Vishwa Goudar

Public Physical Unclonable Functions

A physical unclonable function (PUF) is an integrated circuit (IC) that serves as a hardware security primitive due to its complexity and the unpredictability between its outputs and the applied inputs. PUFs have received a great deal of research interest and significant commercial activity. Public PUFs (PPUFs) address the crucial PUF limitation of being a secret-key technology. To some extent, the first generation of PPUFs are similar to SIMulation Possible, but Laborious (SIMPL) systems and one-time hardware pads, and employ the time gap between direct execution and simulation. The second PPUF generation employs both process variation and device aging which results in matched devices that are excessively difficult to replicate. The third generation leaves the analog domain and employs reconfigurability and device aging to produce digital PPUFs. We survey representative PPUF architectures, related public protocols and trusted information flows, and related testing issues. We conclude by identifying the most important, challenging, and open PPUF-related problems.

Fault-Tolerant and Low-Power Sampling Schedules for Localized BASNs

Recent advances in the scope of wearable devices and networks make body area sensor networks (BASNs) an extremely attractive tool to the fields of mobile and tele-health, owing to the range of medical applications they can serve and the diagnostic richness of patient data they can offer. However, for BASNs to achieve true ubiquity, they must be scalable in their support of automated patient data collection, making usability and reliability key considerations. Its designers must wrestle with the tradeoff between usability, hindered by device intrusiveness into the behaviors it measures, and lifetime, enhanced by large power supplies and expensive, sturdy components. Furthermore, the validity and reliability of the collected data are paramount. In this paper, we consider these issues in the context of localized multi-sensory wearable networks and present a method to generate low-power sampling schedules that are resilient to sensor faults while achieving high diagnostic fidelity. We jointly formulate this as a power-constrained sampling problem wherein the number of sensors sampled per epoch are limited, and, a fault tolerant scheduling problem wherein the sampling scheme offers enough redundancy to endure up to a predefined number of sensor faults while maintaining diagnostic accuracy. This formulation is based on, 1) the localized scope of BASNs that engenders strong spatio-temporal interactions in the samples, and, 2) the periodic nature of human behaviors measured. We present our algorithm in the context of gait diagnostics derived from a foot plantar pressure measurement platform and illustrate its performance based on real datasets collected by it.

Optimizing the Output of a Human-Powered Energy Harvesting System With Miniaturization and Integrated Control

We propose a novel harvesting technology to inconspicuously transduce mechanical energy from human foot-strikes and explore its configuration and control toward optimized energy output. Dielectric elastomers (DEs) are high-energy density, soft material that electrostatically transduce mechanical energy. These properties enable increased energy-transduction efficiency without sacrificing user comfort, if configured and controlled properly. We expose key statistical properties of human gait, which show that an array of miniaturized harvesters across the foot-sole will improve energy output. Further, the gait properties naturally yield a closed-loop control strategy to individually control harvesters in the array in a manner that maximizes net energy output. We propose statistical techniques that guide the configuration and control of the harvester array, and evaluate system behavior from detailed analytical and empirical models of DE behavior. System evaluations based on experimentally collected foot pressure data sets show that the proposed system can achieve up to 120 mJ per foot-strike.

A model of order-selectivity based on dynamic changes in the balance of excitation and inhibition produced by short-term synaptic plasticity

Determining the order of sensory events separated by a few hundred milliseconds is critical to many forms of sensory processing, including vocalization and speech discrimination. Although many experimental studies have recorded from auditory order-sensitive and order-selective neurons, the underlying mechanisms are poorly understood. Here we demonstrate that universal properties of cortical synapses-short-term synaptic plasticity of excitatory and inhibitory synapses-are well suited for the generation of order-selective neural responses. Using computational models of canonical disynaptic circuits, we show that the dynamic changes in the balance of excitation and inhibition imposed by short-term plasticity lead to the generation of order-selective responses. Parametric analyses predict that among the forms of short-term plasticity expressed at excitatory-to-excitatory, excitatory-to-inhibitory, and inhibitory-to-excitatory synapses, the single most important contributor to order-selectivity is the paired-pulse depression of inhibitory postsynaptic potentials (IPSPs). A topographic model of the auditory cortex that incorporates short-term plasticity accounts for both context-dependent suppression and enhancement in response to paired tones. Together these results provide a framework to account for an important computational problem based on ubiquitous synaptic properties that did not yet have a clearly established computational function. Additionally, these studies suggest that disynaptic circuits represent a fundamental computational unit that is capable of processing both spatial and temporal information.

Power constrained sensor sample selection for improved form factor and lifetime in localized BANs

Wearable sensing systems are paving the way for significant advances in diagnosis, preventative healthcare and tele-healthcare, by facilitating a variety of wireless health applications for medical signal and diagnostic monitoring and assessment. However, the considerable spatial and temporal sampling for multiple sensed modalities that enable these applications, also makes them power hungry, requiring large, heavy power supplies, and leading to a tradeoff between usability and lifetime. We propose a sampling algorithm to overcome this trade-off by capitalizing on the spatio-temporal redundancy inherent to Body Area Networks owing to their localized nature, as well as, assessing sample relevance based on its contribution to the predicted diagnostic(s). Our approach improves energy-efficiency and raises contextual sample quality, by tackling sample selection simultaneously in the spatial and temporal domains, yielding improved diagnostic accuracy under power-constraints. We present our algorithm in the context of diagnostics gleaned from a foot plantar pressure measurement platform and illustrate its efficacy based on real datasets collected by the platform.

Spatiotemporal assignment of energy harvesters on a self-sustaining medical shoe

We present a new method for spatiotemporal assignment and scheduling of energy harvesters on a medical shoe tasked with measuring gait diagnostics. While prior work exists on the application of dielectric elastomers (DEs) for energy scavenging on shoes, current literature does not address the issues of placement and timing of these harvesters, nor does it address integration into existing sensing systems. We solve these issues and present a self-sustaining medical shoe that harvests energy from human ambulation while simultaneously measuring gait characteristics most relevant to medical diagnosis.

Useful dynamic regimes emerge in recurrent networks

The internal dynamics of recurrent cortical circuits is crucial to brain function. We now learn that simply increasing the strengths of recurrent connections shifts neural dynamics to a potentially powerful computational regime.

Energy-efficient sampling schedules for Body Area Networks

We present a novel sampling method to overcome the tradeoff between sensing fidelity and energy-efficiency in the context of localized sensor arrays used by Body Area Networks (BANs). Prior research has tackled this tradeoff as a coverage problem, wherein a subset of sensors must cover the sensor field. Instead, we formulate it as a power-constrained sampling problem, limiting the number of samples taken per epoch to produce schedules with enhanced coverage and energy savings. This formulation capitalizes on the periodic nature and the strong spatio-temporal interactions that are innate to BAN sensor samples. Our algorithm produces schedules with over 170% in energy savings with increased sensor coverage that yields up to a 41% improvement in diagnostic estimates.

Dielectric Elastomer Generators for foot plantar pressure based energy scavenging

Parasitic energy scavenging from human-generated vibrations with piezoelectric materials has long been studied in contrast to electromagnetic or conventional electrostatic transducers. Dielectric Elastomers (DEs) are now gaining notice as low-cost electrostatic transducers with high energy densities. However, their transduction mechanism is more intricate. DE Generators (DEGs) are functionally variable capacitors, which require fine-grained control of their charging cycles in order to maximize the energy transduced. Based on a detailed DEG model that incorporates an effective method to time the charge cycles, we contrast the energy scavenged from shoe strikes by DEGs that are virtually embedded into the shoe sole, to similar piezoelectric generators. This comparison for a plantar pressure dataset of a walking subject demonstrates a multiple order-of-magnitude improvement in harvested energy.

Encoding sensory and motor patterns as time-invariant trajectories in recurrent neural networks

Much of the information the brain processes and stores is temporal in nature—a spoken word or a handwritten signature, for example, is defined by how it unfolds in time. However, it remains unclear how neural circuits encode complex time-varying patterns. We show that by tuning the weights of a recurrent neural network (RNN), it can recognize and then transcribe spoken digits. The model elucidates how neural dynamics in cortical networks may resolve three fundamental challenges: first, encode multiple time-varying sensory and motor patterns as stable neural trajectories; second, generalize across relevant spatial features; third, identify the same stimuli played at different speeds—we show that this temporal invariance emerges because the recurrent dynamics generate neural trajectories with appropriately modulated angular velocities. Together our results generate testable predictions as to how recurrent networks may use different mechanisms to generalize across the relevant spatial and temporal features of complex time-varying stimuli.