Odorless and tasteless, VX is an oily liquid that is amber in color and very slow to evaporate; about as slowly as motor oil. Ricin, the toxin found naturally in castor beans, is poisonous if inhaled, injected, or ingested. It can be in the form of a powder, a mist, or a pellet, or it can be dissolved in water or weak acid.
Castor beans are processed throughout the world to make castor oil and ricin is part of the waste "mash" produced when castor oil is made. Ricin works by getting inside the cells of a person's body and inhibiting protein synthesis. The major symptoms of ricin poisoning depend on the route of exposure and the dose received, though many organs may be affected in severe cases.
Treatment is available, but long-term organ damage in survivors is likely. Death from ricin poisoning could take place within 36 to 72 hours of exposure, depending on the route of exposure inhalation, ingestion, or injection and the dose received. Ahmed Ressam, sentenced to 22 years in prison in for planning to bomb Los Angeles International Airport on 31 December , acquired this ostensibly genuine Canadian passport by using a fraudulent baptismal certificate; the obscured name was Western-sounding in an attempt to evade scrutiny at the border.
False travel documents are often based on lost or stolen passports. Prompt reporting of missing documents can be an effective deterrent to this threat.
For more information, please visit: travel. It can be very unstable and sensitive to heat, shock, and friction. TATP is made of a mixture of hydrogen peroxide and acetone with the addition of an acid, such as sulfuric, nitric, or hydrochloric acid. Formula: C 9 H 18 O 6. ANFO ammonium nitrate fuel oil An explosive mixture of ammonium nitrate and an organic fuel. Because of its ready availability and cheap material cost, ANFO has been used extensively as the main charge in improvised weapons around the world and is the most common commercial explosive.
Formula: NH 4 NO 3. TNT trinitrotoluene One of the most commonly used explosives for military and industrial purposes. Its insensitivity to shock and friction reduces the risk of accidental detonation. It appears as a yellow solid and is commonly mixed with other explosives materials in commercial boosters and military munitions or used as a main charge.
HMTD hexamethlene triperoxide diamine Improvised primary explosive prepared from three basic precursors: hexamine, a weak acid, and hydrogen peroxide. The product is highly sensitive to friction, impact, and electrostatic discharge. HMTD is corrosive in contact in metals and can degrade quickly if improperly synthesized or stored. UN urea nitrate High explosive produced by combining dissolved urea fertilizer with nitric acid.
Formula: CH 5 N 3 O 4. The Islamic State in Afghanistan may be down, but it's not out. But even though Zawahiri has conjured less of a personality cult, al Qaida's current leader is just as dangerous to the United States as its old one. Nineteen years after the terrorist attacks of September 11, , does al Qaida still pose a significant threat to U. Among researchers, military and intelligence officials, and policymakers who study the group, there is little consensus.
An accurate assessment of al Qaida's organizational health must take into account the group's recent and dramatic resurrection.
In this report, RAND researchers provide a snapshot of the terrorist and extremist threats facing the Philippines and the countering violent extremism efforts that the Philippine government and nongovernmental agencies have undertaken in response. Featured Terrorist organizations have long threatened the security, infrastructure, and citizens of nations and communities throughout the world. Report Who Are America's Jihadists? Sep 11, Brennan, Rick, Jr. Brown, Harold Brown, Michael A. Clutterbuck, Lindsay Cohen, Raphael S.
Daddario, Richard Daly, Sara A. Davenport, Steven Davis, Lynn E. Davis, Paul K. Fair, C. Christine Frelinger, David R. Godges, John P. Gompert, David C. Hiatt, Liisa Hlavka, Jakub P. Hoorens, Stijn Hunter, Robert E.
Ingram, Haroro J. Jackson, Brian A. Johnston, Patrick B. Jones, Gregory S. Jones, Seth G. Jung, Danielle F. Stephen Larson, Eric V. Libicki, Martin C. Miller, Paul D. Palimaru, Alina I. Parachini, John V. Also, most studies have been concerned with analysis of a single network. A further complication is that traditional social network measures are not designed for time-series analysis of dynamic networks. While static analysis may be adequate for slow-changing interpersonal networks, covert networks are characterized by their fluidity and dynamism.
Thus, analysis of covert networks needs to be approached from a dynamic perspective, tracking change inside the network as well as its static parameters.
Nevertheless, from an organizational perspective, it is important to look beyond social networks. Krackhardt and Carley [ 17 ] proposed concentrating knowledge about an organization in a format that could be analyzed using standard network methods, called the MetaMatrix.
The MetaMatrix analysis represents organizations as evolving networks in which the nodes in the social network are actively engaged in realistically specified tasks. This conceptualization made it possible to link performance to social networks and ask, at a concrete level, how changes in the social network could effect changes in performance.
Carley [ 18 ] [ 19 ] generalized this approach and extended the perspective into the realm of knowledge networks, enabling the researcher to ask how changes in the social network could effect changes in the distribution of information and the resultant impact of knowledge disruption strategies on organizational performance. By taking an information processing perspective, we are explicitly linking knowledge management and social networks [ 1 ] and enabling network evolution through learning mechanisms.
From a conceptual and data perspective, this means that we examine the co-evolution of all networks in the MetaMatrix as described in table 1. Moreover, we explicitly focus on the fact that the organization, and so these underlying networks, evolve. People Knowledge Tasks People Stuctural knowledge: Command and control structures, information pathways and relationships between organization members. Knowledge Distribution: Who has access to what knowledge within the organization.
Task Assignment: Who does which tasks within the target organization. Knowledge Knowledge Precedence: Which types of skills go together. Skill Requirements: Which skills are needed to accomplish a particular task. Task Task Precedence: On a tactical level, the sequencing and precedence of tasks that the target organization can accomplish.
A number of social networks metrics have been proposed for identifying the key actors who should be targeted in order to destabilize covert networks. Such metrics include, but are not limited to, those focused on centrality, random attacks, and from a more dynamic network perspective, cognitive demand [ 7 ]. Identifying an actor as key, using one of these metrics and then isolating that actor is a destabilization strategy.
We now consider several such strategies. The centrality approach , consisting of measuring the centrality [ 20 ] of each node in the network, then selecting a small number of most central nodes as targets for further action, is an intuitive approach to finding a core group of leaders within a terrorist network.
However, it is known from available intelligence that terrorist networks function in tightly connected cells and maintain only loose connections with the rest of the organization. Therefore, a search for highly central individuals is more likely to turn up a large number of agents that do not constitute the leadership circle, but are members of a densely connected cell. Moreover, as Borgatti [ 21 ] stated, none of the centrality metrics is guaranteed to disconnect the network into discreet components.
Bienenstock and Bonacich [ 22 ] have conducted a simulation study on vulnerability of networks to random and strategic attacks. The study suggests that as average connectedness of each individual node rises and high betweenness nodes are methodically attacked, the impact on overall performance of the network is minimal. However, if neighborhoods nodes connected to a high-centrality node are attacked along with the node, the opposite is true.
The implication of that result is that the cells of covert networks that are connected by a few individuals with high betweenness are very vulnerable to discovery of these individuals. Johnson et al. The results of this study suggest that perhaps the structural position of a gatekeeper is not important to the functioning of an isolated cell. However if two cells of the organization are to function in concert, the best position for the charismatic leader is in a gatekeeper role. The cognitive load approach described by Carley [ 6 ] combines static measures of centrality with dynamic measures of information flow, task performance and resource distribution.
These measures are based on the meta-matrix knowledge about the organization and have been shown to accurately detect emergent leaders. Consequently, cognitive load metrics can potentially be useful for detecting key members of terrorist networks. Based on the foregoing review of the literature we have identified a suite of destabilization strategies.
Each strategy identifies actor criticality in a different way. All strategies rely on data in one or more cells in the meta-matrix. The identified strategies are: Highest degree centrality: Isolate one agent from the covert network that has the highest degree centrality [ 20 ].
Highest betweenness centrality: Isolate one agent from the covert network that has the highest betweenness centrality [ 20 ]. Highest task accuracy: Isolate the best performing agent in the organization. This corresponds to standard police practice of arresting agents implicated in commission of a terrorist act. Amount of unique knowledge: Isolate the agent that has the highest expertise. When a destabilization strategy is applied, an actor is identified and isolated.
This results in one or more changes in the underlying networks in the meta-matrix and possibly a cascade of future changes [ 24 ]. Since the overall network is a complex adaptive system there is no guarantee that such cascades will destabilize the overall network, particularly in the long run. Thus, an examination of these destabilization strategies needs to be done in a dynamic context.
NetWatch: A Multi-Agent Network Model of Covert Network Surveillance and Destabilization NetWatch is a multi-agent network model for examining the destabilization of covert networks under varying levels and types of surveillance. Computational models, particularly, multi-agent network models, are a valuable tool for studying complex adaptive systems like organizations in general [ 14 ] [ 25 ] and covert networks in particular [ 7 ]. In multi-agent models, social behavior grows out of the ongoing interactions among, and activities of, the intelligent adaptive agents within the system.
From the meta-matrix perspective, actions of each agent or actor are constrained and enabled not just by the activities of other agents but by what resources or knowledge they have, what tasks they are doing, the order in which tasks need to be done, the structure of communication and authority, and so on. Further, the agents are intelligent, adaptive and computational information processing systems. The goals of NetWatch are to: Simulate the communication patterns, information and resource flows in a dynamic covert cellular network; Model the process of gathering signal intelligence on a cellular network and evaluate a variety of heuristics for intelligence gathering; Model and evaluate strategies for destabilizing a covert network based on intelligence obtained; Model reactions of a covert network to these destabilization strategies.
The social and cognitive underpinnings of the actors and the network in which they operate are based upon the CONSTRUCT model of the co-evolution of social and knowledge networks [ 18 ] [ 24 ]. The agents in the model perform a classification task that is information-intensive i. In the beginning of the simulation, agents are endowed with relatively little knowledge and must engage in learning behaviors in order to increase their task performance. Agents learn by interaction: trading facts with other agents or asking direct questions in hope of getting an accurate answer.
Agents also forget little-used facts. In keeping with the research in cognitive science, the agents representing humans are both cognitively and socially constrained [ 26 ] [ 27 ] [ 28 ] [ 29 ].
Thus, their decision-making ability, actions, and performance depend on their knowledge, structural position, procedures and abilities to manage and traverse these networks. Unlike Construct agents, the NetWatch agents are implemented as non-deterministic finite automata, with states of the automaton representing low-level behaviors and transitions governing the way the agent switches between them.
Some transitions are deterministic, others rely on probabilistic equations. Low-level behaviors include chatter, knowledge seeking, resource seeking, task execution and information reporting. Chatter is the simplest of the low-level behaviors. It can be thought of as non-goal-directed socializing, where some information is exchanged but it may or may not be relevant to the task the agent is engaged in.
Partners for chatter interaction are randomly picked from the agent's ego network peer group. Chatter uses the Knowledge Exchange Protocol see section 4. Knowledge and Resource seeking behaviors use the same protocol as chatter, however assign a higher priority to the messages.
Communication partners are determined by estimating the probability of a successful interaction, informed by the MetaMatrix representation of the agent's ego network. Processes that govern selection of partners are described in section 4. Task execution is described in detail in section 4. It is governed by a simple challenge-response protocol that is executed over one time period.
Task messages have the highest priority in the system and will preempt both knowledge exchange and chatter messages.
It is important to note that due to asynchronous execution of agents and multi-tiered message priorities, it is possible that some interactions will never complete or will complete after a significant delay. Each agent stores incoming unprocessed messages in a queue sorted by message priority. Thus, if an agent is overwhelmed with tasks or goal-oriented information exchange, most chatter requests will never be processed.
To prevent deadlocks, each of the messages is time-stamped at the time of sending, and interactions are set to time out after a fixed number of time periods. Also, agents are capable of handling multiple interactions at the same time, with task preempting based on priority of incoming messages. For example, if an agent was in the middle of a chatter interaction when a resource request or a task request arrived, the chatter will not be resumed until higher-priority interactions have been finished.
Formal and Informal Networks in NetWatch In NetWatch, the formal structure of the organization is specified as a directed weighted graph that specifies the communication channels that are open as well as their throughput or cost of communication. The directed nature of the graph allows one to specify one-way relationships and chain-of-command relationships.
The beliefs about the informal structure are individual to every agent, and also consist of a weighted directed graph. However, when an agent joins a network, its informal relationship graph is empty, and it must learn about the informal network before it can be used for communication. In NetWatch, the agents' interactions are governed by the formal structure of the organization, and agents' beliefs about the informal structure.
The agents communicate solely on the basis of networks that they belong to. Each of the networks is represented as a directed graph structure representing probability of communication or social proximity:. The agents do not have access to full information about the network, but rather every agent can only access a probability vector where is a probability of agent communicating with all agents.
This means that each agent may only know who it may interact with or is close to - but does not have access to interaction patterns of any other agents. Each agent also possesses a belief matrix that it uses to store any information it learns about interrelationships of other agents within the network. However, this information is far from complete and is often inaccurate. The directionality of the network also means that the communication may be asymmetric - thus allowing full representation of command networks as well as more symmetric friendship networks.
For example, in NetWatch, a cellular organization like Al Qaeda can be represented as a cellular network structure. The formal network consists of small densely connected cells that maintain a small number of connections to other cells.
The ties in experimental networks are generated from a profile of a cellular organization, such as one described in section 2. The profile contains the following information: For communication networks: Mean and standard deviation of size of cells, connection density inside cells and outside cells, density of one-way links and probability of triad closure; For knowledge networks: Amount of common knowledge doctrine , distributed knowledge group member specialties and specialized expert knowledge; For resource networks: Amount of resources needed to accomplish tasks, amount of common and distributed resources; For task networks: Branching factor and depth of the task precedence network.
The profile is used to create a probability distribution for each edge within a network, thus generating a space of random networks that all conform to the original profile. A number of sample organizations is then drawn from that space and run through the simulation, and mean and standard deviation of each of the resulting variables are taken.
Processes Governing Communication Each of the agents in NetWatch maintains a perception of its surroundings, via the notion of MetaMatrix 1. The perceptive MetaMatrix consists of the agent's ego network agents that it is directly connected with , agent's own knowledge, resources and task assignments, and is augmented by the agent's perception of other agents' ego networks, knowledge, resources and task assignments. However, an agent may only learn of other agents outside its ego network via interaction with agents that are in its ego network - and therefore any agent's perception of other agents' networks or knowledge is generally inaccurate.
Moreover, it has been shown [ 30 ] that knowledge of people outside a person's ego network decreases exponentially as graph distance between the actors increases. In the context of a cellular organization, this translates to agent's initial knowledge of its network including its cell because of dense communication patterns inside cells and a small number of agents outside the cell with whom cell members regularly communicate. Agents may later acquire further knowledge of the organization through interactions.
The choice of communication partner at every time period is based on two factors: Social proximity of the agents and their motivation to communicate.
Social proximity is defined as closeness of a relationship between two agents, scaled between 0 and 1 where 0 means "no relationship" and 1 is "very close relationship. We define homophyly to be based on a measure of relative similarity between agent and agent : the amount of knowledge that and have in common divided by the amount shares with all other agents, or. In both cases, agents operate on their beliefs about what the other agents know.
Thus, their predictions of relative expertise or similarity can be inaccurate. However, as interaction progresses and agents learn more and more about each other, they learn an increasingly complete picture of their world. Processes Governing Knowledge Exchange In a multi-agent network, the agents do not have perfect knowledge about the world. The only way to obtain information about the world is via ineraction with other agents - either through direct query or through information exchange.
Tracing back to its roots with Construct [ 18 ] model, the NetWatch model is based upon the concept of knowledge, knowledge manipulation and learning. In NetWatch, each agent's knowledge is represented by a bit string. A value of 1 in the position means that the agent knows fact and the value of 0 means that it does not. This allows for only a minimally acceptable performance and thus a very low utility , giving agents an incentive to communicate with other agents and attempt to gain more information.
To learn new facts, the agents execute the Construct Knowledge Exchange Protocol. For ease of description, we shall refer to the parties in knowledge exchange as Alice agent and Bob agent.
Note that Alice and Bob can be any two agents. Determine who to communicate with: Alice does this by evaluating Relative Similarity Eqn. After the probability of communication for each of the agents is computed, Alice throws a dice that reflects the computed probability vector and determines an agent to communicate with, or Bob.
Determine what to communicate: This is done by weighing information seeking vs. If Alice is in information seeking mode, it chooses at random a part of the knowledge string that is not known i. In similarity-based communication, Alice chooses a part of the known knowledge string and sends it to its counterpart. Determine proper response: On receipt of a query, Bob determines if it should answer it by checking whether the sender of the query is a part of its network and whether it has the knowledge in question - and, if all is good, sends a reply.
If Bob does not know the facts requested, it checks its internal belief matrix and may respond to Alice with a name of another agent Clare that may be better suited to answer Alice's question.
In this case, the agents exchange referential data. On receipt of knowledge, Bob determines if the knowledge is useful i. If all is good, the agent will choose some knowledge from its knowledge base and send it back. Update internal knowledge base: On receipt of the reply, Alice determines the usefulness of the reply and uses that to update its internal knowledge of Bob "Bob knows fact " as well as its knowledge base "I now also know fact ". If Alice receives referential data, it uses that to update both its knowledge of Bob "Bob does not know fact " and "Bob knows Clare" and its knowledge of Clare "Clare may know fact ".
This may be followed by a query to Clare - which may or may not be honored. Note that Clare may not have been originally a part of Alice's network - but now, through Bob, Alice has learned about her existence. Thus, agents within the organization use referential data about each other to form an informal network.
Due to asynchronicity of communication, the agents may not be able to conduct a knowledge exchange transaction in one time period. It is also possible that agents may be too busy to be able to respond to a query, and may either delay or terminate the transaction.
The knowledge exchange protocol, however, provides for a robust deadlock resolution, allowing agents to detect a transaction that is deadlocked, terminate it and start anew, finding a different party to communicate with. The task is merely defined as a function that maps a problem vector and agent's knowledge and resource vectors onto a result vector. In NetWatch, we measure agent performance as accuracy in performance of a ternary classification task.
The classification task is represented by a vector of binary values. An agent can only access bits in the task vector that correspond to non-zero values in agent's knowledge vector.
The task is then decided by a "majority rule.
0コメント