Adaptive Controls for In-Line Checked Baggage Screening Systems - - PDF document

• !

Adaptive Controls for In-Line Checked Baggage Screening Systems (CBSS)

Shalom Dolev and Cathal Flynn

SecureLogic Corporation, New York, New York, USA Abstract The first principle of risk management is to reduce risk to the lowest affordable level. In checked baggage screening, the principle dictates that every bag should be so screened as to achieve the highest probability of bomb detection (Pd), given its risk category, the availability of resources (numbers of

• perating EDS and TSO’s on duty), volume of bags entering the CBSS, current and projected loads at

potential chokepoints, and time to departure. That requires a computer-based, upper-level, adaptive management, control, and communication system, installed above the programmable logic control (PLC)- based system that suffices for baggage handling but is insufficiently agile for CBSS. The adaptive controls implement the ability of EDS to switch detection modes on the fly, and facilitate alarm resolution processes at each subsequent level of the CBSS. They permit rapid, efficient adjustment to heightened security conditions (e.g., Orange or Red) and to focused alerts that require more stringent treatment of specific bags (e.g., because of their airports of origin, or higher probability that they contain certain difficult-to-detect types or configurations of explosives). Using adaptive controls, the screening authority can achieve the highest Pd and most effective security in all circumstances. The adaptive control system should also constantly record every event in the CBSS, to enable accurate performance assessment and informed CBSS adjustments. The paper describes the necessity for an adaptive control system, its design and attributes. Givens for CBSS One element in checked baggage screening is the division of bags according to their estimated level of risk or, in other words, their relative levels of threat. In the United States, there are two broad categories: selectee and non-selectee (normal) bags. The security authorities of some other nations divide bags into three or more standing risk categories. In the United States, selectee bags comprise 5 to 10% of the total, and non-selectees therefore comprise the remaining 90-95%. A second element is the limit on screening resources, such as the number of available, operating EDS, and particularly the number of Transportation Security Officers (TSOs) at work in the second and third levels of the CBSS, where EDS alarms are resolved, on-screen at Level 2 and, if not resolved at Level 2, at Level 3 through several options of search and Explosives Trace Detection (ETD) procedures associated with the X-ray image from the EDS. A third element is the ability of EDS to have three or more detection modes installed, and to switch on a bag-by-bag basis without re-booting (on the fly) among those modes. For EDS employed in the United Sates, three modes are pertinent; we will call them Normal (detection and false positives meeting the standards for certification established by the FAA in 1993, but with a lower target explosive mass recently established by the TSA), High (with a higher probability of detecting a wider range of explosive configurations, but with a consequently higher false positive rate), and Speed (with a lower probability of detection than the Normal mode, but with faster automatic processing and a much lower false positive rate). Even though Speed mode has a lower probability of detection (Pd) than Normal mode, its Pd is at least equal to that of approved alternative “mitigation” screening processes currently in use. The Problem of Peak Baggage Loads The problem of peak baggage loads is twofold. First, the peaks may overwhelm the CBSS resources (EDS, but more often the staff performing alarm resolution procedures), thus causing some bags to be late for their flights, unless the CBSS is equipped and staffed to screen all the peak bag loads that might

• !
• ccur during specific periods. Second, it may be unaffordable to build, equip and staff the CBSS to cope

with all bag load eventualities. There are three pertinent categories of peak baggage loads in CBSS. The first comprises diurnal peaks, which occur at predictable periods of each day, and are associated with increased flight departures per unit of time. Also included in diurnal peaks are the additional bags associated with particularly busy days

• f every week.

The second category consists of ultra-peaks: those that occur on a very few days of exceptionally heavy airline travel, such as the Sunday of Thanksgiving weekend at many stations. Ultra-peaks can substantially exceed diurnal peaks in both bag volume per unit of time and in their duration. A third category is related to the resources available for screening at any given time. These resource- related peaks occur the forecast lulls (long periods of low baggage volume entering the CBSS) between diurnal peaks. During the lulls, for economy’s sake, staffing (the number of Transportation Security Officers, TSOs, on duty) is reduced at Level 2, where some alarms from the EDS are resolved by image interpretation on-screen, and at Level 3 where the remaining alarms are resolved by manual search and use of explosives trace detection (ETD) equipment. Normally, the reduced staff is able to perform the alarm resolution procedures thoroughly on the lower volumes of bags during lulls. If, however, unexpected peaks of baggage loads then occur in the CBSS, the reduced staff at Levels Two and Three may be unable to complete alarm resolution procedures on all alarm bags within the time allowed, and flight delays will then result. (The effects of the unexpected bag surges will be most heavily felt at Level Three, because bags that might have been cleared by on-screen procedures, had there been sufficient staff at Level Two, will by default be sent to Level Three, adding to the bags awaiting resolution there.) Design Capacity of In-Line CBSS Since diurnal peaks are defined as everyday occurrences, CBSS should be designed and constructed to cope with them. In operation, the CBSS should be sufficiently staffed at the times of expected diurnal peaks to resolve alarms in the allowed time. We note that the TSA Aviation Security Advisory Committee’s Baggage Screening Investment Study Working Group Report, August 2006, suggests that future growth in checked baggage loads will be accommodated by expected increases in EDS efficiency (higher scanning rates, lower false positives), and that current capacity therefore will be sufficient over the CBSS lifetime. We believe that assumption would be reasonable if the explosives (types, masses, configurations) to be detected remained constant, but that, unfortunately but clearly, is not at all likely. Out of necessity or by preference, terrorists are using an ever-widening range of explosives, many of which have considerable similarity in EDS detection- relevant physical characteristics to innocuous materials commonly packed in baggage. Detecting the additional explosives is bound to decrease scanning rates and increase EDS false positives to an extent that will exceed the otherwise reasonably expected efficiencies cited by the BSIS Working Group. We therefore believe that CBSS design capacity must be based on forecast diurnal peaks and the necessity of detecting, at least during prolonged periods of heightened threat, a wider range of types and configurations of explosives. In contrast, it is not clear that ultra-peaks should determine CBSS capacity. Given their rare occurrence, the incremental cost of building CBSS to cope with them might be excessive. The rate of increase in capital cost of CBSS capacity is not linear. If the load increment is 10% over the diurnal peaks, a CBSS with capacity to screen the ultra-peaks could cost 20% more. In operation, economy requires that CBSS be staffed sufficiently to deal with the differing expected loads during the peaks and lulls of flight departures through the day. Over-staffing to deal with all possible load conditions during TSO shifts is probably not affordable. Risk Management in CBSS An alternative to designing and staffing CBSS to deal with all possible peaks is to screen bags according to their individual levels of risk, applying risk management with diligence, foresight, and accountability. In its essence, it means screening bags in the following framework:

• 1. Bags that are assessed (because of their origin, for example) as having a significant probability of

containing “new” explosives that are not detectable by EDS (possibly because detection algorithms

• !

have not yet been developed for them), are routed to Level 3 even if it has been determined at Level 1 that they do not contain other explosives for which the EDS has activated detection algorithms. Level 3 staff are informed of the bags’ status and given step-by-step clearance procedures for them.

• 2. Other selectee bags are always screened at least with Normal mode at Level 1, but with High mode

whenever the resulting increase in alarm bags can be accommodated at Levels 2 and 3.

• 3. Non-selectee bags are normally screened with Normal mode at Level 1, but with High mode when

Levels 2 and 3 can accommodate the increase in alarm bags.

• 4. When Levels 2 and 3 are about to be over-loaded with alarms bags requiring resolution, Level 1 EDS

will employ Speed mode to as high a proportion of non-selectee bags as is needed to reduce bag arrival at Level 2 and 3 to the rate that the staff on duty can clear on time.

• 5. Speed mode will also be similarly employed when, as at ultra-peaks or if some EDS are out of
• peration, Level 1 capacity would otherwise be exceeded.

In summary, this screening framework entails the least increase in risk due to using Speed mode during peak conditions that otherwise cannot be affordably accommodated, and on balance, due to its use of High mode whenever possible, assures the highest level of security from the available screening

• resources. The framework conforms to an operating principle established by Secretary Chertoff’s Second

Stage review, to “use risk/value analysis to make operational decisions.” The framework requires the use of a CBSS control and management system with several attributes that are not provided by PLC-level systems alone. The control system must provide for: ready set-up of screening procedures by the appropriate TSA official; system self-awareness of every bag’s identity and risk level, the rate of arrival of bags and the screening rate at each level of the CBSS; adaptation of the CBSS to (sometimes rapidly) varying conditions of load and threat; and a record of every bag’s screening process, CBSS components’ availability and performance. SecureLogic’s iScreen-Line is such an adaptive, self-aware, responsive, and accountable system. iScreen-Line Components For a typical CBSS, iScreen-Line is composed of four main subsystems: The Screening Process Editor (SPE) The Screening Process Editor (SPE) is a ‘stand alone’ application that serves as the security director’s interface with iScreen-Line. It allows the director to define and modify the screening processes and alarm resolution protocols for all combinations of bag risk levels, real time loads, and system capacities. No programming skills are required for the operation of the SPE interactive ‘step by step’ editing tool. The editing process consists of adjective structured tables and graphical tools enabling the FSD (or the FSD’s designee) to set the specific screening process for each risk level at different load levels. Each and every screening step is discretely defined and set. The next screening step is defined as a function of the results of the former screening. The setting of the screening processes is open so as to enable the efficient deployment of any security rule and procedure. Infrastructure information derived from the physical layout and design of the screening system is fed into iScreen-Line via the SPE module. This information includes (but is not limited to): risk levels of bags (selectee and non-selectee); other risk factors (e.g. origin, destination, airline); load levels; definition of significant points in the CBSS layout; types of EDS; possible screening setups and results that these EDS may produce; number of Level 2 workstations; number of search/TD stations in Level 3; possible routes between various locations in the layout. A compiler performs a wide set of validity and integrity checks in order to detect erroneous or missing rules.When all tests are successfully passed, the full set of routing rules is converted into the Graphical Logical Editor runtime application.

• !

The SPE screen The Graphical Logical Editor (GLE) An object-oriented parametric tool, the Graphical Logic Editor (GLE) enables the project implementer and users to define, set up and configure the logic control for the specific CBSS layout. This is done using predefined libraries of common system components (such as various EDS and alarm resolution practices), as well as the physical layout of the system (such as decision points and available routes). The GLE defines the logic of the screening and the routing of bags throughout the screening system. It

• ptimally assigns screening scenarios to bags in accordance with preset screening processes, as set by

the SPE. The GLE then monitors and directs the CBS control system to route each bag through the screening process from entry to clear-for-sortation. In the design stage, the GLE is used to program the logical flow of the screening system, dictated by the actual CBS layout. Programming will include the screening process for each category of bag, EDS and routes, defining report points and decision points and interfaces with the baggage handling system control. In run-time, the GLE monitors the CBS components via the BHS control system for status information and the screening results of each bag in the system. The status information and screening results are processed for further real-time routing commands based on the routing table generated with the SPE. The GLE monitors and measures in real-time the total number of bags within the CBSS and at each of its segments, e.g.,Level 1 EDS mainframe, Level 2 OSAR, and Level 3 AR. In addition the system monitors the available screening resources (including the number of TSOs on duty) at each of those CBSS

• segments. These data and risk criteria inputs are processed by a multi-dimensional algorithm to assign

the optimal screening scenario for every bag. The GLE also feeds the statistics subsystem with data gathered as the bag proceeds through the screening system to serve as a database for key performance indicators (KPI) analysis. The GLE consists of control software objects arranged in libraries and a generation engine. The required control software for each BHS element or EDS machine is packed into a software object with its specific properties and parameters. To set the process logic for a specific CBS layout, the adaptation team will select and place the components in a logical layout using ”drag and drop” functionality and will then connect the objects according to the specified process logic for the site. Once the control logic settings are complete, the GLE runs an automated simulation, testing the process for validation, verification and debugging of the process logic. The built-in simulator simulates and emulates the flow of baggage throughout the CBS, showing the bags’ route options and screening

• procedures. The bags move on the graphical logical conveyors from the entrance to the exit (always

downstream). The GLE includes the layers that are needed for communication with the BHS control system, as well as PLC communication layers with all standard controllers, as required.

• !

The GLE interfaces (via the BHS PLCs) with the EDS machines in the screening system, with the purpose of pre-setting and configuring the machines as required for each bag and to acquire the screening results for all bag screening runs (alarm, no result, dark alarm, clear, etc.). The GLE activates the RLR (Risk Level Retriever) in order to automatically retrieve an incoming bag’s Risk Level from external data systems (e.g., its BSM) or from the bag tag in accordance with the airline. The SPE and GLE tools allow fast and simple system modifications that may be required as a result of future changes in security policies and procedures, as well as new models or versions of EDS machines. The Runtime GLE Screen The GLE Simulator Screen The Statistical Analysis (STA) Subsystem The Statistical Analysis subsystem gathers all the information received during the system run-time, and then generates system performance analyses. This includes analysis of the following KPI: Object processing time. Clear / Reject ratios. EDS performance (e.g. machine scanning times and idle time). Overall system load throughout the day - may be broken down and indexed according to machine, route and conveyor. Identification of system bottlenecks. Comparison of different EDS machines’ alarm results to the norm. Human operator response times and tendencies (as affected by risk level, flight destinations, etc.). The data required for generating reports by the STA subsystem are copied periodically from the

• perational database to a secondary one. This secondary database serves as a historical repository,

responsible for both concatenating all the historical data and archiving it for statistical purposes. The following is a partial list of reports provided as standard by the STA subsystem: EDS Statistics

• Number of bags alarmed by a specific EDS
• Number of bags cleared by a specific EDS
• EDS machine fault
• EDS hours of operation

OSAR Statistics

• Total number of bags through OSAR
• Total number of bags through OSAR by

EDS

• Total number of bags cleared by OSAR
• Average time to clear bag by OSAR

ETD Area Statistics

• Total number of bags received in ETD
• Total number of bags cleared by ETD
• Total number of bags by station

Screening Baggage Volume

• By EDS machine
• By ETD area
• By ETD station

• !

The Search Area Management Module (SAM) The Search Area Management Module (SAM) is a software application that provides a comprehensive computerized management, monitoring, control and audit solution for the ETD/physical search operation at Level 3. SAM is a modular tool that combines a screening process control with a management application, based

• n load monitoring, queue balancing, prioritized management and time optimization.

SAM monitors, directs and controls the screening operation for each and every bag within the alarm resolution Search Area, according to pre-defined rules and requirements regarding security protocols, throughput and timing considerations. The system automatically processes all relevant real-time and preset data to provide the optimal screening process sequence and timing for each and every bag. SAM generates a status table of bags within the Search Area, available on line to external users. This table enables the terminal or airline baggage system operator/supervisor to check which bags are in the Search Area and to estimate the lead time to release in order to monitor the loading of the baggage onto the aircraft and thereby minimize bags being delayed or missing their flights. The SAM module will alleviate the TSO’s requirement to manually fill in alarm resolution reports. The forms currently filled manually are to be generated automatically, using information accumulated during the system's operation. This automated reporting feature applies to both alarm resolution operations, the Level 2 OSAR and the ETD/ manual alarm resolution at Level 3. SAM touch screen WS SAM User Interface Screen