the circuit through a Data Flow Graph (DFG) representation of the hardware design. We evaluate our model on a custom dataset by expanding the Trusthub HT benchmarks trusthub1. The results demonstrate that our approach detects unknown HTs with 97% recall (true positive rate) very fast in 21.1ms for RTL and 84% recall in 13.42s for Gate-Level Netlist.

Machine Learning (ML) has proven effective in Integrated Circuits (IC) security, particularly in Hardware Trojan (HT) detection. However, a model’s generalization potential depends on its ability to address distribution shifts (DS) in unseen data. Mitigating DS enhances a model’s adaptability to novel variations and threats within the dynamic realm of IC designs and HTs. We formulate HT detection as a DS problem, introducingDART, a novelDistribution-AwareHT detection framework, to enhance model generalization. ApplyingDARTon state-of-the-art Graph Neural Network architecture yields up to 22.96% and 17.37% F1-score improvements for unseen IC designs diverging significantly from the training data.

Autonomous vehicles (AVs) can operate in complex environments that require multiple types of sensors in their perception approaches. However, a single sensing configuration is not tenable in all environments, and adapting the perception approach to different domains is a challenging task. Both sensor fusion and ensemble learning methods utilize the diversity of multimodal sensor data (e.g., cameras, radar, lidar) to increase perception performance under challenging sensing conditions. In this paper, we conduct the first analysis examining how sensor diversity can impact performance across different AV perception tasks. We propose ensembles of multimodal models that leverage diversity and assess their ability to address the challenge of adaptation in AV perception. We introduce a novel, model-agnostic framework,DivFusE, that identifies the level of diversity that individual sensor modalities contribute to perception ensembles and employs adaptation based on this diversity. We benchmark our approach on five public datasets containing semantic segmentation and object detection tasks, highlighting scenarios where our proposed framework can improve perception performance by 70% (fuse it) and scenarios where fusion can reduce perception performance (lose it). Our findings indicate that the performance improvements of using multimodal ensembles is greater on objection detection compared to semantic segmentation, and that sensing conditions with higher (lower) levels of diversity further increase (decrease) performance. Ultimately, the analysis in this work highlights scenarios where these multimodal ensembles should and should not be deployed and can be used to develop more adaptive multimodal sensor fusion systems.

Differential Pressure Sensors are widely deployed to monitor critical environments. However, our research unveils a previously overlooked vulnerability: their high sensitivity to pressure variations makes them susceptible to acoustic side-channel attacks. We demonstrate that the pressure-sensing diaphragms in DPS can inadvertently capture subtle air vibrations caused by speech, which propagate through the sensor’s components and affect the pressure readings. Exploiting this discovery, we introduce BaroVox, a novel attack that reconstructs speech from DPS readings, effectively turning DPS into "a fly on the wall." We model the effect of sound on DPS, exploring the limits and challenges of acoustic leakage. To overcome these challenges, we propose two solutions: a signal-processing approach using a unique spectral subtraction method and a deep learning-based approach for keyword classification. Evaluations under various conditions demonstrate BaroVox’s effectiveness, achieving a word error rate of 0.29 for manual recognition and 90.51% accuracy for automatic recognition. Our findings highlight the significant privacy implications of this vulnerability. We also discuss potential defense strategies to mitigate the risks posed by BaroVox.

Many real-world applications of the Internet of Things (IoT) employ machine learning (ML) algorithms to analyze time series information collected by interconnected sensors. However, distribution shift, a fundamental challenge in data-driven ML, arises when a model is deployed on a data distribution different from the training data and can substantially degrade model performance. Additionally, increasingly sophisticated deep neural networks (DNNs) are required to capture intricate spatial and temporal dependencies in multi-sensor time series data, often exceeding the capabilities of today's edge devices. In this paper, we propose SMORE, a novel resource-efficient domain adaptation (DA) algorithm for multi-sensor time series classification, leveraging the efficient and parallel operations of hyperdimensional computing. SMORE dynamically customizes test-time models with explicit consideration of the domain context of each sample to mitigate the negative impacts of domain shifts. Our evaluation on a variety of multi-sensor time series classification tasks shows that SMORE achieves on average 1.98% higher accuracy than state-of-the-art (SOTA) DNN-based DA algorithms with 18.81x faster training and 4.63x faster inference.

In the current technological landscape, Internet of Things (IoT) systems are deeply embedded in numerous facets of daily life, from domestic settings to critical infrastructure, which underscores the importance of these systems security and integrity. The constrained nature of IoT devices, in terms of computational capacity, economic limitations, or time-to-market, makes them vulnerable to security breaches and system failures. Additionally, the hybrid essence of IoT- combining the physical domain via sensor interfaces and the cyber domain through communication networks and cloud connectivity- further complicates mitigating these threats. While numerous techniques for either network intrusion detection or sensor anomaly detection exist, an integrated approach that synergistically combines information from both domains is absent. This paper proposes a multi-modal data fusion technique, which, for the first time, melds sensor and communication data. This approach underscores the interdependencies between the components, provides contextual embeddings for data from each element, and integrates the system's physical and cyber features into a graph-based representation. Harnessing the power of Graph Neural Networks (GNNs), we capture the normal state and context of the system, facilitating the detection of anomalies and intrusions. Additionally, our model discerns between network and sensor-based attacks, pinpointing the anomaly's origin, thereby expediting post-incident recovery. Optimized for fog-computing environments, our solution ensures real-time oversight. Rigorous testing on greenhouse IoT systems indicates the efficacy of our model, with a commendable 22% improvement in Fl-score over singular modal techniques.

Emerging multi-model workloads with heavy models like recent large language models significantly increased the compute and memory demands on hardware. To address such increasing demands, designing a scalable hardware architecture became a key problem. Among recent solutions, the 2.5D silicon interposer multi-chip module (MCM)-based AI accelerator has been actively explored as a promising scalable solution due to their significant benefits in the low engineering cost and composability. However, previous MCM accelerators are based on homogeneous architectures with fixed dataflow, which encounter major challenges from highly heterogeneous multi-model work-loads due to their limited workload adaptivity. Therefore, in this work, we explore the opportunity in the heterogeneous dataflow MCM AI accelerators. We identify the scheduling of multi-model workload on heterogeneous dataflow MCM AI accelerator is an important and challenging problem due to its significance and scale, which reaches$\mathbf{O}(10^{56})$even for a two-model workload on 6×6 chiplets. We develop a set of heuristics to navigate the huge scheduling space and codify them into a scheduler, SCAR, with advanced techniques such as inter-chiplet pipelining. Our evaluation on ten multi-model workload scenarios for datacenter multitenancy and AR/VR use-cases has shown the efficacy of our approach, achieving on average 27.6% and 29.6% less energy-delay product (EDP) for the respective applications settings compared to homogeneous baselines.

The metaverse is a virtual space that combines physical and digital elements, creating immersive and connected digital worlds. For autonomous mobility, it enables new possibilities with edge computing and digital twins (DTs) that offer virtual prototyping, prediction, and more. DTs can be created with 3D scene reconstruction methods that capture the real world's geometry, appearance, and dynamics. However, sending data for real-time DT updates in the metaverse, such as camera images and videos from connected autonomous vehicles (CAVs) to edge servers, can increase network congestion, costs, and latency, affecting metaverse services. Herein, a new method is proposed based on distributed radiance fields (RFs), multi-access edge computing (MEC) network for video compression and metaverse DT updates. RF-based encoder and decoder are used to create and restore representations of camera images. The method is evaluated on a dataset of camera images from the CARLA simulator. Data savings of up to 80% were achieved for H.264 I-frame - P-frame pairs by using RFs instead of I-frames, while maintaining high peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) qualitative metrics for the reconstructed images. Possible uses and challenges for the metaverse and autonomous mobility are also discussed.

Effectively capturing intricate interactions among road users plays a critical role in achieving safe navigation for autonomous vehicles. While graph learning (GL) has emerged as a promising approach to tackle this challenge, existing GL models rely on predefined domain-specific graph extraction rules and often fail in real-world dynamic scenarios. Additionally, these graph extraction rules severely impede the capability of existing GL methods to generalize knowledge across domains. To address this issue, we propose RoadScene2Graph (RS2G), an innovative autonomous scenario understanding framework with a novel data-driven graph extraction and modeling approach that dynamically captures the diverse relations among road users. Our evaluations show that on average RS2G outperforms the state-of-the-art (SOTA) rule-based graph extraction method by 4.47% and the SOTA deep learning model by 22.19% in subjective risk assessment. RS2G also delivers notably better performance in transferring knowledge gained from simulations to unseen real-world scenarios.

Split Computing has become a prominent resource-efficient method to enable machine learning (ML) applications on user-constrained edge devices, where compute-intensive ML workloads can be delegated to remote cloud servers for processing. However, the exposure of data to the cloud service providers as such raises privacy alarms due to the possible leakage of sensitive user information. On a relevant note, typical deep neural network (DNN) design frameworks do not take into account model splitting and its complications at the early design stages of DNNs. Thus, a natural question arises on how to bridge this gap and optimize the DNN design process such that split computing operations can meet the requirements of accuracy, performance, and privacy. In this paper, we strive to address this question through adopting a privacy-by-design approach, where privacy is characterized either as a constraint or an objective to realize privacy-aware models tailored for split computing. Using the -differential privacy standard for our case study, we conduct intensive empirical analysis on the relation between architectural parameters and intrinsic privacy budgets, and propose PrivyNAS – a privacy-aware Neural Architecture Search framework for split computing. On the CIFAR-10 dataset, our approach has demonstrated promising results in providing DNN architectures that balance the required design trade-offs.

Thanks to the rapid growth in wearable technologies and advancements in machine learning, monitoring complex human contexts becomes feasible, paving the way to develop human-in-the-loop IoT systems that naturally evolve to adapt to the human and environment state autonomously. Nevertheless, a central challenge in designing many of these IoT systems arises from the requirement to infer the human mental state, such as intention, stress, cognition load, or learning ability. While different human contexts can be inferred from the fusion of different sensor modalities that can correlate to a particular mental state, the human brain provides a richer sensor modality that gives us more insights into the required human context. This paper proposes ERUDITE, a human-in-the-loop IoT system for the learning environment that exploits recent wearable neurotechnology to decode brain signals. Through insights from concept learning theory, ERUDITE can infer the human state of learning and understand when human learning increases or declines. By quantifying human learning as an input sensory signal, ERUDITE can provide adequate personalized feedback to humans in a learning environment to enhance their learning experience. ERUDITE is evaluated across 15 participants and showed that by using the brain signals as a sensor modality to infer the human learning state and providing personalized adaptation to the learning environment, the participants’ learning performance increased on average by 26%. Furthermore, to evaluate ERUDITE practicality and scalability, we showed that ERUDITE can be deployed on an edge-based prototype consuming 75 mW power on average with 100 MB memory footprint.

Network-based security has emerged as an increasingly critical challenge in the domain of the Internet of Things (IoT). A number of network intrusion detection systems (NIDS), typically relying on sophisticated machine learning (ML) algorithms, have been proposed to monitor network traffic and detect malicious activity. However, these NIDS designs require extensive memory and computational power, exceeding the capability of today’s IoT devices, and often fail to provide timely detection of network attacks. To tackle this issue, we propose HyperDetect, the first attempt at NIDS modeling that leverages the highly efficient and parallel operations of brain-inspired hyperdimensional computing (HDC). Our innovative model updating method effectively mitigates model saturation and significantly reduces the number of retraining iterations needed to reach convergence. Additionally, we employ a novel dynamic encoding technique to regenerate insignificant dimensions, considerably lowering the dimensionalities required to achieve high-quality performance and further accelerating the learning process. HyperDetect delivers on average 5.02× faster training and 31.83× faster inference compared to state-of-the-art (SOTA) learning approaches on a wide range of network intrusion classification tasks. We also extensively evaluate HyperDetect on embedded hardware to demonstrate its low-latency and resource-efficient characteristics.

Autonomous vehicles are cyber-physical systems that combine embedded computing and deep learning with physical systems to perceive the world, predict future states, and safely control the vehicle through changing environments. The ability of an autonomous vehicle to accurately predict the motion of other road users across a wide range of diverse scenarios is critical for both motion planning and safety. However, existing motion prediction methods do not explicitly model contextual information about the environment, which can cause significant variations in performance across diverse driving scenarios. To address this limitation, we propose CASTNet : a dynamic, context-aware approach for motion prediction that (i) identifies the current driving context using a spatio-temporal model, (ii) adapts an ensemble of motion prediction models to fit the current context, and (iii) applies novel trajectory fusion methods to combine predictions output by the ensemble. This approach enables CASTNet to improve robustness by minimizing motion prediction error across diverse driving scenarios. CASTNet is highly modular and can be used with various existing image processing backbones and motion predictors. We demonstrate how CASTNet can improve both CNN-based and graph-learning-based motion prediction approaches and conduct ablation studies on the performance, latency, and model size for various ensemble architecture choices. In addition, we propose and evaluate several attention-based spatio-temporal models for context identification and ensemble selection. We also propose a modular trajectory fusion algorithm that effectively filters, clusters, and fuses the predicted trajectories output by the ensemble. On the nuScenes dataset, our approach demonstrates more robust and consistent performance across diverse, real-world driving contexts than state-of-the-art techniques.

Current power system state estimation and control methods are susceptible to false data injection attacks (FDIAs), which introduce faulty measurements throughout the grid that decrease grid stability. Fusing sensor measurements can reduce errors in state estimation, and data-driven approaches have been increasingly used for defense against FDIAs. However, current methods often lack adaptability and focus only on detection while failing to address attack effects on estimation and control. This work proposes AstroFusion, an adaptive data fusion framework that makes power grid state estimation and control more resilient to attacks. AstroFusion employs deep multilayer perceptrons to identify which sensors may be under attack and adaptively selects from an ensemble of data-driven models to improve the state estimation. This work is the first to characterize the performance of autonomous power grid controllers in the presence of varying attacks. Results are shown on IEEE 14-bus, IEEE 36-bus, and IEEE 118-bus systems.

Heterogeneous MPSoCs comprise diverse processing units of varying compute capabilities. To date, the mapping strategies of neural networks (NNs) onto such systems are yet to exploit the full potential of processing parallelism, made possible through both the intrinsic NNs’ structure and underlying hardware composition. In this paper, we propose a novel framework to effectively map NNs onto heterogeneous MPSoCs in a manner that enables them to leverage the underlying processing concurrency. Specifically, our approach identifies an optimal partitioning scheme of the NN along its ‘width’ dimension, which facilitates deployment of concurrent NN blocks onto different hardware computing units. Additionally, our approach contributes a novel scheme to deploy partitioned NNs onto the MPSoC as dynamic multi-exit networks for additional performance gains. Our experiments on a standard MPSoC platform have yielded dynamic mapping configurations that are 2.1x more energy-efficient than the GPU-only mapping while incurring 1.7x less latency than DLA-only mapping.

Runtime energy management has become quintessential for multi-sensor autonomous systems at the edge for achieving high performance given the platform constraints. Typical for such systems, however, is to have their controllers designed with formal guarantees on safety that precede in priority such optimizations, which in turn limits their application in real settings. In this paper, we propose a novel energy optimization framework that is aware of the autonomous system’s safety state, and leverages it to regulate the application of energy optimization methods so that the system’s formal safety properties are preserved. In particular, through the formal characterization of a system’s safety state as a dynamic processing deadline, the computing workloads of the underlying models can be adapted accordingly. For our experiments, we model two popular runtime energy optimization methods, offloading and gating, and simulate an autonomous driving system (ADS) use-case in the CARLA simulation environment with performance characterizations obtained from the standard Nvidia Drive PX2 ADS platform. Our results demonstrate that through a formal awareness of the perceived risks in the test case scenario, energy efficiency gains are still achieved (reaching 89.9%) while maintaining the desired safety properties.

Dynamic neural networks (DyNNs) have become viable techniques to enable intelligence on resource-constrained edge devices while maintaining computational efficiency. In many cases, the implementation of DyNNs can be sub-optimal due to its underlying backbone architecture being developed at the design stage independent of both: (i) potential support for dynamic computing, e.g. early exiting, and (ii) resource efficiency features of the underlying hardware, e.g., dynamic voltage and frequency scaling (DVFS). Addressing this, we present HADAS, a novel Hardware-Aware Dynamic Neural Architecture Search framework that realizes DyNN architectures whose backbone, early exiting features, and DVFS settings have been jointly optimized to maximize performance and resource efficiency. Our experiments using the CIFAR-100 dataset and a diverse set of edge computing platforms have shown that HADAS can elevate dynamic models' energy efficiency by up to 57% for the same level of accuracy scores. Our code is available at https://github.com/HalimaBouzidi/HADAS

With the rapid evolution of the Internet of Things, many real-world applications utilize heterogeneously connected sensors to capture time-series information. Edge-based machine learning (ML) methodologies are often employed to analyze locally collected data. However, a fundamental issue across data-driven ML approaches is distribution shift. It occurs when a model is deployed on a data distribution different from what it was trained on, and can substantially degrade model performance. Additionally, increasingly sophisticated deep neural networks (DNNs) have been proposed to capture spatial and temporal dependencies in multi-sensor time series data, requiring intensive computational resources beyond the capacity of today's edge devices. While brain-inspired hyperdimensional computing (HDC) has been introduced as a lightweight solution for edge-based learning, existing HDCs are also vulnerable to the distribution shift challenge. In this paper, we propose DOMINO, a novel HDC learning framework addressing the distribution shift problem in noisy multi-sensor time-series data. DOMINO leverages efficient and parallel matrix operations on high-dimensional space to dynamically identify and filter out domain-variant dimensions. Our evaluation on a wide range of multi-sensor time series classification tasks shows that DOMINO achieves on average 2.04% higher accuracy than state-of-the-art (SOTA) DNN-based domain generalization techniques, and delivers$16.34\times$faster training and$2.89\times$faster inference. More importantly, DOMINO exhibits notably better performance when learning from partially labeled data and highly imbalanced data, and provides$10.93\times$higher robustness against hardware noises than SOTA DNNs.

Autonomous systems (AS) are systems that can adapt and change their behaviors in response to unanticipated events and include systems such as aerial drones, autonomous vehicles, and ground/aquatic robots. AS require a wide array of sensors, deep learning models, and powerful hardware platforms to perceive the environment and safely operate in real-time. However, in many contexts, some sensing modalities negatively impact perception while increasing the system's overall energy consumption. Since AS are often energy-constrained edge devices, energy-efficient sensor fusion methods have been proposed. However, existing methods either fail to adapt to changing scenario conditions or to optimize system-wide energy efficiency. We propose CARMA, a context-aware sensor fusion approach that uses context to dynamically reconfigure the computation flow on a field-programmable gate array (FPGA) at runtime. By clock gating unused sensors and model sub-components, CARMA significantly reduces the energy used by a multi-sensory object detector without compromising performance. We use a deep learning processor unit (DPU) based reconfiguration approach to minimize the latency of model reconfiguration. We evaluate multiple context identification strategies, propose a novel system-wide energy-performance joint optimization, and evaluate scenario-specific perception performance. Across challenging real-world sensing contexts, CARMA outperforms state-of-the-art methods with up to 1.3× speedup and 73% lower energy consumption.

This article presents a resource-efficient adaptive sampling methodology for classifying electrocardiogram (ECG) signals into different heart rhythms. We present our methodology in two folds: ( i ) the design of a novel real-time adaptive neural network architecture capable of classifying ECG signals with different sampling rates and ( ii ) a runtime implementation of sampling rate control using deep reinforcement learning (DRL). By using essential morphological details contained in the heartbeat waveform, the DRL agent can control the sampling rate and effectively reduce energy consumption at runtime. To evaluate our adaptive classifier, we use the MIT-BIH database and the recommendation of the AAMI to train the classifiers. The classifier is designed to recognize three major types of arrhythmias, which are supraventricular ectopic beats (SVEB), ventricular ectopic beats (VEB), and normal beats (N). The performance of the arrhythmia classification reaches an accuracy of 97.2% for SVEB and 97.6% for VEB beats. Moreover, the designed system is 7.3× more energy-efficient compared to the baseline architecture, where the adaptive sampling rate is not utilized. The proposed methodology can provide reliable and accurate real-time ECG signal analysis with performances comparable to state-of-the-art methods. Given its time-efficient, low-complexity, and low-memory-usage characteristics, the proposed methodology is also suitable for practical ECG applications, in our case for arrhythmia classification, using resource-constrained devices, especially wearable healthcare devices and implanted medical devices.

Wearable medical technology has become increasingly popular in recent years. One function of wearable health devices is stress detection, which relies on sensor inputs to determine a patient’s mental state. This continuous, real-time monitoring can provide healthcare professionals with vital physiological data and enhance the quality of patient care. Current methods of stress detection lack: (i) robustness—wearable health sensors contain high levels of measurement noise that degrades performance, and (ii) adaptation—static architectures fail to adapt to changing contexts in sensing conditions. We propose to address these deficiencies with SELF-CARE, a generalized selective sensor fusion method of stress detection that employs novel techniques of context identification and ensemble machine learning. SELF-CARE uses a learning-based classifier to process sensor features and model the environmental variations in sensing conditions known as the noise context. SELF-CARE uses noise context to selectively fuse different sensor combinations across an ensemble of models to perform robust stress classification. Our findings suggest that for wrist-worn devices, sensors that measure motion are most suitable to understand noise context, while for chest-worn devices, the most suitable sensors are those that detect muscle contraction. We demonstrate SELF-CARE’s state-of-the-art performance on the WESAD dataset. Using wrist-based sensors, SELF-CARE achieves 86.34% and 94.12% accuracy for the 3-class and 2-class stress classification problems, respectively. For chest-based wearable sensors, SELF-CARE achieves 86.19% (3-class) and 93.68% (2-class) classification accuracy. This work demonstrates the benefits of utilizing selective, context-aware sensor fusion in mobile health sensing that can be applied broadly to Internet of Things applications.

Edge computing is to be widely adopted for autonomous systems (ASs) applications as compute-intensive processing tasks can be offloaded to compute-capable servers located at the edge of the network infrastructure. Given the critical nature of numerous AS applications, their tasks are mostly governed by strict execution deadlines to alleviate any safety concerns from delayed responses. Although wireless link uncertainty has prompted recent works to designate redundant local execution as an offloading fail-safe to ensure these deadlines are met, frequent invocation of such fail-safe mechanisms can potentially undermine the extent of performance gains from offloading. In this article, we thoroughly analyze how redundant execution overheads can influence the overall performance. Then, we present TESTUDO, a methodology to optimize the energy consumption for latency-sensitive AS applications employing collaborative edge computing. Primarily, our methodology encompasses two main stages: 1) designing processing pipelines supporting optimal offloading points and fail-safe integration using modular design techniques and 2) developing a context-aware adaptive runtime solution based on deep reinforcement learning to adapt the mode of operation according to the wireless network status. Our experiments for end-to-end control and object detection use-cases have shown that TESTUDO achieved energy gains reaching up to 31% and 13.4% (15.9% and 5.3% on average) for the former and latter, respectively, while incurring little-to-no degradation in prediction scores (< 1% change) from state-of-the-art strategies.

Graph Neural Networks (GNNs) are becoming increasingly popular for vision-based applications due to their intrinsic capacity in modeling structural and contextual relations between various parts of an image frame. On another front, the rising popularity of deep vision-based applications at the edge has been facilitated by the recent advancements in heterogeneous multi-processor Systems on Chips (MPSoCs) that enable inference under real-time, stringent execution requirements. By extension, GNNs employed for vision-based applications must adhere to the same execution requirements. Yet contrary to typical deep neural networks, the irregular flow of graph learning operations poses a challenge to running GNNs on such heterogeneous MPSoC platforms. In this paper, we propose a novel unified design-mapping approach for efficient processing of vision GNN workloads on heterogeneous MPSoC platforms. Particularly, we develop MaGNAS, a mapping-aware Graph Neural Architecture Search framework. MaGNAS proposes a GNN architectural design space coupled with prospective mapping options on a heterogeneous SoC to identify model architectures that maximize on-device resource efficiency. To achieve this, MaGNAS employs a two-tier evolutionary search to identify optimal GNNs and mapping pairings that yield the best performance trade-offs. Through designing a supernet derived from the recent Vision GNN (ViG) architecture, we conducted experiments on four (04) state-of-the-art vision datasets using both ( i ) a real hardware SoC platform (NVIDIA Xavier AGX) and ( ii ) a performance/cost model simulator for DNN accelerators. Our experimental results demonstrate that MaGNAS is able to provide 1.57 × latency speedup and is 3.38 × more energy-efficient for several vision datasets executed on the Xavier MPSoC vs. the GPU-only deployment while sustaining an average 0.11% accuracy reduction from the baseline.

Over the last six years, several papers used memory deduplication to trigger various security issues, such as leaking heap-address and causing bit-flip in the physical memory. The most essential requirement for successful memory deduplication is to provide identical copies of a physical page. Recent works use a brute-force approach to create identical copies of a physical page that is an inaccurate and time-consuming primitive from the attacker’s perspective.

A Negative Pressure Room (NPR) is an essential requirement by the Bio-Safety Levels (BSLs) in biolabs or infectious-control hospitals to prevent deadly pathogens from being leaked from the facility. An NPR maintains a negative pressure inside with respect to the outside reference space so that microbes are contained inside of an NPR. Nowadays, differential pressure sensors (DPSs) are utilized by the Building Management Systems (BMSs) to control and monitor the negative pressure in an NPR. This paper demonstrates a non-invasive and stealthy attack on NPRs by spoofing a DPS at its resonant frequency. Our contributions are: (1) We show that DPSs used in NPRs typically have resonant frequencies in the audible range. (2) We use this finding to design malicious music to create resonance in DPSs, resulting in an overshooting in the DPS's normal pressure readings. (3) We show how the resonance in DPSs can fool the BMSs so that the NPR turns its negative pressure to a positive one, causing a potential leak of deadly microbes from NPRs. We do experiments on 8 DPSs from 5 different manufacturers to evaluate their resonant frequencies considering the sampling tube length and find resonance in 6 DPSs. We can achieve a 2.5 Pa change in negative pressure from a ~7 cm distance when a sampling tube is not present and from a ~2.5 cm distance for a 1 m sampling tube length. We also introduce an interval-time variation approach for an adversarial control over the negative pressure and show that the forged pressure can be varied within 12 - 33 Pa. Our attack is also capable of attacking multiple NPRs simultaneously. Moreover, we demonstrate our attack at a real-world NPR located in an anonymous bioresearch facility, which is FDA approved and follows CDC guidelines. We also provide countermeasures to prevent the attack.

Autonomous vehicles use multiple sensors, large deep-learning models, and powerful hardware platforms to perceive the environment and navigate safely. In many contexts, some sensing modalities negatively impact perception while increasing energy consumption. We propose EcoFusion: an energy-aware sensor fusion approach that uses context to adapt the fusion method and reduce energy consumption without affecting perception performance. EcoFusion performs up to 9.5% better at object detection than existing fusion methods with approximately 60% less energy and 58% lower latency on the industry-standard Nvidia Drive PX2 hardware platform. We also propose several context-identification strategies, implement a joint optimization between energy and performance, and present scenario-specific results.

Detecting human stress levels and emotional states with physiological body-worn sensors is a complex task, but one with many health-related benefits. Robustness to sensor measurement noise and energy efficiency of low-power devices remain key challenges in stress detection. We propose SELF-CARE, a fully wrist-based method for stress detection that employs context-aware selective sensor fusion that dynamically adapts based on data from the sensors. Our method uses motion to determine the context of the system and learns to adjust the fused sensors accordingly, improving performance while maintaining energy efficiency. SELF-CARE obtains state-of-the-art performance across the publicly available WESAD dataset, achieving 86.34% and 94.12% accuracy for the 3-class and 2-class classification problems, respectively. Evaluation on real hardware shows that our approach achieves up to 2.2× (3-class) and 2.7× (2-class) energy efficiency compared to traditional sensor fusion.

Sensors are one of the most pervasive and integral components of today's safety-critical systems. Sensors serve as a bridge between physical quantities and connected systems. The connected systems with sensors blindly believe the sensor as there is no way to authenticate the signal coming from a sensor. This could be an entry point for an attacker. An attacker can inject a fake input signal along with the legitimate signal by using a suitable spoofing technique. As the sensor's transducer is not smart enough to differentiate between a fake and legitimate signal, the injected fake signal eventually can collapse the connected system. This type of attack is known as the transduction attack. Over the last decade, several works have been published to provide a defense against the transduction attack. However, the defenses are proposed on an ad-hoc basis; hence, they are not well-structured. Our work begins to fill this gap by providing a checklist that a defense technique should always follow to be considered as an ideal defense against the transduction attack. We name this checklist as the Golden reference of sensor defense. We provide insights on how this Golden reference can be achieved and argue that sensors should be redesigned from the transducer level to the sensor electronics level. We point out that only hardware or software modification is not enough; instead, a hardware/software (HW/SW) co-design approach is required to ride on this future roadmap to the robust and resilient sensor.

Due to the high performance and safety requirements of self-driving applications, the complexity of modern autonomous driving systems (ADS) has been growing, instigating the need for more sophisticated hardware which could add to the energy footprint of the ADS platform. Addressing this, edge computing is poised to encompass self-driving applications, enabling the compute-intensive autonomy-related tasks to be offloaded for processing at compute-capable edge servers. Nonetheless, the intricate hardware architecture of ADS platforms, in addition to the stringent robustness demands, set forth complications for task offloading which are unique to autonomous driving. Hence, we present ROMANUS, a methodology for robust and efficient task offloading for modular ADS platforms with multi-sensor processing pipelines. Our methodology entails two phases: (i) the introduction of efficient offloading points along the execution path of the involved deep learning models, and (ii) the implementation of a runtime solution based on Deep Reinforcement Learning to adapt the operating mode according to variations in the perceived road scene complexity, network connectivity, and server load. Experiments on the object detection use case demonstrated that our approach is 14.99% more energy-efficient than pure local execution while achieving a 77.06% reduction in risky behavior from a robust-agnostic offloading baseline.

Providing health monitoring devices with machine intelligence is important for enabling automatic mobile healthcare applications. However, this brings additional challenges due to the resource scarcity of these devices. This work introduces a neural contextual bandits based dynamic sensor selection methodology for high-performance and resource-efficient body-area networks to realize next generation mobile health monitoring devices. The methodology utilizes contextual bandits to select the most informative sensor combinations during runtime and ignore redundant data for decreasing transmission and computing power in a body area network (BAN). The proposed method has been validated using one of the most common health monitoring applications: cardiac activity monitoring. Solutions from our proposed method are compared against those from related works in terms of classification performance and energy while considering the communication energy consumption. Our final solutions could reach 78.8% AU-PRC on the PTB-XL ECG dataset for cardiac abnormality detection while decreasing the overall energy consumption and computational energy by 3.7 × and 4.3 ×, respectively.

Several papers have been published over the last ten years to provide a defense against intentional spoofing to sensors. However, these defenses would only work against those spoofing signals, which have a separate frequency from the original signal being measured. These defenses would not work if the spoofing attack signal (i) has a frequency equal to the frequency of original signals, (ii) has zero frequency, and (iii) is strong enough to drive the sensor output close to its saturation region. More specifically, these defenses are not designed for a magnetic spoofing attack on passive Hall sensors.

The recent developments in wearable devices and the Internet of Medical Things (IoMT) allow real-time monitoring and recording of electrocardiogram (ECG) signals. However, continuous monitoring of ECG signals is challenging in low-power wearable devices due to energy and memory constraints. Therefore, in this article, we present a novel and energy-efficient methodology for continuously monitoring the heart for low-power wearable devices. The proposed methodology is composed of three different layers: 1) a noise/artifact detection layer to grade the quality of the ECG signals; 2) a normal/abnormal beat classification layer to detect the anomalies in the ECG signals; and 3) an abnormal beat classification layer to detect diseases from ECG signals. Moreover, a distributed multioutput convolutional neural network (CNN) architecture is used to decrease the energy consumption and latency between the edge–fog/cloud. Our methodology reaches an accuracy of 99.2% on the well-known MIT-BIH Arrhythmia Data Set. Evaluation on real hardware shows that our methodology is suitable for devices having a minimum RAM of 32 kb. Moreover, the proposed methodology achieves$7\times $more energy efficiency compared to state-of-the-art works.

In autonomous vehicles (AVs), early warning systems rely on collision prediction to ensure occupant safety. However, state-of-the-art methods using deep convolutional networks either fail at modeling collisions or are too expensive/slow, making them less suitable for deployment on AV edge hardware. To address these limitations, we propose SG2VEC, a spatiotemporalscene-graphembedding methodology that uses the graph neural network (GNN) and long short-term memory (LSTM) layers to predict future collisions via visual scene perception. We demonstrate that SG2VEC predicts collisions 8.11% more accurately and 39.07% earlier than the state-of-the-art method on synthesized data sets, and 29.47% more accurately on a challenging real-world collision data set. We also show that SG2VEC is better than the state of the art at transferring knowledge from synthetic data sets to real-world driving data sets. Finally, we demonstrate that SG2VEC performs inference$9.3\times $faster with an 88.0% smaller model, 32.4% less power, and 92.8% less energy than the state-of-the-art method on the industry-standard Nvidia DRIVE PX 2 platform, making it more suitable for implementation on the edge.

Human activity recognition (HAR) is one of the key applications of health monitoring that requires continuous use of wearable devices to track daily activities. This article proposes an adaptive convolutional neural network for energy-efficient HAR (AHAR) suitable for low-power edge devices. Unlike traditional adaptive (early-exit) architecture that makes the early-exit decision based on classification confidence, AHAR proposes a novel adaptive architecture that uses an output block predictor to select a portion of the baseline architecture to use during the inference phase. The experimental results show that traditional adaptive architecture suffer from performance loss whereas our adaptive architecture provides similar or better performance as the baseline one while being energy efficient. We validate our methodology in classifying locomotion activities from two data sets—1) Opportunity and 2) w-HAR. Compared to the fog/cloud computing approaches for the Opportunity data set, our baseline and adaptive architectures show a comparable weighted F1 score of 91.79%, and 91.57%, respectively. For the w-HAR data set, our baseline and adaptive architectures outperform the state-of-the-art works with a weighted F1 score of 97.55%, and 97.64%, respectively. Evaluation on real hardware shows that our baseline architecture is significantly energy efficient ($422.38\times $less) and memory-efficient ($14.29\times $less) compared to the works on the Opportunity data set. For the w-HAR data set, our baseline architecture requires$2.04\times $less energy and$2.18\times $less memory compared to the state-of-the-art work. Moreover, experimental results show that our adaptive architecture is 12.32% (Opportunity) and 11.14% (w-HAR) energy efficient than our baseline while providing similar (Opportunity) or better (w-HAR) performance with no significant memory overhead.

Recently, road scene-graph representations used in conjunction with graph learning techniques have been shown to outperform state-of-the-art deep learning techniques in tasks including action classification, risk assessment, and collision prediction. To enable the exploration of applications of road scene-graph representations, we introduce roadscene2vec: an open-source tool for extracting and embedding road scene-graphs. The goal of roadscene2vec is to enable research into the applications and capabilities of road scene-graphs by providing tools for generating scene-graphs, graph learning models to create spatio-temporal scene-graph embeddings, and tools for visualizing and analyzing scene-graph-based methodologies. The capabilities of roadscene2vec include (i) customized scene-graph generation from either video clips or data from the CARLA simulator, (ii) multiple configurable spatio-temporal graph embedding models and baseline CNN-based models, (iii) built-in functionality for using graph and sequence embeddings for risk assessment and collision prediction applications, (iv) tools for evaluating transfer learning, and (v) utilities for visualizing scene-graphs and analyzing the explainability of graph learning models. We demonstrate the utility of roadscene2vec for these use cases with experimental results and qualitative evaluations for both graph learning models and CNN-based models. roadscene2vec is available at https://github.com/AICPS/roadscene2vec.

The articles in this special section are based on selected papers presented at the 2021 ACM/IEEE International Conference on Cyber-Physical Systems (ICCPS 2021), a premier single-track conference that promotes development of fundamental principles that underpin the integration of cyber and physical elements, as well as the development of technologies, tools, architectures, and infrastructure for the design and implementation of CPS. ICCPS 2021 focused on contributions related to smart and connected cities, autonomous CPS, verification and control, security and privacy, and human health and biomedical CPS.

A neuro-cognitive inspired architecture based on the Hierarchical Temporal Memory (HTM) is proposed for anomaly detection and simultaneous data prediction in real-time for smart grid$\mu$PMU data. The key technical idea is that the HTM learns asparse distributed temporal representationof sequential data that turns out to be very useful for anomaly detection and simultaneous data prediction in real-time. Our results show that the proposed HTM can predict anomalies within 83–90 percent accuracy for three different application profiles, namelyStandard, Reward Few False Positive, Reward Few False Negativefor two different datasets. We show that the HTM is competitive to five state-of-the-art algorithms for anomaly detection. Moreover, for the multi-step prediction in the online setting, the same HTM achieves a low 0.0001 normalized mean square error, a low negative log-likelihood score of 1.5 and is also competitive to six state-of-the-art prediction algorithms. We demonstrate that the same HTM model can be used forboth the tasksand can learn online in one-pass, in an unsupervised fashion and adapt to changing statistics.The other state-of-the-art algorithms are either less accurate or are limited to one of the tasks or cannot learn online in one-pass, and adapt to changing statistics.

With newer technologies, the embedded hardware and software in traditional vehicles and traffic control infrastructure continue to become more interconnected and more vulnerable. To assist in dealing with existing and potential vulnerabilities, we present a novel attack modeling methodology, taxonomy, and metrics (relative average waiting time, average network flow, impacts and rate of changes) to model, simulate, and meaningfully evaluate the security of Intelligent Transportation Systems. We implement our work in two different architectures: 1) Newell’s Car-Following Model with Bounded Acceleration (the BA-Newell Model) in Matlab and 2) Intelligent Driver Model in Veins. Our code is entirely open-sourced and will be maintained so that the ITS community may use it as a tool. We observe that the architectural-related metric values for sample attack simulation results are similar and transferable; where, for example, the rate of change values have range of average distances 1.8-3.5% for network flow impact and 3.3-9.6% for wait time impact.

There is considerable evidence that evaluating the subjective risk level of driving decisions can improve the safety of Autonomous Driving Systems (ADS) in both typical and complex driving scenarios. In this paper, we propose a novel data-driven approach that uses scene-graphs as intermediate representations for modeling the subjective risk of driving maneuvers. Our approach includes a Multi-Relation Graph Convolution Network, a Long-Short Term Memory Network, and attention layers. To train our model, we formulate subjective risk assessment as a supervised scene classification problem. We evaluate our model on both synthetic lane-changing datasets and real-driving datasets with various driving maneuvers. We show that our approach achieves a higher classification accuracy than the state-of-the-art approach on both large (96.4% vs. 91.2%) and small (91.8% vs. 71.2%) lane-changing synthesized datasets, illustrating that our approach can learn effectively even from small datasets. We also show that our model trained on a lane-changing synthesized dataset achieves an average accuracy of 87.8% when tested on a real-driving lane-changing dataset. In comparison, the state-of-the-art model trained on the same synthesized dataset only achieved 70.3% accuracy when tested on the real-driving dataset, showing that our approach can transfer knowledge more effectively. Moreover, we demonstrate that the addition of spatial and temporal attention layers improves our model’s performance and explainability. Finally, our results illustrate that our model can assess the risk of various driving maneuvers more accurately than the state-of-the-art model (86.5% vs. 58.4%, respectively).

The globalization of the integrated circuit (IC) supply chain has moved most of the design, fabrication, and testing process from a single trusted entity to various untrusted third-party entities worldwide. The risk of using untrusted third-Party Intellectual Property (3PIP) is the possibility for adversaries to insert malicious modifications known as Hardware Trojans (HTs). These HTs can compromise the integrity, deteriorate the performance, deny the service, and alter the functionality of the design. While numerous HT detection methods have been proposed in the literature, the crucial task of HT localization is overlooked. Moreover, a few existing HT localization methods have several weaknesses: reliance on a golden reference, inability to generalize for all types of HT, lack of scalability, low localization resolution, and manual feature engineering/property definition. To overcome their shortcomings, we propose a novel, golden reference-free HT localization method at the pre-silicon stage by leveraging graph convolutional network (GCN). In this work, we convert the circuit design into its intrinsic data structure, graph, and extract the node attributes. Afterward, the graph convolution performs automatic feature extraction for nodes to classify the nodes as Trojan or benign. Our approach is automated and does not burden the designer with manual code review. It locates the Trojan signals with 99.6% accuracy, 93.1%$F1$-score, and a false-positive rate below 0.009%.

Myocardial Infarction (MI) is a heart disease that damages the heart muscle and requires immediate treatment. Its silent and recurrent nature necessitates real-time continuous monitoring of patients. Nowadays, wearable devices are smart enough to perform on-device processing of heartbeat segments and report any irregularities in them. However, the small form factor of wearable devices imposes resource constraints and requires energy-efficient solutions to satisfy them. In this paper, we propose a design methodology to automate the design space exploration of neural network architectures for MI detection. This methodology incorporates Neural Architecture Search (NAS) using Multi-Objective Bayesian Optimization (MOBO) to render Pareto optimal architectural models. These models minimize both detection error and energy consumption on the target device. The design space is inspired by Binary Convolutional Neural Networks (BCNNs) suited for mobile health applications with limited resources. The models' performance is validated using the PTB diagnostic ECG database from PhysioNet. Moreover, energy-related measurements are directly obtained from the target device in a typical hardware-in-the-loop fashion. Finally, we benchmark our models against other related works. One model exceeds state-of-the-art accuracy on wearable devices (reaching 91.22%), whereas others trade off some accuracy to reduce their energy consumption (by a factor reaching 8.26x).

Edge-Cloud hierarchical systems employing intelligence through Deep Neural Networks (DNNs) endure the dilemma of workload distribution within them. Previous solutions proposed to distribute workloads at runtime according to the state of the surroundings, like the wireless conditions. However, such conditions are usually overlooked at design time. This paper addresses this issue for DNN architectural design by presenting a novel methodology, LENS, which administers multi-objective Neural Architecture Search (NAS) for two-tiered systems, where the performance objectives are refashioned to consider the wireless communication parameters. From our experimental search space, we demonstrate that LENS improves upon the traditional solution’s Pareto set by 76.47% and 75% with respect to the energy and latency metrics, respectively.

Aggressive time-to-market constraints and enormous hardware design and fabrication costs have pushed the semiconductor industry toward hardware Intellectual Properties (IP) core design. However, the globalization of the integrated circuits (IC) supply chain exposes IP providers to theft and illegal redistribution of IPs. Watermarking and fingerprinting are proposed to detect IP piracy. Nevertheless, they come with additional hardware overhead and cannot guarantee IP security as advanced attacks are reported to remove the watermark, forge, or bypass it. In this work, we propose a novel methodology, GNN4IP, to assess similarities between circuits and detect IP piracy. We model the hardware design as a graph and construct a graph neural network model to learn its behavior using the comprehensive dataset of register transfer level codes and gate-level netlists that we have gathered. GNN4IP detects IP piracy with 96% accuracy in our dataset and recognizes the original IP in its obfuscated version with 100% accuracy.

Design and fabrication outsourcing has made integrated circuits (IC) vulnerable to malicious modifications by third parties known as hardware Trojans (HT). Over the last decade, the use of side-channel measurements for detecting the malicious manipulation of the ICs has been extensively studied. However, the suggested approaches often suffer from three major limitations: 1) reliance on a trusted identical chip (i.e. golden chip), 2) untraceable footprints of subtle hardware Trojans which remain inactive during the testing phase, and 3) the need to identify the best discriminative features that can be used for separating side-channel signals coming from HT-free and HT-infected circuits. To overcome these shortcomings, we propose a novel neural network design (i.e. HTNet) and a feature extractor training methodology that can be used for HT detection in run time. We create a library of known hardware Trojans and collect electromagnetic and power side-channel signals for each case and train HTnet to learn the best discriminative features based on this library. Then, in the test time we fine tune HTnet to learn the behavior of the particular chip under test. We use HTnet followed by an anomaly detection mechanism in run-time to monitor the chip behavior and report malicious activities in the side-channel signals. We evaluate our methodology using TrustHub [15] benchmarks and show that HTnet can extract a robust set of features that can be used for HT-detection purpose.

A digital twin is the virtual replica of a physical system. Digital twins are useful because they provide models and data for design, production, operation, diagnostics, and autonomy of machines and products. Hence, the digital twin has been projected as the key enabler of the Visions of Industry 4.0. The digital twin concept has become increasingly sophisticated and capable over time, enabled by many technologies. In this paper, we propose the cognitive digital twin as the next stage of advancement of a digital twin that will help realize the vision of Industry 4.0. Cognition, which is inspired by advancements in cognitive science, machine learning, and artificial intelligence, will enable a digital twin to achieve some critical elements of cognition, e.g., attention (selective focusing), perception (forming useful representations of data), memory (encoding and retrieval of information and knowledge), etc. Our main thesis is that cognitive digital twins will allow enterprises to creatively, effectively, and efficiently exploit implicit knowledge drawn from the experience of existing manufacturing systems and enable the transfer of higher performance decisions and control and improve the performance across the enterprise (at scale). Finally, we present open questions and challenges to realize these capabilities in a digital twin.

The time to market pressure and resource constraints has pushed System-on-Chip (SoC) designers toward outsourcing the design and using third-party Intellectual Property (IP). It has created an opportunity for rogue entities in the Integrated Circuit (IC) supply chain to insert malicious circuits in the hardware design, known as Hardware Trojans (HT). HT detection is a major hardware security challenge, and its early discovery is crucial because postponing the removal of HT to late in design or after the fabrication process would be very expensive. Current works suffer from several shortcomings such as reliance on a golden HT-free reference, unable to identify all types of HTs or unknown ones, burdening the designer with the manual review of code, or scalability issues. To overcome these limitations, we propose GNN4TJ, a novel golden reference-free HT detection method in the register transfer level (RTL) based on Graph Neural Network (GNN). GNN4TJ represents the hardware design as its intrinsic data structure, a graph, and generates the data flow graphs for RTL codes. We utilize GNN to extract the features from DFG, learn the circuit's behavior, and identify the presence of HT, in a fully automated pipeline. We evaluate our model on a dataset that we create by expanding the Trusthub [1] HT benchmarks. The results demonstrate that GNN4TJ detects unknown HT with 97% recall (true positive rate) very fast in 21.1ms.

Deep Neural Network (DNN) models have been extensively developed by companies for a wide range of applications. The development of a customized DNN model with great performance requires costly investments, and its structure (layers and hyper-parameters) is considered intellectual property and holds immense value. However, in this paper, we found the model secret is vulnerable when a cloud-based FPGA accelerator executes it. We demonstrate an end-to-end attack based on remote power side-channel analysis and machine-learning-based secret inference against different DNN models. The evaluation result shows that an attacker can reconstruct the layer and hyper-parameter sequence at over 90% accuracy using our method, which can significantly reduce her model development workload. We believe the threat presented by our attack is tangible, and new defense mechanisms should be developed against this threat.

Equipping wearable devices with intelligence is essential for promoting mobile healthcare applications. However, challenges remain due to the resource limitations of these devices. In this work, we introduce EExNAS, a methodology for designing high-performance and resource-efficient dynamic Neural Architecture solutions for wearable devices. The methodology incorporates a platform-aware Neural Architecture Search (NAS) that accounts for energy efficiency at runtime through an Early-Exit (EEx) option. We showcase our methodology’s merit across 2 wearable applications, Myocardial Infarction (MI) detection and Human Activity Recognition (HAR). Solutions from EExNAS are compared against those from related works in terms of accuracy and performance. For MI detection, our final solutions with EEx capability could reach 98.54% accuracy on the PTB ECG dataset.

Wireless, battery-free Body Area Networks (BAN) enable reliable long-term health monitoring with minimal intervention, and have the potential to transform patient care via mobile health monitoring. Current approaches for achieving such battery-free networks are limited in the number, capability, and positioning of sensing nodes-this is related to constraints in power supply, data rate, and working distance requirements between the wireless power source and sensing nodes. Here, we investigate a Qi-based, near-field power transfer scheme that can effectively drive wireless, battery-free, multi-node and multi-sensor BAN over long distances. This consists of a single Qi power source (such as a cellphone), a detached/untethered Passive Intermediate Relay (PIR) (facilitates power transfer from a central Qi source to multiple nodes on the body), and finally individual/detached sensing nodes placed throughout the body. Alongside this power scheme we implement the star network topology of a Gazell protocol to enable the continuous connection of one host to many sensing nodes while minimizing data loss over long temporal periods. The high-power transmission capabilities of Qi enables wireless support for a multitude of sensors (up to 12), and sensing nodes (up to 6) with a single transmitter at long distances (60 cm) and a sample rate of 20 Hz. This scheme is studied both in-vitro and in-vivo on the body.

Eating activity recognition (EAR) plays an important role in ensuring healthy eating habits. Recent advancements of the Internet of Things (IoT) have bolstered automated EAR through various wearable edge devices. State-of-the-art work uses some offline trained classifiers at the fog device to recognize eating activities. However, the eating habits of a person change quite frequently and vary from person to person. Therefore, the classifiers should be updated continuously with new data to adapt to these changes and be personalized over time through online learning. To the best of our knowledge, no state-of-the-art work has addressed this issue so far. In this article, we propose an online learning methodology called human eating activity recognition (HEAR) by introducing an online update phase. We also design an algorithm to be used in the online update phase that provides approximate true labels for the new data. Moreover, we also design a wearable neckband as the edge device to capture eating activity data (Chewing, Swallowing, Talking, and Idle) in a lab environment. Through a detailed experimental evaluation on 12 users, we show that an online learned neural network (OLNN) classifier using our HEAR methodology performs better than any state-of-the-art offline trained classifier. We also demonstrate that our OLNN classifier is energy efficient compared to the competitive offline trained classifiers.

Pykg2vec is a Python library for learning the representations of the entities and relations in knowledge graphs. Pykg2vec's flexible and modular software architecture currently implements 25 state-of-the-art knowledge graph embedding algorithms, and is designed to easily incorporate new algorithms.The goal of pykg2vec is to provide a practical and educational platform to accelerate research in knowledge graph representation learning. Pykg2vec is built on top of PyTorch and Python's multiprocessing framework and provides modules for batch generation, Bayesian hyperparameter optimization, evaluation of KGE tasks, embedding, and result visualization. Pykg2vec is released under the MIT License and is also available in the Python Package Index (PyPI). The source code of pykg2vec is available at https://github.com/Sujit-O/pykg2vec.

In cyber-physical additive manufacturing systems, side-channel attacks have been used to reconstruct the G/M-code (which are instructions given to a manufacturing system) of 3D objects being produced. This method is effective for stealing intellectual property from an organization, through least expected means, during prototyping stage before the product goes through a large-scale fabrication and comes out in the market. However, an attacker can be far from being able to completely reconstruct the G/M-code due to lack of enough information leakage through the side-channels. In this paper, we propose a novel way to amplify the information leakage and thus boost the chances of recovery of G/M-code by surreptitiously altering the compiler. By using this compiler, an adversary may easily control various parameters to magnify the leakage of information from a 3D printer while still producing the desired object, thus remaining hidden from the authentic users. This type of attack may be implemented by strong attackers having access to the tool chain and seeking high level of stealth. We have implemented such a compiler and have demonstrated that it increases the success rate of recovering G/M-codes from the four side-channels (acoustic, power, vibration, and electromagnetic) by up to 39 percent compared to previously proposed attacks.

Autonomous vehicles (AV) are expected to revolutionize transportation and improve road safety significantly. However, these benefits do not come without cost; AVs require large Deep-Learning (DL) models and powerful hardware platforms to operate reliably in real-time, requiring between several hundred watts to one kilowatt of power. This power consumption can dramatically reduce vehicles’ driving range and affect emissions. To address this problem, we propose SAGE: a methodology for selectively offloading the key energy-consuming modules of DL architectures to the cloud to optimize edge, energy usage while meeting real-time latency constraints. Furthermore, we leverage Head Network Distillation (HND) to introduce efficient bottlenecks within the DL architecture in order to minimize the network overhead costs of offloading with almost no degradation in the model’s performance. We evaluate SAGE using an Nvidia Jetson TX2 and an industry-standard Nvidia Drive PX2 as the AV edge, devices and demonstrate that our offloading strategy is practical for a wide range of DL models and internet connection bandwidths on 3G, 4G LTE, and WiFi technologies. Compared to edge-only computation, SAGE reduces energy consumption by an average of 36.13% , 47.07% , and 55.66% for an AV with one low-resolution camera, one high-resolution camera, and three high-resolution cameras, respectively. SAGE also reduces upload data size by up to 98.40% compared to direct camera offloading.

Since 2007, the use of side-channel measurements for detecting Hardware Trojan (HT) has been extensively studied. However, the majority of works either rely on a golden chip, or they rely on methods that are not robust against subtle acceptable changes that would occur over the life-cycle of an integrated circuit (IC). In this paper, we propose using a brain-inspired architecture called Hierarchical Temporal Memory (HTM) for HT detection. Similar to the human brain, our proposed solution is resilient againstnaturalchanges that might happen in the side-channel measurements while being able to accurately detect abnormal behavior of the chip when the HT gets triggered. We use a self-referencing method for HT detection, which eliminates the need for the golden chip. The effectiveness of our approach is evaluated using TrustHub benchmarks, which shows 92.20% detection accuracy on average.

Deep Neural Network (DNN) models have been extensively developed by companies for a wide range of applications. The development of a customized DNN model with great performance requires costly investments, and its structure (layers and hyper-parameters) is considered intellectual property and holds immense value. However, in this paper, we found the model secret is vulnerable when a cloud-based FPGA accelerator executes it. We demonstrate an end-to-end attack based on remote power side-channel analysis and machine-learning-based secret inference against different DNN models. The evaluation result shows that an attacker can reconstruct the layer and hyper-parameter sequence at over 90% accuracy using our method, which can significantly reduce her model development workload. We believe the threat presented by our attack is tangible, and new defense mechanisms should be developed against this threat.

If machine failures can be detected preemptively, then maintenance and repairs can be performed more efficiently, reducing production costs. Many machine learning techniques for performing early failure detection using vibration data have been proposed; however, these methods are often power and data-hungry, susceptible to noise, and require large amounts of data preprocessing. Also, training is usually only performed once before inference, so they do not learn and adapt as the machine ages. In this article, we propose a method of performing online, real-time anomaly detection for predictive maintenance using hierarchical temporal memory (HTM). Inspired by the human neocortex, HTMs learn and adapt continuously and are robust to noise. Using the Numenta Anomaly Benchmark, we empirically demonstrate that our approach outperforms state-of-the-art algorithms at preemptively detecting real-world cases of bearing failures and simulated 3-D printer failures. Our approach achieves an average score of 64.71, surpassing state-of-the-art deep-learning (49.38) and statistical (61.06) methods.

Due to the sensitivity of the information exchanged in Vehicle to Vehicle (V2V) and Vehicle to Infrastructure (V2I) communication, generating secret keys is critical to secure these communications. As nature is open access, distributed keys are more vulnerable to attacks in the vehicular environment. Physical layer key generation methods using wireless channel characteristics show promise in preventing such attacks, generating keys independently, and removing the need for distribution. In this paper, we present a novel key generation approach in a real vehicular environment based on Channel State Information (CSI), including a new algorithm for key bit extraction. We implemented our algorithm using USRP B210 Software-Defined Radios (SDR) and the industry-standard V2X communication protocol: IEEE 802.11p. The proposed key generation protocol uses the CSI values of each sub-carrier as a source of randomness, from which bits are extracted using a new QAM demodulator quantizer (QAM-Dem-Quan). We compared our technique to state-of-the-art Received Signal Strength (RSS)-based approaches, and show that our method achieves better performance. Moreover, we reached a min-entropy of approximately 70% for the generated keys and a key generation rate of less than$150~\mu \text{s}$/key for key lengths ranging from 16 to 128 bits.

Keeping up with the rapid growth of Deep Learning (DL) research is a daunting task. While existing scientific literature search systems provide text search capabilities and can identify similar papers, gaining an in-depth understanding of a new approach or an application is much more complicated. Many publications leverage multiple modalities to convey their findings and spread their ideas - they include pseudocode, tables, images and diagrams in addition to text, and often make publicly accessible their implementations. It is important to be able to represent and query them as well. We utilize RDF Knowledge graphs (KGs) to represent multimodal information and enable expressive querying over modalities. In our demo we present an approach for extracting KGs from different modalities, namely text, architecture images and source code. We show how graph queries can be used to get insights into different facets (modalities) of a paper, and its associated code implementation. Our innovation lies in the multimodal nature of the KG we create. While our work is of direct interest to DL researchers and practitioners, our approaches can also be leveraged in other scientific domains.

Machine learning has been attracting strong interests in recent years. Numerous companies have invested great efforts and resources to develop customized deep-learning models, which are their key intellectual properties. In this work, we investigate to what extent the secret of deep-learning models can be inferred by attackers. In particular, we focus on the scenario that a model developer and an adversary share the same GPU when training a Deep Neural Network (DNN) model. We exploit the GPU side-channel based on context-switching penalties. This side-channel allows us to extract the fine-grained structural secret of a DNN model, including its layer composition and hyper-parameters. Leveraging this side-channel, we developed an attack prototype named MosConS, which applies LSTM-based inference models to identify the structural secret. Our evaluation of MosConS shows the structural information can be accurately recovered. Therefore, we believe new defense mechanisms should be developed to protect training against the GPU side-channel.

The deployment of Internet of Things (IoT) devices in cyber-physical applications has introduced a new set of vulnerabilities. The new security and reliability challenges require a holistic solution due to the cross-domain, cross-layer, and interdisciplinary nature of IoT systems. However, the majority of works presented in the literature primarily focus on the cyber aspect, including the network and application layers, and the physical layer is often overlooked.

Recent decades have observed the proliferation of sensors in embedded and cyber-physical systems (ECPSs). Sensors are an essential part of embedded and CPSs and serve as a bridge between physical quantities and connected systems. The tight coupling between sensors and systems enables many critical applications where decisions are taken by using the information from various sensors at different time-scales. This tight coupling opens the “Pandora's Box” of unknown threats that could come from very unconventional ways. An unconventional attack model could be to noninvasively attack sensors using forged spoofing signals and trigger unwanted behavior in connected systems. This paper introduces this type of new, strong, and unorthodox attack model and elaborates how important this will be in the near future when sensors will pervade our lives. Moreover, this paper presents a motivational example of a sensor-spoofing attack on Hall sensors in the context of smart grids to demonstrate the harmful consequences of this type of attack in ECPSs.

Cyber-Physical Production Systems (CPPS) will usher a new era of smart manufacturing. However, CPPS will be vulnerable to cross-domain attacks due to the interactions between the cyber and physical domains. To address the challenges of modeling cross-domain security in CPPS, we are proposing GAN-Sec, a novel conditional Generative Adversarial Network based modeling approach to abstract and estimate the relations between the cyber and physical domains. Using GAN-Sec, we are able to determine if various security requirements such as confidentiality, availability, and integrity are met. We provide a security analysis of an additive manufacturing system to demonstrate the applicability of GAN-Sec.

Cyber-Physical Systems (CPS) are growing with added complexity and functionality. Multidisciplinary interactions with physical systems are the major keys to CPS. However, sensors, actuators, controllers, and wireless communications are prone to attacks that compromise the system. Machine learning models have been utilized in controllers of automotive to learn, estimate, and provide the required intelligence in the control process. However, their estimation is also vulnerable to the attacks from physical or cyber domains. They have shown unreliable predictions against unknown biases resulted from the modeling. In this paper, we propose a novel control design using conditional generative adversarial networks that will enable a self-secured controller to capture the normal behavior of the control loop and the physical system, detect the anomaly, and recover from them. We experimented our novel control design on a self-secured BMS by driving a Nissan Leaf S on standard driving cycles while under various attacks. The performance of the design has been compared to the state-of-the-art; the self-secured BMS could detect the attacks with 83% accuracy and the recovery estimation error of 21% on average, which have improved by 28% and 8%, respectively.

Modern automotive Cyber-Physical Systems (CPS) are increasingly adopting wireless communications for Intra-Vehicular, Vehicle-to-Vehicle (V2V), and Vehicle-to-Infrastructure (V2I) protocols as a promising solution for challenges such as the wire harnessing problem, collision detection, and collision avoidance, traffic control, and environmental hazards. Regrettably, this new trend results in new security challenges that can put the safety and privacy of the automotive CPS and passengers at great risk. In addition, automotive wireless communication security is constrained by strict energy and performance limitations of electronic controller units and sensors. As a result, the key generation and management for secure automotive CPS wireless communication is an open research challenge. This article aims to help solve these security challenges by presenting a practical key generation technique based on the reciprocity and high spatial and temporal variation properties of the automotive wireless communication channel. Accompanying this technique is also a key length optimization algorithm to improve performance (in terms of time and energy) for safety-related applications constrained by small communication windows. To validate the practicality and effectiveness of our approach, we have conducted simulations alongside real-world experiments with vehicles and RC cars. Last, we demonstrate through simulations that we can generate keys with high security strength (keys with 67% min-entropy) with 20× reduction in code size overhead in comparison to the state-of-the-art security techniques.

Precision irrigation systems promise to bring significant improvement in resource efficiency and crop yield by providing analytics and smart tools for the growers. While significant amounts of data can be collected in a sensor-rich system, there are no rigorously designed models that can provide actionable intelligence to the user. This paper proposes the integration of circuit-inspired modeling of natural phenomena and man-made artifacts to generate end-to-end irrigation system circuit models. Such models can take advantage of existing circuit design and simulation tools that have been perfected over the past decades to efficiently process large input sets. We show that circuit-inspired models are indeed qualitatively sound and quantitatively accurate in capturing both natural phenomena and engineered physical irrigation systems.

Graphics Processing Units (GPUs) have been employed in embedded systems to handle increased amounts of computation and to satisfy the timing requirement. Due to the small feature size, chip aging and within-die parameter variations have been considered to be among the challenging problems for state-of-the-art processors, including GPUs. In order to deal with the process variation, several processors use chip-level guardbanding, which uses the lowest operating frequency that results in a significant chip-level performance drop. Other processors improve their performance efficiency through core-level guardbanding that may use a different operating frequency for each core. Existing aging management techniques are based on the chip-level guardbanding, which assigns the same number of instructions to the cores that have the same aging status. However, in the presence of the process variation, existing aging management techniques have a limitation in minimizing the aging effect because each core has a different amount of stress for the same number of instructions. In order to tackle this problem, we propose a low-overhead aging and process variation aware workload management technique for embedded GPUs. The proposed technique considers the process variation and the current aging status together, and assigns a different number of instructions to clusters to minimize the aging effect in the presence of process variation. Results show that our technique improves the GPU aging in over 95 percent of cases whereas the state-of-the-art compiler-based technique improves the GPU aging in 72.25 percent of cases. Moreover, compared to the compiler-based technique, our technique reduces the performance overhead by 40 percent while achieving almost the same GPU aging improvement.

In cyber-physical systems, due to the tight integration of the computational, communication, and physical components, most of the information in the cyber-domain manifests in terms of physical actions (such as motion, temperature change, etc.). This leads to the system being prone to physical-to-cyber domain attacks that affect the confidentiality. Physical actions are governed by energy flows, which may be observed. Some of these observable energy flows unintentionally leak information about the cyber-domain and hence are known as the side-channels. Side-channels such as acoustic, thermal, and power allow attackers to acquire the information without actually leveraging the vulnerability of the algorithms implemented in the system. As a case study, we have taken cyber-physical additive manufacturing systems (fused deposition modeling-based three-dimensional (3D) printer) to demonstrate how the acoustic side-channel can be used to breach the confidentiality of the system. In 3D printers, geometry, process, and machine information are the intellectual properties, which are stored in the cyber domain (G-code). We have designed an attack model that consists of digital signal processing, machine-learning algorithms, and context-based post processing to steal the intellectual property in the form of geometry details by reconstructing the G-code and thus the test objects. We have successfully reconstructed various test objects with an average axis prediction accuracy of 86% and an average length prediction error of 11.11%.

Electric Vehicles (EV) as a zero-emission means of transportation encounter challenges in battery design that cause a range anxieties for the drivers. Besides the electric motor, the Heating, Ventilation, and Air Conditioning (HVAC) system is another major contributor to the power consumption that may influence the EV battery lifetime and driving range. In the state-of-the-art methodologies for battery management systems, the battery performance is monitored and improved. While in the automotive climate control, the passenger’s thermal comfort is the main objective. Hence, the influence of the HVAC power on the battery behavior for the purpose of jointly optimized battery management and climate control has not been considered. In this article, we propose an automotive climate control methodology that is aware of the battery behavior and performance, while maintaining the passenger’s thermal comfort. In our methodology, battery parameters and cabin temperature are modeled and estimated, and the HVAC utilization is optimized and adjusted with respect to the electric motor and HVAC power requests. Therefore, the battery stress reduces, while the cabin temperature is maintained by predicting and optimizing the system states in the near-future. We have implemented our methodology and compared its performance to the state-of-the-art in terms of battery lifetime improvement and energy consumption reduction. We have also conducted experiments and analyses to explore multiple control window sizes, drive profiles, ambient temperatures, and modeling error rates in the methodology. It is shown that our battery-aware climate control can extend the battery lifetime by up to 13.2% and reduce the energy consumption by up to 14.4%.

Cyber-physical additive manufacturing systems consist of tight integration of cyber and physical domains. This union, however, induces new cross-domain vulnerabilities that pose unique security challenges. One of these challenges is preventing confidentiality breach, caused by physical-to-cyber domain attacks. In this form of attack, attackers utilize the side-channels (such as acoustics, power, electromagnetic emissions, and so on) in the physical-domain to estimate and steal cyber-domain data (such as G/M-codes). Since these emissions depend on the physical structure of the system, one way to minimize the information leakage is to modify the physical-domain. However, this process can be costly due to added hardware modification. Instead, we propose a novel methodology that allows the cyber-domain tools [such as computer aided-manufacturing (CAM)] to be aware of the existing information leakage. Then, we propose to change either machine process or product design parameters in the cyber-domain to minimize the information leakage. Our methodology aids the existing cyber-domain and physical-domain security solution by utilizing the cross-domain relationship. We have implemented our methodology in a fused-deposition modeling-based Cartesian additive manufacturing system. Our methodology achieves reduction of mutual information by 24.94% in acoustic side-channel, 32.91% in power side-channel, 32.29% in magnetic side-channel, and 55.65% in vibration side-channel. As a case study, to help understand the implication of mutual information drop, we have also presented the calculation of success rate and the reconstruction of the 3D object based on an attack model. For the given attack model, our leakage-aware CAM tool decreases the success rate of an attacker by 8.74% and obstructs the reconstruction of finer geometry details.

The interaction between the cyber domain and the physical domain components and processes can be leveraged to enhance the security of the cyber-physical system. In order to do so, we must first analyze various cyber domain and physical domain information flows, and characterize the relation between them using model functions. In this paper, we present a notion of cross-domain security of cyber-physical systems, whereby we present a security analysis framework that can be used for generating novel cross-domain attack models, attack detection methods, etc. We demonstrate how information flows such as discrete domain signal flows and continuous domain energy flows in the cyber and physical domain can be used to generate model functions using data-driven estimation, and use this model functions for performing various cross-domain security analysis. We also demonstrate the practical applicability of the cross-domain security analysis framework using the cyber-physical manufacturing system as a case study.

GPUs have been employed in the embedded systems to handle increased amount of computation and satisfy the timing requirement. Therefore, the lifetime of embedded GPUs is considered one of the most important aspects to ensure functional correctness over a long period of time. Moreover, existing state-of-the-art compiler-based GPU aging management techniques suffer from a considerable amount of performance overhead. In this paper, we propose a low-overhead aging-aware resource management technique. The proposed technique extends the behavior of the existing warp scheduler and the instruction dispatcher to cluster the computational cores and distribute instructions based on the aging information. Compared to when using the original applications, our technique improves the aging of the embedded GPU by 30% on average. In addition, compared to the state-of-the-art GPU aging management technique, our technique reduces the performance overhead by 16.4% on average while improving the aging by 3% on average.

A modern automotive design contains over a hundred microprocessors, several cyber-physical modules, connectivity to a variety of networks, and several hundred megabytes of software. The future is anticipated to see an even sharper rise in complexity of this electronics, with the imminence of driverless vehicles, the potential of connected automobiles within a few years, and work towards seamless integration of automobiles with smart cities and infrastructure systems. Security is a fundamental challenge in the design of automotive systems. Unfortunately, security considerations in automotive systems are complicated by two factors: (1) need for real-time mitigation against in-field threats; and (2) in-field configurability and extensibility of security features. This paper examines the trade-offs between security countermeasures, real-time requirements, and in-field configurability needs for modern automotive systems. We discuss the current state of the practice in automotive security architecture, as well as gaps and challenges that need to be addressed for a viable security solution in future.

Cyber-physical additive manufacturing systems consists of tight integration of cyber and physical domains. This results in new cross-domain vulnerabilities that poses unique security challenges. One of the challenges is preventing confidentiality breach due to physical-to-cyber domain attacks, where attackers can analyze various analog emissions from the side-channels to steal the cyber-domain information. This information theft is based on the idea that an attacker can accurately estimate the relation between the analog emissions (acoustics, power, electromagnetic emissions, etc.,) and the cyber-domain data (such as G-code). To obstruct this estimation process, it is crucial to quantize the relation between the analog emissions and the cyber-data, and use it as a metric to generate computer aided manufacturing tools, such as slicing and tool-path generation algorithms, that are aware of these information leakage through the side-channels. In this paper, we present a novel methodology that uses mutual information as a metric to quantize the information leakage from the side-channels, and demonstrates how various design variables (such as object orientation, nozzle velocity, etc.,) can be used in an optimization algorithm to minimize the information leakage. Our methodology integrates this leakage aware algorithms to the state-of-the-art slicing and tool-path generation algorithms and achieves 24.76% average drop in the information leakage through acoustic side-channel. To the best of our knowledge, this is the first work that demonstrates the idea of generating information leakage aware computer aided manufacturing tools for protecting the confidentiality of the manufacturing system.

The next industrial revolution will incorporate various enabling technologies. These technologies will make the product lifecycle of the manufacturing system efficient, decentralized, and well-connected. However, these technologies have various security issues, and when integrated in the product lifecycle of manufacturing systems can pose various challenges for maintaining the security requirements such as confidentiality, integrity, and availability. In this paper, we will present the various trends and advances in the security of the product lifecycle of the manufacturing system while highlighting the roles played by the major enabling components of Industry 4.0.

Controllers in cyber-physical systems integrate a design-time behavioral model of the system under design to improve their own quality. In the state-of-the-art control designs, behavioral models of other interacting neighbor systems are also integrated to form a centralized behavioral model and to enable a system-level optimization and control. Although this ideal embedded control design may result in pareto-optimal solutions, it is not scalable to larger number of systems. Moreover, the behavior of the multi-domain physical systems may be too complex for a control designer to model and may dynamically change at run time. In this paper, we propose a novel Adaptive and Cooperative Quality-Aware (ACQUA) control design which addresses these challenges. In this control design, an ACQUA-based controller for the system under design will monitor the quality of the neighbor systems to dynamically learn their behavior. Therefore, it can quickly adapt its control to cooperate with other neighbor controllers for improving the quality of not only itself, but also other neighbor systems. We apply ACQUA to design a cooperative controller for automotive navigation system, motor control unit, and battery management system in an electric vehicle. We use this automotive example to analyze the performance of the design. We show that by using our ACQUA control, we can reach up to 86% improvements achievable by an ideal embedded control design such that energy consumption reduces by 18% and battery capacity loss decreases by 12% compared to the state-of-the-art on average.

In power systems, the traditional, non-interactive, and manually controlled power grid has been transformed to a cyber-dominated smart grid. This cyber-physical integration has provided the smart grid with communication, monitoring, computation, and controlling capabilities to improve its reliability, energy efficiency, and flexibility. A microgrid is a localized and semi-autonomous group of smart energy systems that utilizes the above-mentioned capabilities to drive modern technologies such as electric vehicle charging, home energy management, and smart appliances. Design, upgrading, test, and verification of these microgrids can get too complicated to handle manually. The complexity is due to the wide range of solutions and components that are intended to address the microgrid problems. This article presents a novel Model-Based Design (MBD) methodology to model, co-simulate, design, and optimize microgrid and its multi-level controllers. This methodology helps in the design, optimization, and validation of a microgrid for a specific application. The application rules, requirements, and design-time constraints are met in the designed/optimized microgrid while the implementation cost is minimized. Based on our novel methodology, a design automation, co-simulation, and analysis tool, called GridMAT, is implemented. Our experiments have illustrated that implementing a hierarchical controller reduces the average power consumption by 8% and shifts the peak load for cost saving. Moreover, optimizing the microgrid design using our MBD methodology considering smart controllers has decreased the total implementation cost. Compared to the conventional methodology, the cost decreases by 14% and compared to the MBD methodology where smart controllers are not considered, it decreases by 5%.

Electric vehicles (EVs) have been considered as a solution to the environmental issues caused by transportation, such as air pollution and greenhouse gas emission. However, limited energy capacity, scarce EV supercharging stations, and long recharging time have brought anxiety to drivers who use EVs as their main mean of transportation. Furthermore, EV owners need to deal with a huge battery replacement cost when the battery capacity degrades. Yet in-house EV chargers affect the pattern of the power grid load, which is not favorable to the utilities. The driving route, departure/arrival time of daily trips, and electricity price influence the EV energy consumption, battery lifetime, electricity cost, and EV charger load on the power grid. The EV driving range and battery lifetime issues have been addressed by battery management systems and route optimization methodologies. However, in this article, we are proposing an optimized charge and drive management (OCDM) methodology that selects the optimal driving route, schedules daily trips, and optimizes the EV charging process while considering the driver’s timing preference. Our methodology will improve the EV driving range, extend the battery lifetime, reduce the recharging cost, and diminish the influence of EV chargers on the power grid. The performance of our methodology compared to the state of the art have been analyzed by experimenting on three benchmark EVs and three drivers. Our methodology has decreased EV energy consumption by 27%, improved the battery lifetime by 24.8%, reduced the electricity cost by 35%, and diminished the power grid peak load by 17% while increasing less than 20 minutes of daily driving time. Moreover, the scalability of our OCDM methodology for different parameters (e.g., time resolution and multiday cycles) in terms of execution time and memory usage has been analyzed.

Major challenges of driving range and battery lifetime in Electric Vehicles (EV) have been addressed by designing more efficient power electronics, advanced embedded hardware, and sophisticated embedded software. Besides the electric motor in EVs, Heating, Ventilation, and Air Conditioning (HVAC) has been seen as a significant contributor to the EV power consumption. The main responsibility of automotive climate controls has been to control the HVAC system in order to maintain the passengers' thermal comfort. However, the HVAC power consumption and its dynamic behavior may influence the battery lifetime and driving range significantly. Therefore, modeling and analyzing the HVAC system and its thermodynamic behavior may benefit the control designers to integrate the HVAC control and optimization into Battery Management Systems (BMS) for better battery lifetime and driving range. In this paper, the EV architecture, HVAC system dynamic behavior, and battery characteristics are explained and modeled. Automotive climate controls (e.g. battery lifetime-aware automotive climate control) and the benefits gained by system modeling and estimation for different conditions in terms of battery lifetime and driving range are illustrated. Moreover, present and future challenges regarding the HVAC system and control design are explained.

Additive Manufacturing (AM) uses Cyber-Physical Systems (CPS) (e.g., 3D Printers) that are vulnerable to kinetic cyber-attacks. Kinetic cyber-attacks cause physical damage to the system from the cyber domain. In AM, kinetic cyber-attacks are realized by introducing flaws in the design of the 3D objects. These flaws may eventually compromise the structural integrity of the printed objects. In CPS, researchers have designed various attack detection method to detect the attacks on the integrity of the system. However, in AM, attack detection method is in its infancy. Moreover, analog emissions (such as acoustics, electromagnetic emissions, etc.) from the side-channels of AM have not been fully considered as a parameter for attack detection. To aid the security research in AM, this paper presents a novel attack detection method that is able to detect zero-day kinetic cyber-attacks on AM by identifying anomalous analog emissions which arise as an outcome of the attack. This is achieved by statistically estimating functions that map the relation between the analog emissions and the corresponding cyber domain data (such as G-code) to model the behavior of the system. Our method has been tested to detect potential zero-day kinetic cyber-attacks in fused deposition modeling based AM. These attacks can physically manifest to change various parameters of the 3D object, such as speed, dimension, and movement axis. Accuracy, defined as the capability of our method to detect the range of variations introduced to these parameters as a result of kinetic cyber-attacks, is 77.45%.

By introducing microgrids, energy management is required to control the power generation and consumption for residential, industrial, and commercial domains, e.g., in residential microgrids and homes. Energy management may also help us to reach zero net energy (ZNE) for the residential domain. Improvement in technology, cost, and feature size has enabled devices everywhere, to be connected and interactive, as it is called Internet of Things (IoT). The increasing complexity and data, due to the growing number of devices like sensors and actuators, require powerful computing resources, which may be provided by cloud computing. However, scalability has become the potential issue in cloud computing. In this paper, fog computing is introduced as a novel platform for energy management. The scalability, adaptability, and open source software/hardware featured in the proposed platform enable the user to implement the energy management with the customized control-as-services, while minimizing the implementation cost and time-to-market. To demonstrate the energy management-as-a-service over fog computing platform in different domains, two prototypes of home energy management (HEM) and microgrid-level energy management have been implemented and experimented.

Graphics processing units (GPUs) have been employed in the critical path of applications in embedded systems due to the GPUs' programmability, high-performance, and low power consumption. State-of-the-art GPUs have the capability to process multiple GPU workloads concurrently. Moreover, GPU-based embedded systems have been considered to be essential because of the increased number of throughput-oriented applications and system events. However, existing application scheduling frameworks on a GPU do not have enough flexibility to handle the dynamic behavior of the event-driven applications. This is because in the existing scheduling frameworks: only temporal preemption is considered and one application occupies the GPU at a time. In order to tackle these problems, we propose a novel run-time scheduling framework that considers both temporal and spatial preemptions concurrently. We demonstrate the capability and novelty of our framework compared to the existing scheduling frameworks with realistic benchmark applications and with different execution scenarios. Experimental results show that our scheduling framework is able to guarantee up to 1.37 times as many applications compared to other scheduling frameworks. Moreover, the total amount of timing violation is decreased by up to 54.57%.

Bridging disparate realms of physical and cyber system components requires models and methods that enable rapid evaluation of design alternatives in cyber-physical systems (CPS). The diverse intellectual traditions of physical and mathematical sciences makes this task exceptionally hard. This paper seeks to explore potential solutions by examining specific examples of CPS applications in automobiles and smart buildings. Both smart buildings and automobiles are complex systems with embedded knowledge across several domains. We present our experiences with development of CPS applications to illustrate the challenges that arise when expertise across domains is integrated into the system, and show that creation of models, abstractions, and architectures that address these challenges are key to next generation CPS applications.

Electric Vehicle (EV) optimization involves stringent constraints on driving range and battery lifetime. Sophisticated embedded systems and huge number of computing resources have enabled researchers to implement advanced Battery Management Systems (BMS) for optimizing the driving range and battery lifetime. However, the Heating, Ventilation, and Air Conditioning (HVAC) control and BMS have not been considered together in this optimization. This paper presents a novel automotive climate control methodology that manages the HVAC power consumption to improve the battery lifetime and driving range. Our experiments demonstrate that the HVAC consumption is considerable and flexible in an EV which significantly influences the driving range and battery lifetime. Hence, this influence on the above-mentioned constraints has been modeled and analyzed precisely, then it has been considered thoroughly in the EV optimization process. Our methodology provides significant improvement in battery lifetime (on average 14%) and average power consumption (on average 39% reduction) compared to the state-of-the-art methodologies.

This paper presents a novel Model-Based Design (MBD) approach and associated tool-chain for the Time-Triggered Real-time Embedded (TTRE)1systems. Our tool-chain automatically synthesizes software for manufacturing TTRE systems. A Software-In-the-Loop Simulation (SILS) framework integrated into our tool-chain helps to reduce the design iterations. Using a manufacturing robot-arm use-case, we validate our tool-chain and demonstrate a 39× improvement in the Quality-of-Control (QoC) when compared to the state-of-the-art approach [3]. Our auto-generated scheduler meets all the hard real-time constraints (zero deadline misses) for a given TTRE system when compared to the scheduler (e.g., 145 deadline misses for a CPU utilization of 95%) presented in [1], [25]. Moreover, compared to the traditional MBD approach for the TTRE systems where many design iterations are required [4], our approach and tool-chain can generate high quality and accurate tasks with an associated scheduler in a single design iteration.

Security is one of the major challenges for Cyber-Physical Systems (CPS) design. Identifying flaws as early as possible in the CPS design saves time and money [6]; between 5× to 10× less expensive [7] than finding them during the detailed design stages. This paper makes a case for finding cybersecurity flaws as early as possible. Not only for the temporal and cost benefits, but more importantly, for the integrity of the system once in operation. We introduce a security-aware functional modeling methodology, supported by simulation to validate the robustness of the system in the presence of attacks and countermeasures. Our ideas are implemented in a design automation tool in Amesim and Matlab/Simulink. We use an automotive use-case as an example to validate the methodology and the tool.

This work presents a novel Model-Based Design (MBD) approach and associated tool-chain for the IEC 61131-3 specific Programmable Logical Controllers (PLC) [2]. Our tool-chain automatically synthesizes software for manufacturing Time-Triggered Real-time Embedded (TTRE) systems. A Software-In-the-Loop Simulation (SILS) framework integrated into our tool-chain helps to reduce the design iterations. Using a manufacturing robot-arm use-case, we validate our tool-chain and demonstrate a 39x improvement in the Quality-of-Control (QoC) when compared to the state-of-the-art approach [1]. Our auto-generated scheduler meets all the hard real-time constraints (zero deadline misses) for a given TTRE system when compared to the scheduler (e.g., 145 deadline misses for a CPU utilization of 95%) presented in [13].

Recently, Electric Vehicles (EVs) have been considered as new paradigm of transportation in order to solve environmental concerns, e.g. air pollution. However, EVs pose new challenges regarding their Battery LifeTime (BLT), energy consumption, and energy costs related to battery charging. The EV power consumption may be estimated by having the route information and the EV specifications. Also, by having the battery characteristics, the battery capacity consumption and the BLT may be estimated for each route. In this paper, we propose a driving management which uses the above-mentioned information in order to optimize the driving route by being aware of the EV energy consumption, energy cost, and BLT. Our proposed driving management extends the BLT by 16.8% and reduces the energy consumption by 11.9% and energy cost by 12.6% on average, by selecting the optimized route instead of the fastest route.

Cyber-Physical Energy Systems (CPES) are an amalgamation of both power gird technology, and the intelligent communication and co-ordination between the supply and the demand side through distributed embedded computing. Through this combination, CPES are intended to deliver power efficiently, reliably, and economically. The design and development work needed to either implement a new power grid network or upgrade a traditional power grid to a CPES-compliant one is both challenging and time consuming due to the heterogeneous nature of the associated components/subsystems. The Model Based Design (MBD) methodology has been widely seen as a promising solution to address the associated design challenges of creating a CPES. In this paper, we demonstrate a MBD method and its associated tool for the purpose of designing and validating various control algorithms for a residential microgrid. Our presented co-simulation engine GridMat is a MATLAB/Simulink toolbox; the purpose of it is to co-simulate the power systems modeled in GridLAB-D as well as the control algorithms that are modeled in Simulink. We have presented various use cases to demonstrate how different levels of control algorithms may be developed, simulated, debugged, and analyzed by using our GridMat toolbox for a residential mi-crogrid.

This paper presents our multi-year experience in the development of a Functional Modeling Compiler (FMC), a new model-based design tool for the development of multi-disciplinary automotive cyber-physical systems. We show how system-level simulation models suitable for design-space exploration of complex architectures can be synthesized from functional specifications to test and validate the interactions between ECUs, control algorithms, and the multi-physics.

GPU architecture has traditionally been used in graphics application because of its enormous computing capability. Moreover, GPU architecture has also been used for general purpose computing in these days. Most of the current scheduling frameworks that are developed to handle GPGPU workload operate sequentially. This is problematic since this sequential approach may not be scalable for real-time systems, which is a consequence of the approach's inability to support preemption. We propose a novel scheduling framework that provides real-time support for the GPGPU platform. In contrast to existing frameworks, our proposed framework considers both concurrent execution of applications on the GPU and mapping between streaming multiprocessors and thread blocks. By considering both concurrent execution and mapping, our framework is able to guarantee timing up to 6.4 times as many applications compared to TimeGraph [9] and Global EDF [5]. In addition, our experimental applications use up to 20% less power under our scheduling framework compared to [5], [9].

A novel design methodology, associated algorithms, and tools for the design of complex automotive cyber-physical systems are presented. Rather than supporting the critical path where most resources are spent, we preemptively target the concept design phase that determines 75% of a vehicle's cost. In our methodology, the marriage of systems engineering principles with high-level synthesis techniques results in a Functional Modeling Compiler capable of generating high-fidelity simulation models for the design space exploration and validation of multiple cyber-physical (ECUs+Physics) vehicle architectures. Using real-world automotive use-cases, we demonstrate how functional models capturing cyber-physical aspects are synthesized into high-fidelity simulation models.

In industrial cyber-physical systems (CPS)1, the ability of a system to react quicker to its inputs by just a few milliseconds can be translated to billions of dollars in additional profit over just a few years of uninterrupted operation. Therefore, it is important to reduce the cycle time of industrial CPS applications not only for the economical benefits but also for waste minimization, energy reduction, and safer working environments. In this paper, we present a novel method to reduce the execution time of CPS applications through a holistic software/hardware method that enables automatic parallelization of standardized industrial automation languages and their execution in multi-core processors. Through a realistic CPS, we demonstrate that parallel execution reduces the cycle time of the application and increases the life-cycle through better utilization of the mechanical, electrical, and computing resources.

This paper presents an overview of the the novel technologies that we are experiencing today in the automation industries. We present the opportunities and challenges of having tightly coupled collaborative networks of embedded systems for controlling complex physical processes. Our objective is to motivate the targeted design automation community to tackle some of the grand challenges in the area of such a distributed, intelligent, and collaborative embedded computing platform.

Networsk-on-chip (NoCs) have emerged as a promising on-chip interconnect for future multi/many-core architectures as NoCs are able to scale communication links with the growing number of cores. State-of-the-art NoC designs rely mainly on a static network configuration using fixed routing algorithms and buffer placements. These approaches are not effective in dealing with hard-to-predict system behavior, for instance due to user behavior or varying workloads, since in order for static NoCs to cover these scenarios, they would have to be designed for worst case scenarios. In this paper, we address these problems with a runtime adaptive network-on-chip (AdNoC). Focusing on the architecture-level adaptation, we present an adaptive route allocation algorithm which provides a required level of QoS (guaranteed bandwidth) coupled with an adaptive buffer assignment scheme which reassigns buffer blocks on-demand. Furthermore, the adaptivity requires a comprehensive, hardly intrusive, runtime observability infrastructure, i.e., using monitoring components, in order to gather data on the system state. The area overhead introduced by the adaptive scheme can be traded off against the flexibility gained. Moreover, the area overhead is also reduced by resource multiplexing due to the on-demand buffer assignment at each output port (we achieved on an average 42% buffer saving in our experiments). We demonstrate the advantage by using various digital media applications and compare our approach to the state-of-the-art static NoC architectures e.g., Xpipe, QNoC, and Æthereal.

3D multi-core architectures are seen to provide increased transistor density, reduced power consumption, and improved performance through wire length reduction. However, 3D suffers from increased power density, which exacerbates thermal hotspots. In this paper, we present a novel 3D multi-core architecture that reduces processor activity on the die distant to the heat sink and a core-level dynamic thermal management technique based on the architectural adaptation, e.g. dynamically adapting core-resources depending on diverse application requirements and thermal behavior. The proposed thermal management technique synergistically combines the benefits of the architectural adaptation supported by our 3D multi-core architecture with dynamic voltage and frequency scaling. Our proposed technique provides 19.4% (maximum 24.4%, minimum 15.5%) improvement in the instruction throughput compared to the state-of-the-art thermal management techniques [4, 5] applied to the thermal-aware 3D processor architecture without considering run-time adaptation [10].

Multi core architectures that are built to reap performance and energy efficiency benefits from the parallel execution of applications often employ runtime adaptive techniques in order to achieve, among others, load balancing, dynamic thermal management, and to enhance the reliability of a system. Typically, such runtime adaptation in the system level requires the ability to quickly and consistently migrate a task from one core to another. For distributed memory architectures, the policy for transferring the task context between source and destination cores is of vital importance to the performance and to the successful operation of the system. As its performance is negatively correlated with the communication overhead, energy consumption and the dissipated heat, task migration needs to be runtime adaptive to account for the system load, chip temperature, or battery capacity. This work presents a novel context-aware runtime adaptive task migration mechanism (CARAT) that reduces the task migration latency by 93.12%, 97.03% and 100% compared to three state-of-the-art mechanisms and allows to control the maximum migration delay and the performance overhead tradeoff at runtime. This novel mechanism is built on an in-depth analysis of the memory access behavior of several multi-media and robotic embedded-systems applications.

With the rapid increase of wireless applications, Cognitive Radio (CR) has been considered as a promising concept to improve the utilization of limited radio spectrum resources for future wireless communications and mobile computing. Because of the dynamic behavior of Cognitive Radio Networks (CRNs), secure communication in CRNs becomes more critical than for other conventional Wireless networks. So, in this paper we propose a trust-based security solution for CRNs. Trust is calculated from the requesting node depending on various communication attributes and the evaluated trust is compared with the trust threshold value. Depending on the resultant decision, the requested service is available to the requesting user. We prove the security of our proposed scheme in terms of security analysis.

Reliability concerns associated with upcoming technology nodes coupled with unpredictable system scenarios resulting from increasingly complex systems require considering runtime adaptivity in all possible parts of future on-chip systems. We are presenting a novel configurable link which can change its supported bandwidth on-demand at runtime (2X-Links) for an adaptive on-chip communication architecture. We have evaluated our results using real-time multi-media and the E3S application benchmark suits. Our 2X-Links provide a higher throughput of up to 36%, with an average throughput increase of 21.3%, compared to the Normal-Full-Duplex-Links [12], [14], [17], [20] and keep performance-related guarantees with as low as 50% of the Normal-Full-Duplex-Links capacity. Our simulation shows when some links fail, the NoC with 2X-Links can recover from these faults with an average probability of 82.2% whereas these faults would be fatal for the Normal-Full-Duplex-Links.

A growing challenge in embedded system design is coping with increasing power densities resulting from packing more and more transistors onto a small die area, which in turn transform into thermal hotspots. In the current late silicon era silicon structures have become more susceptible to transient faults and aging effects resulting from these thermal hotspots. In this paper we present an agent-based power distribution approach (TAPE) which aims to balance the power consumption of a multi/many-core architecture in a pro-active manner. By further taking the system's thermal state into consideration when distributing the power throughout the chip, TAPE is able to noticeably reduce the peak temperature. In our simulation we provide a fair comparison with the state-of-the-art approaches HRTM [19] and PDTM [9] using the MiBench benchmark suite [18]. When running multiple applications simultaneously on a multi/many-core architecture, we are able to achieve an 11.23% decrease in peak temperature compared to the approach that uses no thermal management [14]. At the same time we reduce the execution time (i.e. we increase the performance of the applications) by 44.2% and reduce the energy consumption by 44.4% compared to PDTM [9]. We also show that our approach exhibits higher scalability, requiring 11.9 times less communication overhead in an architecture with 96 cores compared to the state-of-the-art approaches.

Design-time decisions can often only cover certain scenarios and fail in efficiency when hard-to-predict system scenarios occur. This drives the development of run-time adaptive systems. To the best of our knowledge, we are presenting the first scheme for a run-time application mapping in a distributed manner using agents targeting for adaptive NoC-based heterogeneous multi-processor systems. Our approach reduces the overall traffic produced to collect the current state of the system (monitoring-traffic), needed for runtime mapping, compared to a centralized mapping scheme. In our experiment, we obtain 10.7 times lower monitoring traffic compared to the centralized mapping scheme proposed in [8] for a 64 x 64 NoC. Our proposed scheme also requires less execution cycles compared to a non-clustered centralized approach. We achieve on an average 7.1 times lower computational effort for the mapping algorithm compared to the simple nearest-neighbor (NN) heuristics proposed in [6] in a 64 x 32 NoC. We demonstrate the advantage of our scheme by means of a robot application and a set of multimedia applications and compare it to the state-of-the-art run-time mapping schemes proposed in [6, 8, 19].

We present a novel methodology for design space exploration using a two-steps scheme to optimize the number of virtual channel buffers (buffers take the premier share of the router in a NoC) used to implement logical channels multiplexed across the physical channel in a router output port for QoS supported on-chip communication. In the first step, the number of virtual channels is minimized during the mapping of tasks to the NoC at the design time of a system on chip (SoC)for which we use a swarm intelligence-based ant colony optimization (ACO) algorithm. In the second step, a probabilistic approach based on the traffic model of the application is used to further minimize the number of virtual channels. We achieve on average 90.2% reduction in the number of virtual channels compared to a fixed state- of-the-art (i.e. QNoC) allocation for the E3S embedded application benchmark suit. The reduction depends on the designer and the QoS parameter, and it is dependent on the specific application driven traffic model. We demonstrate our design space exploration by means of a complete robot application and also extend our exploration by evaluating the E3S embedded application benchmark suit.

Hard-to-predict system behavior and/or reliability issues resulting from migrating to new technology nodes requires considering runtime adaptivity in future on-chip systems. Run-time observability is a prerequisite for runtime adaptivity as it is providing necessary system information gathered on-the-fly. We are presenting the first comprehensive runtime observability infrastructure for an adaptive network on chip architecture which is flexible (e.g. in choosing the routing path), hardly intrusive, and requires little additional overhead (around 0.7% of the total link bandwidth). The hardware overhead is negligible, too, and is in fact less than the hardware savings due to resource multiplexing capabilities that are achieved through runtime observability/adaptivity. As an example, our on-demand buffer assignment scheme increases the buffer utilization and decreases the overall buffer requirements by an average of 42% (the buffer area amounts to about 60% of the entire router area [19]) in our case study analysis compared to a fixed buffer assignment scheme [7]. Our runtime observability on an average also increases the connection success rate by 62% compared to the case without runtime observability for the applications from the E3S benchmark suite [6]. We show the advantages obtained through runtime observability and compare with state-of-the art communication-centric designs.

We propose a scheme to reduce the number of virtual channel buffers used to implement logical channels multiplexed across the physical channel in a router output port for QoS supported on-chip communication. The number of virtual channels is minimized during the mapping of tasks to the NoC during the design time of a System on Chip (SoC) and a swarm intelligence-based Ant Colony Optimization (ACO) algorithm is used. We achieve on average 88% reduction in the number of virtual channels compared to a fixed allocation for the E3S embedded application benchmark suit and a collection of existing real-time embedded applications.

During run-time varying workloads and/or constraints in embedded systems require run-time adaptivity to provide a high degree of efficiency during any operation mode/scenario. Design time decisions can often only cover certain scenarios and fail in efficiency when hard-to-predict system scenarios occur. We are presenting the first approach of an adaptive on-chip communication scheme. It provides an adaptive routing/path allocation algorithm to meet a required level of QoS (guaranteed bandwidth). In our architecture adaptive runtime links are established by re-assigning buffer blocks ondemand. This adaptive buffer allocation scheme increases the buffer utilization and decreases the overall buffer use on an average of 42% in our case study analysis compared to a fixed buffer assignment strategy. The area overhead introduced by the adaptive scheme can be traded-off against the flexibility in order to select an available path and on-demand buffer allocation. We demonstrate the advantage by using various real world digital media applications and compare our approach to the state-ofthe- art static on-chip communication schemes.

Current Application Specific Instruction set Processor (ASIP) design methodologies are mostly based on iterative architecture exploration that uses Architecture Description Languages (ADLs) and retargetable software development tools. However, for improved design efficiency, additional pre-architecture exploration tools are required to help narrow-down the huge design space and making coarsegrained Instruction Set Architecture (ISA) decisions before detailed ADL modeling. Extensive application code profiling is the key in such early design stages. Based on a novel code instrumentation technology, we present a microprofiling approach that fills the current gap between source-level and instruction-level profilers and combines their advantages w.r.t. speed and accuracy. We show how the microprofiler is embedded into an advanced ASIP design flow and justify its use in a case study to design an MP3 decoder ASIP.