This PDF file contains the front matter associated with SPIE Proceedings Volume 9119 including the Title Page, Copyright information, Table of Contents, Introduction, and Conference Committee listing.

I will provide a high level walk-through for three computational approaches derived from Nature. First, evolutionary computation implements what we may call the “mother of all adaptive processes.” Some variants on the basic algorithms will be sketched and some lessons I have gleaned from three decades of working with EC will be covered. Then neural networks, computational approaches that have long been studied as possible ways to make “thinking machines”, an old dream of man’s, and based upon the only known existing example of intelligence. Then, a little overview of attempts to combine these two approaches that some hope will allow us to evolve machines we could never hand-craft. Finally, I will touch on artificial immune systems, Nature’s highly sophisticated defense mechanism, that has emerged in two major stages, the innate and the adaptive immune systems. This technology is finding applications in the cyber security world.

Neural network design has utilized flexible nonlinear processes which can mimic biological systems, but has suffered from a lack of traceability in the resulting network. Graphical probabilistic models ground network design in probabilistic reasoning, but the restrictions reduce the expressive capability of each node making network designs complex. The ability to model coupled random variables using the calculus of nonextensive statistical mechanics provides a neural node design incorporating nonlinear coupling between input states while maintaining the rigor of probabilistic reasoning. A generalization of Bayes rule using the coupled product enables a single node to model correlation between hundreds of random variables. A coupled Markov random field is designed for the inferencing and classification of UCI’s MLR ‘Multiple Features Data Set’ such that thousands of linear correlation parameters can be replaced with a single coupling parameter with just a (3%, 4%) reduction in (classification, inference) performance.

Machine learning is continuing to gain popularity due to its ability to solve problems that are difficult to model using conventional computer programming logic. Much of the current and past work has focused on algorithm development, data processing, and optimization. Lately, a subset of research has emerged which explores issues related to security. This research is gaining traction as systems employing these methods are being applied to both secure and adversarial environments. One of machine learning’s biggest benefits, its data-driven versus logic-driven approach, is also a weakness if the data on which the models rely are corrupted. Adversaries could maliciously influence systems which address drift and data distribution changes using re-training and online learning. Our work is focused on exploring the resilience of various machine learning algorithms to these data-driven attacks. In this paper, we present our initial findings using Monte Carlo simulations, and statistical analysis, to explore the maximal achievable shift to a classification model, as well as the required amount of control over the data.

We introduce the motivations behind AHaH computing, an emerging new form of adaptive computing with many applications in machine learning. We then present a technology stack or specification describing the multiple levels of abstraction and specialization needed to support AHaH computing.

It is now accepted that the traditional von Neumann architecture, with processor and memory separation, is ill suited to process parallel data streams which a mammalian brain can efficiently handle. Moreover, researchers now envision computing architectures which enable cognitive processing of massive amounts of data by identifying spatio-temporal relationships in real-time and solving complex pattern recognition problems. Memristor cross-point arrays, integrated with standard CMOS technology, are expected to result in massively parallel and low-power Neuromorphic computing architectures. Recently, significant progress has been made in spiking neural networks (SNN) which emulate data processing in the cortical brain. These architectures comprise of a dense network of neurons and the synapses formed between the axons and dendrites. Further, unsupervised or supervised competitive learning schemes are being investigated for global training of the network. In contrast to a software implementation, hardware realization of these networks requires massive circuit overhead for addressing and individually updating network weights. Instead, we employ bio-inspired learning rules such as the spike-timing-dependent plasticity (STDP) to efficiently update the network weights locally. To realize SNNs on a chip, we propose to use densely integrating mixed-signal integrate-andfire neurons (IFNs) and cross-point arrays of memristors in back-end-of-the-line (BEOL) of CMOS chips. Novel IFN circuits have been designed to drive memristive synapses in parallel while maintaining overall power efficiency (<1 pJ/spike/synapse), even at spike rate greater than 10 MHz. We present circuit design details and simulation results of the IFN with memristor synapses, its response to incoming spike trains and STDP learning characterization.

Side-channel attacks (SCAs), specifically differential power attacks (DPA), target hardware vulnerabilities of cryptosystems. Next generation computing systems, integrated with emerging technologies such as RRAM, offer unique opportunities to mitigate DPAs with their inherent device characteristics. We propose two different approaches to mitigate DPA attacks using memristive hardware. The first approach, obfuscates the power profile using dual RRAM modules. The power profile stays almost uniform for any given data access. This is achieved by realizing a memory and its complementary module in RRAM hardware. Balancing logic, which ensures the parallel access, is implemented in CMOS. The power consumed with the dual-RRAM balancing is an order lower than the corresponding pure CMOS implementation. The second exploratory approach, uses a novel neuromemristive architecture to compute an AES transformation and mitigate DPAs. Both the proposed approaches were tested on a 128-bit AES algorithm. A customized simulation framework, integrating CAD tools, is developed to mount the DPA attacks. In both the designs, the attack mounted on the baseline architectures (CMOS only) was successful and full key was recovered. However, DPA attacks mounted on the dual RRAM modules and neuromemristive hardware modules of an AES cryptoprocessor yielded no successful keys, demonstrating their resiliency to DPA attacks.

The advent of nanoscale metal-insulator-metal (MIM) structures with memristive properties has given birth to a new generation of hardware neural networks based on CMOS/memristor integration (CMHNNs). The advantage of the CMHNN paradigm compared to a pure CMOS approach lies in the multi-faceted functionality of memristive devices: They can efficiently store neural network configurations (weights and activation function parameters) via non-volatile, quasi-analog resistance states. They also provide high-density interconnects between neurons when integrated into 2-D and 3-D crossbar architectures. In this work, we explore the combination of CMHNN classifiers with manifold learning to reduce the dimensionality of CMHNN inputs. This allows the size of the CMHNN to be reduced significantly (by ≈ 97%). We tested the proposed system using the Caltech101 database and were able to achieve classification accuracies within ≈ 1:5% of those produced by a traditional support vector machine.

A fully parallel, silicon-based artificial neural network (CM1K) built on zero instruction set computer (ZISC) technology was used for change detection and object identification in video data. Fundamental pattern recognition capabilities were demonstrated with reduced neuron numbers utilizing only a few, or in some cases one, neuron per category. This simplified approach was used to validate the utility of few neuron networks for use in applications that necessitate severe size, weight, and power (SWaP) restrictions. The limited resource requirements and massively parallel nature of hardware-based artificial neural networks (ANNs) make them superior to many software approaches in resource limited systems, such as micro-UAVs, mobile sensor platforms, and pocket-sized robots.

This paper presents a decentralized multi-robot motion control strategy to facilitate a multi-robot system, comprised of collaborative mobile robots coordinated through wireless communications, to form and maintain desired wireless communication coverage in a realistic environment with unstable wireless signaling condition. A fuzzy neural network controller is proposed for each robot to maintain the wireless link quality with its neighbors. The controller is trained through reinforcement learning to establish the relationship between the wireless link quality and robot motion decision, via consecutive interactions between the controller and environment. The tuned fuzzy neural network controller is applied to a multi-robot deployment process to form and maintain desired wireless communication coverage. The effectiveness of the proposed control scheme is verified through simulations under different wireless signal propagation conditions.

To solve problems when an analytical solution is not available, more and more bio-inspired computation techniques have been applied in the last years. Thus, an efficient algorithm is the Genetic Algorithm (GA), which imitates the biological evolution process, finding the solution by the mechanism of “natural selection”, where the strong has higher chances to survive. A genetic algorithm is an iterative procedure which operates on a population of individuals called "chromosomes" or "possible solutions" (usually represented by a binary code). GA performs several processes with the population individuals to produce a new population, like in the biological evolution. To provide a high speed solution, pipelined based FPGA hardware implementations are used, with a nstages pipeline for a n-phases genetic algorithm. The FPGA pipeline implementations are constraints by the different execution time of each stage and by the FPGA chip resources. To minimize these difficulties, we propose a bio-inspired technique to modify the crossover step by using non identical twins. Thus two of the chosen chromosomes (parents) will build up two new chromosomes (children) not only one as in classical GA. We analyze the contribution of this method to reduce the execution time in the asynchronous and synchronous pipelines and also the possibility to a cheaper FPGA implementation, by using smaller populations. The full hardware architecture for a FPGA implementation to our target ALTERA development card is presented and analyzed.

We present a proof-of-concept of a lightweight and low-power network intrusion detection system (NIDS) using a commercially available neural network chip. Such a system is well-suited to the increasing deployment of low-power devices with ubiquitous internet connectivity. Our proposal makes use of previous work on extracting a feature vector from network packets using a histogram of hashed n-grams. The commercially available CogniMem CM1K device implements a version of the Restricted Coulomb Energy neural network classifier, which was used to classify the resulting feature vectors at high speed and low power. In this paper, we describe our feature extraction technique for network packets and the classification algorithm used by the CM1K chip, and present initial classification results on a fabricated test set. Despite the generality of the RCE algorithm and our ‘plug-in’ approach to the classification task, with no fine-tuning of the hardware to our problem domain, we obtain surprisingly good classification results even on highly imbalanced and restricted training sets.

Current computer systems are dumb automatons, and their blind execution of instructions makes them open to attack. Their inability to reason means that they don't consider the larger, constantly changing context outside their immediate inputs. Their nearsightedness is particularly dangerous because, in our complex systems, it is difficult to prevent all exploitable situations. Additionally, the lack of autonomous oversight of our systems means they are unable to fight through attacks. Keeping adversaries completely out of systems may be an unreasonable expectation, and our systems need to adapt to attacks and other disruptions to achieve their objectives. What is needed is an autonomous controller within the computer system that can sense the state of the system and reason about that state. In this paper, we present Self-Awareness Through Predictive Abstraction Modeling (SATPAM). SATPAM uses prediction to learn abstractions that allow it to recognize the right events at the right level of detail. These abstractions allow SATPAM to break the world into small, relatively independent, pieces that allow employment of existing reasoning methods. SATPAM goes beyond classification-based machine learning and statistical anomaly detection to be able to reason about the system, and SATPAM's knowledge representation and reasoning is more like that of a human. For example, humans intuitively know that the color of a car is not relevant to any mechanical problem, and SATPAM provides a plausible method whereby a machine can acquire such reasoning patterns. In this paper, we present the initial experimental results using SATPAM.

Unfortunately, there is no metric, nor set of metrics, that are both general enough to encompass all possible types of applications yet specific enough to capture the application and attack specific details. As a result we are left with ad-hoc methods for generating evaluations of the security of our systems. Current state of the art methods for evaluating the security of systems include penetration testing and cyber evaluation tests. For these evaluations, security professionals simulate an attack from malicious outsiders and malicious insiders. These evaluations are very productive and are able to discover potential vulnerabilities resulting from improper system configuration, hardware and software flaws, or operational weaknesses. We therefore propose the index of cyber integrity (ICI), which is modeled after the index of biological integrity (IBI) to provide a holistic measure of the health of a system under test in a cyber-environment. The ICI provides a broad base measure through a collection of application and system specific metrics. In this paper, following the example of the IBI, we demonstrate how a multi-metric index may be used as a holistic measure of the health of a system under test in a cyber-environment.

What is presented here is a sequence of evolving concepts for network intrusion detection. These concepts start with neuromorphic structures for XOR-based signature matching and conclude with computationally based network intrusion detection system with an autonomous structuring algorithm. There is evidence that neuromorphic computation for network intrusion detection is fractal in nature under certain conditions. Specifically, the neural structure can take fractal form when simple neural structuring is autonomous. A neural structure is fractal by definition when its fractal dimension exceeds the synaptic matrix dimension. The authors introduce the use of fractal dimension of the neuromorphic structure as a factor in the autonomous restructuring feedback loop.

Much of the traffic in modern computer networks is conducted between clients and servers, rather than client-toclient. As a result, servers represent a high-value target for collection and analysis of network traffic. As they reside at a single network location (i.e. IP/MAC address) for long periods of time. Servers present a static target for surveillance, and a unique opportunity to observe the network traffic. Although servers present a heightened value for attackers, the security community as a whole has shifted more towards protecting clients in recent years leaving a gap in coverage. In addition, servers typically remain active on networks for years, potentially decades. This paper builds on previous work that demonstrated a proof of concept leveraging existing technology for increasing attacker workload. Here we present our clean slate approach to increasing attacker workload through a novel hypervisor and micro-kernel, utilizing next generation virtualization technology to create synthetic diversity of the server's presence including the hardware components.

As datasets with time-series records, such as computer logs or financial transactions, grow larger, indexing solutions are needed that can efficiently filter out irrelevant records while retrieving most of relevant ones. These methods must capture essential temporal properties present in the data, and provide a scalable way to generate the index and update it as the new records are presented. Current time-series analysis and indexing methods are insufficient, because the fixed features they rely on capture only limited periodicity in time-series data and become brittle when the time-series encode heterogeneous temporal behaviors and are noisy and incomplete. New indexing solutions must not only cluster the data, but also infer the meaningful characteristics and present them to the users to improve their understanding of the data. In this paper, we develop an indexing procedure based on typical latent behaviors within the time series. Our method (1) converts the data to a quantized format, (2) learns identifying behaviors generating the data, and (3) produces an index for the time series based on these behaviors. The method is found to outperform standard approaches to time series indexing in terms of recall and precision for varying degrees of data noise.

In this paper, we present a model for processing distributed relational data across multiple autonomous heterogeneous computing resources in environments with limited control, resource failures, and communication bottlenecks. Our model exploits dependencies in the data to enable collaborative distributed querying in noisy data. The collaboration policy for computational resources is efficiently constructed from the belief propagation algorithm. To scale to large data sizes, we employ a combination of priority-based filtering, incremental processing, and communication compression techniques. Our solution achieved high accuracy of analysis results and orders of magnitude improvements in computation time compared to the centralized graph matching solution.

Trust is an important concept for machine intelligence and is not consistent across many applications. In this paper, we seek to understand trust from a variety of factors: humans, sensors, communications, intelligence processing algorithms and human-machine displays of information. In modeling the various aspects of trust, we provide an example from machine intelligence that supports the various attributes of measuring trust such as sensor accuracy, communication timeliness, machine processing confidence, and display throughput to convey the various attributes that support user acceptance of machine intelligence results. The example used is fusing video and text whereby an analyst needs trust information in the identified imagery track. We use the proportional conflict redistribution rule as an information fusion technique that handles conflicting data from trusted and mistrusted sources. The discussion of the many forms of trust explored in the paper seeks to provide a systems-level design perspective for information fusion trust quantification.

In environmental monitoring, small differences are expected in data allowing the discrimination of harmless from dangerous, e.g. discern passing trucks from an earth quake. For conventional data analysis it is a challenging problem to identify these signals in noisy data. Electronic data processing consumes extra time, or much energy when processed in parallel. This is not suitable for quick automatic decisions nor useful in nearly autonomous systems with very low energy budget. This paper presents methods inspired by human vision, applied to quickly determine the content of signals collected by a monitoring system. It is shown that the system is capable to distinguish different signals, using a simple filter-set.

A modified version of the intelligent water drop algorithm for performing planning for air and ground robots based on telemetry provided by satellites has been created. The IWD algorithm works by simulating the flow of water drops in a stream-network, dynamically adapting drop and network characteristics. This paper presents the base IWD algorithm, a simplified version of the algorithm (SIWD) and a derivative of this simplified version that has been adapted and applied to planning air and ground robot paths based upon orbital (for aerial) and aerial (for ground) imagery. An analysis of the performance of the algorithm is presented.

Detecting and identifying targets in unmanned aerial vehicle (UAV) images and videos have been challenging problems due to various types of image distortion. Moreover, the significantly high processing overhead of existing image/video processing techniques and the limited computing resources available on UAVs force most of the processing tasks to be performed by the ground control station (GCS) in an off-line manner. In order to achieve fast and autonomous target identification on UAVs, it is thus imperative to investigate novel processing paradigms that can fulfill the real-time processing requirements, while fitting the size, weight, and power (SWaP) constrained environment. In this paper, we present a new autonomous target identification approach on UAVs, leveraging the emerging neuromorphic hardware which is capable of massively parallel pattern recognition processing and demands only a limited level of power consumption. A proof-of-concept prototype was developed based on a micro-UAV platform (Parrot AR Drone) and the CogniMemTMneural network chip, for processing the video data acquired from a UAV camera on the y. The aim of this study was to demonstrate the feasibility and potential of incorporating emerging neuromorphic hardware into next-generation UAVs and their superior performance and power advantages towards the real-time, autonomous target tracking.