Today’s “intelligent” security systems only confirm security breaches that have already taken place. They do not and cannot anticipate and prevent impending breaches before they occur, in real time. This sums up the most serious problem faced by today's security infrastructure.
Today’s security systems continue to rely on a human operator’s experience and instincts to process collected surveillance data and attempt to anticipate an impending security breach. Therefore, there is a critical need for a cognitive security system that simulates how we, as humans, instinctively predict, adapt and react to security breaches.
Cognitive security systems are based on sophisticated mathematical techniques that simulate the instinctive and emotional properties of a human brain, and are designed to anticipate impending security breaches in environments that are not well defined and are constantly changing. This forward-looking, predictive capability represents a fundamentally disruptive advance in security technology, which will dramatically reduce costs and collateral damage, and -- most important -- save lives.
Security systems claiming to be intelligent are really “knowledge-based” and very limited in their abilities. They are only marginally effective, in very rare cases where security environments and the aggressors’ threat profiles are identical to those upon which the systems’ knowledge is based.
Typical security environments are highly dynamic and constantly changing, especially in venues that are crowded with people, vehicles and objects moving about and continuously being re-positioned. Even in relatively uncluttered environments, multiple intruders entering the environment can cause it to become dynamic, through techniques such as diversions and decoys, to name just two. And these techniques can easily confuse a non-intuitive, knowledge-based security system. If conditions exist that were not part of a knowledge-based system’s training data set, the system’s results would be questionable.
The fallback position continues to rely on a human operator’s experience and instincts to process collected surveillance data and attempt to anticipate an impending security breach. Therefore, there is a critical need for a cognitive security system that simulates how we, as humans, instinctively predict, adapt and react to security breaches.
No pre-programmed rules
Cognitive security systems are not based upon preprogrammed “rules.” Instead, they are based on sophisticated mathematical techniques that simulate the instinctive properties of a human brain. The security environment within which such systems operate is subject to baseline adversary goals and threat profiles directed against adversary targets that often become more sophisticated as adversaries learn to take advantage of weaknesses within the security environment.
The state of security is defined by processing data acquired by multiple security peripherals, such as security cameras and radar, as well as motion, biometric, chemical and radioactive sensors, etc. Processing techniques and algorithms facilitate security-event detection and a cognitive security system utilizes symbolic cognitive architectures and inference process algebras that enable the system to possess instinctive qualities and autonomously “learn” and adapt to a dynamically changing security environment.
These symbolic cognitive architectures and inference process algebras enable the system to infer intentions or activities of aggressors through the detection of their actions. These algebras and architectures have built-in, cost-optimization mechanisms that allow them to deal with undetermined, incomplete and uncertain information.
These algebras have been successfully applied to the U.S. Office of Naval Research robotics test-bed to derive Generic Behavior Message-passing Language for behavior planning, control and communication of heterogeneous Autonomous Underwater Vehicles operating in hostile environments.
$-calculus expresses all variables as “cost” expressions. One of the cost functions used in this system is “uncertainty,” which operates using an internal value system that is not only dependent on physical conditions of the real-time security environment, but also depends upon meta-states of the environment associated with unforeseen changes and/or conditions that lie outside the baseline goals and threat profiles of known adversaries.
The system’s “internal values”
These internal values are designed in accordance with psychological terms that human beings associate with “drives” and “emotions.” These internal values do not actually realize real drives and emotions, but the system is designed to exhibit behavior that is governed by drives and emotions.
The system imitates emotionally-driven human behavior and responds to dynamic changes in the security state, just as humans might. Specifically, the ”emotional state“ of the security system is strongly influenced by psychological internal values simulated, for example, by “suspicion,” which is associated with an increase in unusual or atypical sensory inputs from security peripherals, and ”curiosity,” when there are dramatic fluctuations in the sensory data.
These internal values help define the emotional state of the security system with “fear” being associated with increased suspicion and “happiness” being associated with increased curiosity, through use of symbolic cognitive architectures and inference process algebras.
A cognitive system autonomously updates in real time with new and/or emerging adversary goals and threat profiles that could characterize new or unforeseen security breaches.
Such a system strives to minimize uncertainty, suspicion and fear cost expressions, as in a manner that simulates the cognitive processing abilities of human beings, given the same conditions. The result is a security system that is able to instinctively predict impending security breaches before they actually occur.
A gaming application
For example, casino surveillance directors are well aware of threat profiles of gamblers who use statistical card-counting techniques to maximize blackjack winnings. The behavior of card counters is quite deliberate; they do not wander or linger, but, instead, move directly to an empty table, using only two decks of cards. Variation in their wagers is abnormally large and is correlated with statistics of prior hands played.
Security cameras provide detailed video of cards being dealt, payouts and losses. Behaviors are analyzed by processing video data and continuously updating a database of hands played, cards dealt and wagers associated with wins and losses. The data are processed by symbolic cognitive architectures and inference process algebras.
Using “probability” as its cost function to build a ranked set of hypotheses for prediction and interpolation, the cognitive system looks for probable correlations that suggest suspicious behavior patterns.
While the presence of two or more players at a blackjack table normally wouldn’t be cause for alarm, the symbolic cognitive architectures and inference process algebras drive the system to be “curious,” looking for unusual correlations, such as the behavior of a team of card counters at the table. In this case, the team’s playing strategy is correlated in a manner atypical of recreational players.
Recreational players strive to win, but card-counting teams have an opposite behavior -- one player deliberately loses a small-wager hand in order to optimize the other member’s chances of winning with a significantly higher wager. As the system detects these correlations, it learns the new threat profile for a team, then it reconfigures security cameras to provide more details of their playing pattern. This system continues to process data using these architectures and algebras until a user-defined threshold probability suggests that a security breach related to card counting is about to take place.
Cognitive systems are the cost-effective answer
Today, only highly trained security officers are capable of correlating suspicious behaviors of individuals, which, although effective, is expensive and very inefficient. What is needed is a security system designed to replicate the instinctive and cognitive capabilities of these highly trained officers.