Improving rapid counter terrorism decision making COGITO Artificial Intelligence based human pattern recognition General Terrorists are threatening world peace. While intelligence is limited and cultural gap are a challenge COGITO is an effective tool to know who among travelers might pose a serious threat to security. While most technologies focus in identifying specific people based on prior intelligence COGITO concept and approach is focusing in the human factor and Intent Detection. COGITO systems are deployed and used by governments in US, Israel, Singapore, Mexico, Latin America and other parts of the world. This field proven COGITO technology led to detection of criminals, terrorists and employees harboring hostile intent or actually conducted crimes. The main obstacle to COGITO becoming a standard the fear of authorities of privacy concern and public opinion. The COGITO is an automated decision making system capable of collecting and analyzing psycho-physiological indications and cross-referencing these indications with additional objective (and available) information. In test duration of 5 minutes, the system can isolate those examinees (suspects) that qualify for further procedures. The system performs this examination with a high level of accuracy and reliability. The COGITO is specially designed for terrorist and Internal Threat detection and was tested successfully by the DHS. The system high performance reach as low as false-positive 5% (false positive) and 90% hit. This is achieved through the specificities of the COGITO methodology and software - Cross-referencing objective information with subjective reactions to specific terror related issues and stimulation, Using the Guilty Knowledge Test (GKT) method as opposed to the Control Question Technique (CQT). COGITO Concept COGITO presents a significant conceptual breakthrough that can assist international aviation and homeland security authorities in responding to increasingly sophisticated means of international terrorism. This concept is based on several well-established paradigms and assumptions. The COGITO system is a technology-based concept and solution for the detection of suspects harboring malicious intent serves for detection of Internal Threat (employees of governmental agencies and enterprises that have
destructive intents), Police interrogations and border security. The COGITO concept is derived from extensive interdisciplinary know-how in security, polygraph testing and field-proven security-related interrogation techniques. The COGITO core technology is based on proprietary software an expert system that emulates an investigator s Modus Operandi by incorporating soft decision-making algorithms such as Neural Networks and Fuzzy Logic. All hardware elements are best-of-breed off-the-shelf third-party components. The technical solution is comprised of a front-end, the Test Station, and a back-office where multiple-station and multiple-site data is stored, managed and distributed. Stimulated Psycho Physical Reaction (SPPR) The COGITO method is based on stimulating examinees with specific terrorism-related triggers using a direct contact, interaction, conscious, portal approach: The COGITO method postulates that specific words or questions can force terrorist to generate a SPPR that is identifiably different than that of a non-terrorist s SPPR to the same words or questions. Based on extensive field experience accumulated by Israeli security agencies, the only common characteristic to all suicide bombers and effective terrorists is their desire not to be caught by security authorities. The terrorist s fundamental motivation to successfully perform the terrorist act and not be caught by security authorities clearly differentiates him from the innocent person not harboring such intent. This identifiable motivation is known as the terrorist hunting hunted syndrome (THHS). In order to identify and isolate the terrorist, one needs to stimulate and detect the THHS. In order to expose the terrorist, the COGITO system generates a specific monitored stimulation of the THHS (i.e. to check the terrorist individually with his full awareness to the fact that he is being checked, and with wording or questions having specific relevance to terrorism or terrorist culture and terminology). This stimulation will engender an uncontrolled reaction (SPPR) from the terrorist, different than the reaction common to the so-called innocent person. Direct contact will ensure accurate calibration of the sensors to each individual and to each specific stimulated interaction. Terrorists can use sedation and control their psycho-physical output. Only full-contact sensors can detect the alleged sedation by examining the individual s reactions to specific stimulation. There is no academic or field experience supporting the assumption that a terrorist will tend to externalize a higher level of nervousness than an ordinary traveler. This is accentuated by the fact that there are many reasons for which an ordinary traveler may show signs of nervousness. Conversely, and based on field experience accumulated by Israeli law-enforcement and security agencies, the typical suicide bomber is highly mission-focused, and in most cases, does not show his stress in a visible and detectable manner. Uses of general excitement detection methods have been shown to yield unacceptably high positive-false alarms (30% and higher). Thus, any artificial statistic calibration to reduce positive false alarms will be of little value if not based on operational and field experience.
Low false alarm The system achieves results as low as 5% false-positive and negligible false-negative results. This is achieved through the specificities of the COGITO concept Cross-referencing objective information with subjective reactions to specific terror related issues and stimulation Using the Guilty Knowledge Test (GKT) method as opposed to the Control Question Technique (CQT) One of the tenets behind the COGITO method is the Guilty Knowledge Test (GKT)*. This method differs from the classic Control Question Technique (CQT), more commonly used by polygraphists. This method offers a methodology that enables investigators to identify the perpetrators of a criminal act. This is achieved by questioning the suspect based on information that can only be known to the actual perpetrator and that is not available to the general public. The underlying assumption of this theory is that when an individual performs an emotionally affecting act involving guilt or fear (known only to himself and to the investigator) his reaction to a specific event-related stimulation will be different than that of a non-involved individual. This example demonstrates how the chance to randomly generate a false-positive result (i.e. that an innocent person will react strongly to the RSOs) when using the GKT method is extremely low. 1/5*1/5*1/5 = 1/125 or less then 1% Using the GKT method enables COGITO to build and use an algorithm that can significantly reduce the levels of false-alarms. Cross-referencing the GKT results with additional objective passenger information reduces the level of false-alarm even further. * http://en.wikipedia.org/wiki/lie_detection#the_control_question_test_and_the_guilty_knowledge_test COGITO Algorithms The software component that is handling the decision process is a multi-layer algorithm: The Signal Analysis Input Algorithm The GSR output is analyzed by algorithms which are based originally on an expert system. This expert system resembles the way a polygraph specialist analyzes the
polygraph visual graphs. COGITO engineers have studied thousands of polygraph tests under the supervision of leading polygraph specialists. This knowledge base has been transformed into 4 basic algorithms that have been improved upon based on trials and studies. These 4 different algorithms are being used to analyze 12 different parameters (signal slope, amplitude, etc.). Signal analysis algorithm utilizes Sugeno type fuzzy interface system. The system uses several rules. For example: the height of the peak and the delay between the rise time of the reaction and the question. In the algorithm there are several rules with up to 4 inputs. The second part of the algorithm is Neural Network based. One of the inputs for the fuzzy system described above is the dissimilarity of a reaction. In order to accomplish this subjective parameter, the COGITO signal analysis algorithm uses LMS linear neural network. The network is trained on the input signal [P(t)]. After the learning, the network is used as a linear predictor for the signal and the error represents the relative dissimilarity of the response signal. The system uses several rules of the type described below. In the figure below, Input 1 and Input 2 are, for example, the height of the peak and the delay between the rise time of the reaction and the question. The output of the algorithm as described below, represent the relative excitation for each question. In the algorithm there are several rules with up to 4 inputs.
The second algorithm is Neural network based. One of the inputs for the fuzzy system described above is the dissimilarity of a reaction. In order to accomplish this subjective parameter, the COGITO signal analysis algorithm uses LMS linear neural network. The network is trained on the input signal [P(t)]. The learning is done with the Least Mean Square algorithm as described below: Where α is the learning rate. After the learning, the network is used as a linear predictor for the signal and the error represents the relative dissimilarity of the response signal. Peak of tension algorithm that compares only the signal altitude.