Gaussian data augmentation has been used to improve the robustness of a deep leaning neural netwo... more Gaussian data augmentation has been used to improve the robustness of a deep leaning neural network model against adversarial attacks. This paper presents a design and evaluation of a Keras implementation of injecting colored noise during the training phase of deep networks. The evaluation was made using empirical robustness; a metric from the IBM adversarial robustness toolbox (ART). Results show robustness improvement over the base model when injecting black, pink and brown noise; and even better performance when combinations of ensemble noises are injected. Keywords—Colored Noise Augmentation, Adversarial Example, Machine Learning Robustness, Security of Deep Learning.
This paper considers a mobile ad hoc network (MANET) as a socio-technical system with human-machi... more This paper considers a mobile ad hoc network (MANET) as a socio-technical system with human-machine- characteristics. From this perspective, it is argued that a cognitive MANET will need certain human behaviours if it is to acquire intelligence reminiscence of a human social system. To achieve this, the spatial relationships among the MANET nodes require algorithms to process the node actions in order to create robust system-level behaviours from those relationships. We develop a technique known as KANA (a Knowledge Action Network for Agents) for use in developing such intelligence behaviours. KANA has two primary knowledge elements: a network- and a fuzzy matrix- representation of the nodes and edges in a MANET. The KANA algorithms have been used to model and simulate MANET behaviors in order to discover its sociometric properties
: The paper presents a preliminary result of using cognitive- and behavior-based modeling framewo... more : The paper presents a preliminary result of using cognitive- and behavior-based modeling framework to simulate a network of MANETs (Mobile Adhoc NETworks) as intelligent agents in a tactical battlefield. We demonstrate the efficacies of an agent-based modeling and the need for developing formal methods for multi-agent simulations from a system of systems (SoS) perspective. The development of our model framework informs the need to describe behaviors and relationships of actors and objects in the context of a mission space, and 2) to provide a foundation for modeling agent behaviors in a way that is plausible with respect to human behavior, a specially from the standpoint of human-system interactions.
2016 IEEE International Symposium on Systems Engineering (ISSE), 2016
Requirements, stakeholders' visions, drive any systems engineering (SE) process. But in desig... more Requirements, stakeholders' visions, drive any systems engineering (SE) process. But in design process practices, there is a gap often present between stakeholders' visions and requirements representation. This limits the impact that the stakeholders can provide input to the SE process. Then, the process may end up with errors or provide a product which is unwanted. We proposed a design methodology called the Responsive and Formal Design (RFD) process that directly involves the stakeholders input at each level of the requirement elicitation, shows its effect on the total system performance, and integrates high level requirements with domain specific considerations and verifies formally. It consists of a set of levels of representaion. Each level represents a set of requirements and its associated models, simulations, and the relationship between them. The levels of representation are related with refinement and abstraction relations. The refinement helps to make clear the connection with parametric considerations. As levels of RFD proceed towards refinement, the design process becomes a local or discipline specific activity, though always with a global perspective. Following our framework, this paper presents the implementation of the refinement process. Each level of the RFD process has its own level of granularity. This is true for the model, logical representation, and simulation of the requirements. In this paper, we define the pair of functions, refinement and abstraction, that exist between two models (system and component models in table form called classifications) and their logical representations (called theories) with different levels of granularity. We also show how high level requirements are interpreted in the refined level. We use an example of data from three small satellites, whose goal is to image the auroral ovals around Earth's magnetic poles, to demonstrate our development.
Barwise and Seligman stated the first principle of information flow: ”Information flow results fr... more Barwise and Seligman stated the first principle of information flow: ”Information flow results from regularities in the distributed system.” They represent a distributed system in terms of a classification consisting of a set of objects or tokens to be classified, a set of types used to classify tokens, and a binary relation between tokens and types that tells one which tokens are classified as being of which types. We aim to further this investigation and proceed with a dynamic or evolving system instead of a static system. We claim that a classification is a snapshot of a distributed system at a given moment or context. We then aim to answer the question posed by an evolving context. As the context or configuration changes, how do regularities evolve. This paper is a continuation of an investigation we started in [2], where we initiated how to capture a dynamism of information flow with a Kripke structure. Here we develop the same procedure with colored Petri net(CPN). We first ex...
We present a design framework for small-satellite systems that ensures that (1) each satellite ha... more We present a design framework for small-satellite systems that ensures that (1) each satellite has a consistent theory to infer new information from information it perceives and (2) the theory for the entire system is consistent so that a satellite can infer new information from information communicated to it. This research contributes to our Reliable and Formal Design (RFD) process, which strives for designs that are ”correct by construction” by introducing formal methods early. Our framework uses Barwise’s channel theory, founded on category theory, and allied work in situation semantics and situation theory. Each satellite has a ”classification”, which consists of tokens (e.g., observed situations) and types (e.g., situation features) and a binary relation classifying tokens with types. The core of a system of classifications is a category-theoretic construct that amalgamates the several classifications. We show how to derive the theory associated with a classification and the th...
Abstract. The firing rule of Petri nets relies on a residuation operation for the commutative mon... more Abstract. The firing rule of Petri nets relies on a residuation operation for the commutative monoid of natural numbers. We identify a class of residuated commutative monoids, called Petri algebras, for which one can mimic the token game of Petri nets to define the behaviour of generalized Petri net whose flow relation and place contents are valued in such algebraic structures. We show that Petri algebras coincide with the positive cones of lattice-ordered commutative groups and constitute the subvariety of the (duals of) residuated lattices generated by the commutative monoid of natural numbers. We introduce a class of nets, termed lexicographic Petri nets, that are associated with the positive cones of the lexicographic powers of the additive group of real numbers. This class of nets is universal in the sense that any net associated with some Petri algebras can be simulated by a lexicographic Petri net. All the classical decidable properties of Petri nets however are undecidable o...
In this paper, we develop analytical models to calibrate trust and distrust in cyberspace as a fu... more In this paper, we develop analytical models to calibrate trust and distrust in cyberspace as a function of risk. Trust and distrust are viewed as economic gains and losses, while risk represents adverse influencing factors such as privacy loss or virus attacks. The models are constructive and reflect the behaviors often used by people when they attempt to intrinsically calculate whether to trust or distrust cyberspace activities such as on-line banking or on-line shopping under available information on different subjective and objective risk conditions. The analytical models, although simplified, are able to capture the tacit and inherent human behaviors attributed to the user’s perception of risk and trust, as well as represent the economic utility derived from cyberspace. Sample distributions of the trust and distrust as functions risk are derived.
This paper begins by considering requirements for verification tools for NextGen, the US National... more This paper begins by considering requirements for verification tools for NextGen, the US National Airspace System under development. Category theory and channel are reviewed as well as their contribution to computer science, and how these theories can be useful in the development of NextGen is outlined. Category theory is a branch of mathematics that can provide a unified presentation of much of abstract mathematics, and channel theory uses concepts from category theory to provide an abstract and rigorous account of how information about some system component can carry information about other components. The second half of this paper introduces a notion of change into channel theory, which involves augmenting channel theory with situation theory. A “change” relation is defined on the basic units of information (“infons”), which induces a topological space on situations. The topology supports a notion of information updated as time progresses but receding as we project into the futur...
We present preliminary work on a framework for constructing identities that takes as its point of... more We present preliminary work on a framework for constructing identities that takes as its point of departure Barwise’s channel theory, founded on category theory, and allied work on situation semantics. The framework takes actions relating to identity as fundamental. Actions and events are situations, although situations can include larger context. Identity-related situations are classified under (identity) types, which can be fused. Identity-related activity is partitioned according to the ‘form of life’ involved into (identity) environments, similar to Barwise’s classifications. An identity-related action is often based on another identity-related action as a resource, and all these actions may reference events in an individual’s life. The result is a layered picture of identity-related situations. Identities retain reference to the situations from which they are constructed, so provenance, context, and narrative detail are retained. The category-theoretical constructs of Barwise’s...
This paper discusses the use of self-adaptive Petri net to model the behavior of agent nodes in a... more This paper discusses the use of self-adaptive Petri net to model the behavior of agent nodes in a mobile ad-hoc network (MANET). The main objective is to capture and measure the adaptability of these agents in various hostile environments. Agents and their environments are represented by different interleaving and interacting components of a Petri net. For survival, each agent’s transition can self-organize in order to meet the challenges from environment signal that may perturb the agent’s functionality. While the aim of this paper is theoretical, simulation experiments are used to validate the cognitive claims associated to the agent’s behaviors
The concept of multiple satellite missions is becoming attractive because of their potential to p... more The concept of multiple satellite missions is becoming attractive because of their potential to perform coordinated measurements of remote space with greater temporal and spatial resolution, which can be classified as a sensor network or generally as Cyber-Physical System (CPS). Multi-satellite missions require a reliable and cohesive network between the small satellites. This paper represents our initial work on developing an inter-satellite communication link (ISL) that takes into account the symbiotic behavior amongst the system of satellites. The Open Systems Interconnection model (OSI) will be used as a framework for ISL. The data link, in particular the Multiple Access Control (MAC) layer, will be our primary focus. This paper discusses the communication design issues of PNSats within a sensor network construct. A hybrid combination of Time Division Multiple Access (TDMA)/Code Division Multiple Access (CDMA) scheme for a cluster of satellites is proposed. It consist of a TDMA-...
2014 IEEE International Inter-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support (CogSIMA), 2014
ABSTRACT We present a prototype sensor web that uses multiagent, Web, and Semantic Web technologi... more ABSTRACT We present a prototype sensor web that uses multiagent, Web, and Semantic Web technologies so that users may maintain awareness of the evolving weather situation as it relates to an activity in which they engage, in this case running. We apply Devlin's theory of information flow to explain how this system allows a user to maintain situation awareness. Ontologies are critical in this account for the flow of information. The system incorporates weather data into an ontology based on the OGC's O&M encoding schema and transforms it into data in a Running ontology via alignment of the ontologies. Devlin's situation semantics explains how the system provides a user with information on the running situation.
This paper proposed and developed socio-metrics for MANET performance evaluation. It is assumed t... more This paper proposed and developed socio-metrics for MANET performance evaluation. It is assumed that a MANET node is a socio-cognitive-technology system whose entities assume the behaviors of human and technology agents. Using the principles of network connectivity, we identify both intrinsic- and system level extrinsic- factors for the model development. The values of the intrinsic- and extrinsic- variables are generated using known fuzzy models. We analyze the contributions of the socio-metrics to the overall MANET system performance. The simulation outcomes inform that: (a) MANET nodes seek to achieve the same system goals after a significant period of interaction. We observe this as (a) MANET with less trust (10%) initially increased its trust level to about 25% after nine iterations of a simulation experiment; (b) Nodes will collaborate to achieve a singular system level goal as shown in balancing elements of the system level performance where high cooperation and self-awarenes...
Gaussian data augmentation has been used to improve the robustness of a deep leaning neural netwo... more Gaussian data augmentation has been used to improve the robustness of a deep leaning neural network model against adversarial attacks. This paper presents a design and evaluation of a Keras implementation of injecting colored noise during the training phase of deep networks. The evaluation was made using empirical robustness; a metric from the IBM adversarial robustness toolbox (ART). Results show robustness improvement over the base model when injecting black, pink and brown noise; and even better performance when combinations of ensemble noises are injected. Keywords—Colored Noise Augmentation, Adversarial Example, Machine Learning Robustness, Security of Deep Learning.
This paper considers a mobile ad hoc network (MANET) as a socio-technical system with human-machi... more This paper considers a mobile ad hoc network (MANET) as a socio-technical system with human-machine- characteristics. From this perspective, it is argued that a cognitive MANET will need certain human behaviours if it is to acquire intelligence reminiscence of a human social system. To achieve this, the spatial relationships among the MANET nodes require algorithms to process the node actions in order to create robust system-level behaviours from those relationships. We develop a technique known as KANA (a Knowledge Action Network for Agents) for use in developing such intelligence behaviours. KANA has two primary knowledge elements: a network- and a fuzzy matrix- representation of the nodes and edges in a MANET. The KANA algorithms have been used to model and simulate MANET behaviors in order to discover its sociometric properties
: The paper presents a preliminary result of using cognitive- and behavior-based modeling framewo... more : The paper presents a preliminary result of using cognitive- and behavior-based modeling framework to simulate a network of MANETs (Mobile Adhoc NETworks) as intelligent agents in a tactical battlefield. We demonstrate the efficacies of an agent-based modeling and the need for developing formal methods for multi-agent simulations from a system of systems (SoS) perspective. The development of our model framework informs the need to describe behaviors and relationships of actors and objects in the context of a mission space, and 2) to provide a foundation for modeling agent behaviors in a way that is plausible with respect to human behavior, a specially from the standpoint of human-system interactions.
2016 IEEE International Symposium on Systems Engineering (ISSE), 2016
Requirements, stakeholders' visions, drive any systems engineering (SE) process. But in desig... more Requirements, stakeholders' visions, drive any systems engineering (SE) process. But in design process practices, there is a gap often present between stakeholders' visions and requirements representation. This limits the impact that the stakeholders can provide input to the SE process. Then, the process may end up with errors or provide a product which is unwanted. We proposed a design methodology called the Responsive and Formal Design (RFD) process that directly involves the stakeholders input at each level of the requirement elicitation, shows its effect on the total system performance, and integrates high level requirements with domain specific considerations and verifies formally. It consists of a set of levels of representaion. Each level represents a set of requirements and its associated models, simulations, and the relationship between them. The levels of representation are related with refinement and abstraction relations. The refinement helps to make clear the connection with parametric considerations. As levels of RFD proceed towards refinement, the design process becomes a local or discipline specific activity, though always with a global perspective. Following our framework, this paper presents the implementation of the refinement process. Each level of the RFD process has its own level of granularity. This is true for the model, logical representation, and simulation of the requirements. In this paper, we define the pair of functions, refinement and abstraction, that exist between two models (system and component models in table form called classifications) and their logical representations (called theories) with different levels of granularity. We also show how high level requirements are interpreted in the refined level. We use an example of data from three small satellites, whose goal is to image the auroral ovals around Earth's magnetic poles, to demonstrate our development.
Barwise and Seligman stated the first principle of information flow: ”Information flow results fr... more Barwise and Seligman stated the first principle of information flow: ”Information flow results from regularities in the distributed system.” They represent a distributed system in terms of a classification consisting of a set of objects or tokens to be classified, a set of types used to classify tokens, and a binary relation between tokens and types that tells one which tokens are classified as being of which types. We aim to further this investigation and proceed with a dynamic or evolving system instead of a static system. We claim that a classification is a snapshot of a distributed system at a given moment or context. We then aim to answer the question posed by an evolving context. As the context or configuration changes, how do regularities evolve. This paper is a continuation of an investigation we started in [2], where we initiated how to capture a dynamism of information flow with a Kripke structure. Here we develop the same procedure with colored Petri net(CPN). We first ex...
We present a design framework for small-satellite systems that ensures that (1) each satellite ha... more We present a design framework for small-satellite systems that ensures that (1) each satellite has a consistent theory to infer new information from information it perceives and (2) the theory for the entire system is consistent so that a satellite can infer new information from information communicated to it. This research contributes to our Reliable and Formal Design (RFD) process, which strives for designs that are ”correct by construction” by introducing formal methods early. Our framework uses Barwise’s channel theory, founded on category theory, and allied work in situation semantics and situation theory. Each satellite has a ”classification”, which consists of tokens (e.g., observed situations) and types (e.g., situation features) and a binary relation classifying tokens with types. The core of a system of classifications is a category-theoretic construct that amalgamates the several classifications. We show how to derive the theory associated with a classification and the th...
Abstract. The firing rule of Petri nets relies on a residuation operation for the commutative mon... more Abstract. The firing rule of Petri nets relies on a residuation operation for the commutative monoid of natural numbers. We identify a class of residuated commutative monoids, called Petri algebras, for which one can mimic the token game of Petri nets to define the behaviour of generalized Petri net whose flow relation and place contents are valued in such algebraic structures. We show that Petri algebras coincide with the positive cones of lattice-ordered commutative groups and constitute the subvariety of the (duals of) residuated lattices generated by the commutative monoid of natural numbers. We introduce a class of nets, termed lexicographic Petri nets, that are associated with the positive cones of the lexicographic powers of the additive group of real numbers. This class of nets is universal in the sense that any net associated with some Petri algebras can be simulated by a lexicographic Petri net. All the classical decidable properties of Petri nets however are undecidable o...
In this paper, we develop analytical models to calibrate trust and distrust in cyberspace as a fu... more In this paper, we develop analytical models to calibrate trust and distrust in cyberspace as a function of risk. Trust and distrust are viewed as economic gains and losses, while risk represents adverse influencing factors such as privacy loss or virus attacks. The models are constructive and reflect the behaviors often used by people when they attempt to intrinsically calculate whether to trust or distrust cyberspace activities such as on-line banking or on-line shopping under available information on different subjective and objective risk conditions. The analytical models, although simplified, are able to capture the tacit and inherent human behaviors attributed to the user’s perception of risk and trust, as well as represent the economic utility derived from cyberspace. Sample distributions of the trust and distrust as functions risk are derived.
This paper begins by considering requirements for verification tools for NextGen, the US National... more This paper begins by considering requirements for verification tools for NextGen, the US National Airspace System under development. Category theory and channel are reviewed as well as their contribution to computer science, and how these theories can be useful in the development of NextGen is outlined. Category theory is a branch of mathematics that can provide a unified presentation of much of abstract mathematics, and channel theory uses concepts from category theory to provide an abstract and rigorous account of how information about some system component can carry information about other components. The second half of this paper introduces a notion of change into channel theory, which involves augmenting channel theory with situation theory. A “change” relation is defined on the basic units of information (“infons”), which induces a topological space on situations. The topology supports a notion of information updated as time progresses but receding as we project into the futur...
We present preliminary work on a framework for constructing identities that takes as its point of... more We present preliminary work on a framework for constructing identities that takes as its point of departure Barwise’s channel theory, founded on category theory, and allied work on situation semantics. The framework takes actions relating to identity as fundamental. Actions and events are situations, although situations can include larger context. Identity-related situations are classified under (identity) types, which can be fused. Identity-related activity is partitioned according to the ‘form of life’ involved into (identity) environments, similar to Barwise’s classifications. An identity-related action is often based on another identity-related action as a resource, and all these actions may reference events in an individual’s life. The result is a layered picture of identity-related situations. Identities retain reference to the situations from which they are constructed, so provenance, context, and narrative detail are retained. The category-theoretical constructs of Barwise’s...
This paper discusses the use of self-adaptive Petri net to model the behavior of agent nodes in a... more This paper discusses the use of self-adaptive Petri net to model the behavior of agent nodes in a mobile ad-hoc network (MANET). The main objective is to capture and measure the adaptability of these agents in various hostile environments. Agents and their environments are represented by different interleaving and interacting components of a Petri net. For survival, each agent’s transition can self-organize in order to meet the challenges from environment signal that may perturb the agent’s functionality. While the aim of this paper is theoretical, simulation experiments are used to validate the cognitive claims associated to the agent’s behaviors
The concept of multiple satellite missions is becoming attractive because of their potential to p... more The concept of multiple satellite missions is becoming attractive because of their potential to perform coordinated measurements of remote space with greater temporal and spatial resolution, which can be classified as a sensor network or generally as Cyber-Physical System (CPS). Multi-satellite missions require a reliable and cohesive network between the small satellites. This paper represents our initial work on developing an inter-satellite communication link (ISL) that takes into account the symbiotic behavior amongst the system of satellites. The Open Systems Interconnection model (OSI) will be used as a framework for ISL. The data link, in particular the Multiple Access Control (MAC) layer, will be our primary focus. This paper discusses the communication design issues of PNSats within a sensor network construct. A hybrid combination of Time Division Multiple Access (TDMA)/Code Division Multiple Access (CDMA) scheme for a cluster of satellites is proposed. It consist of a TDMA-...
2014 IEEE International Inter-Disciplinary Conference on Cognitive Methods in Situation Awareness and Decision Support (CogSIMA), 2014
ABSTRACT We present a prototype sensor web that uses multiagent, Web, and Semantic Web technologi... more ABSTRACT We present a prototype sensor web that uses multiagent, Web, and Semantic Web technologies so that users may maintain awareness of the evolving weather situation as it relates to an activity in which they engage, in this case running. We apply Devlin's theory of information flow to explain how this system allows a user to maintain situation awareness. Ontologies are critical in this account for the flow of information. The system incorporates weather data into an ontology based on the OGC's O&M encoding schema and transforms it into data in a Running ontology via alignment of the ontologies. Devlin's situation semantics explains how the system provides a user with information on the running situation.
This paper proposed and developed socio-metrics for MANET performance evaluation. It is assumed t... more This paper proposed and developed socio-metrics for MANET performance evaluation. It is assumed that a MANET node is a socio-cognitive-technology system whose entities assume the behaviors of human and technology agents. Using the principles of network connectivity, we identify both intrinsic- and system level extrinsic- factors for the model development. The values of the intrinsic- and extrinsic- variables are generated using known fuzzy models. We analyze the contributions of the socio-metrics to the overall MANET system performance. The simulation outcomes inform that: (a) MANET nodes seek to achieve the same system goals after a significant period of interaction. We observe this as (a) MANET with less trust (10%) initially increased its trust level to about 25% after nine iterations of a simulation experiment; (b) Nodes will collaborate to achieve a singular system level goal as shown in balancing elements of the system level performance where high cooperation and self-awarenes...
Uploads
Papers by Jules Chenou