In 2013, as Brazil was preparing to host the 2014 World Cup and the 2016 Olympic Games, the Brazilian Customs administration started exploring ways to improve air passenger control in order to be able to effectively process the athletes and millions of tourists who were expected to attend or participate in these events. The remarkable growth of international air traffic in the last decade was already challenging the Customs service whose traditional targeting techniques were deemed inadequate to properly manage the security risk posed by travellers.

Two major investments were made by the Federal Revenue Service of Brazil (RFB): a risk assessment system that would enable Advanced Passenger Information (API) and Passenger Name Records (PNR) to be processed was developed, and a passenger Facial Recognition System (IRIS) was purchased. By combining the two technologies, Brazilian Customs managed to achieve vital leaps in efficiencies when it came to the processing of international travellers.

The deployment of these tools was done in cooperation with the National Commission of Airport Authorities, generating an important spinoff: the various stakeholders in the airport environment, such as airline companies, airport operators, Customs and immigration services as well as other government agencies and policy regulators, entered into a constructive dialogue. This cooperation has been formalized through the creation of the National Committee of Airport Authorities (CONAERO) and its technical thematic subcommittees where representatives of all stakeholders that are active in the airport environment meet to discuss issues related to air transportation.

Until 2014, Customs officers posted at the airport based their decision on whether to select and control a passenger and his/her baggage on the results of behavioural analysis, questioning, baggage surveillance, and on other random factors.

The implementation of new analytical tools to screen passengers in 2015 was a game changer, especially with regards to identifying travellers who fit drug smuggling profiles. There was an increase of more than 360% in drug seizures between 2014 and 2016, a record in the history of the Customs service (see figure 1). Other common crimes or offences include the transport of undeclared cash, the illicit acquisition of cultural objects, and the smuggling of guns.

The system in place can analyse both API and PNR data but, with strong privacy protection regulations in place within the European Union (EU) that prevent countries who do not negotiate a bilateral agreement with the EU from receiving PNR data from European airlines, it became clear that the Brazilian Customs administration should focus on how best it should use available API data.

API data is essentially collected at check-in. The best option, from a data quality perspective, is the collection of machine readable information present in many passports, via an automated process. But it is not unusual for API to be collected manually by check-in staff, with the risk that manually entered information could contain errors or even be incorrect.

After observing several intrinsic characteristics of API during the first two years of the system’s implementation, the RFB developed a brand new software for API data analysis. Using a broad range of algorithms, the software is able to analyse received API messages, identify travellers, create a unique traveller archive, and store information on the border movements of every traveller in the archive. This functionality enables the system to automatically correct API that contains errors such as incorrect passport numbers, complete missing data, and solve other issues that could lead to flawed conclusions.

Once travellers are selected for control, one of the most challenging tasks for Customs officers is to identify and segregate them without disrupting the flow of other travellers. In 2016, Brazilian Customs deployed a tailored solution that uses facial recognition technology at 14 international airports. Two high-resolution cameras have been installed in Customs’ “nothing to declare” line area at each airport. The images taken by the cameras are then processed using facial recognition software which scan a person’s face and analyses whether it matches against an image library of people considered as representing a risk and who must be controlled or prevented from entering Brazilian territory.

Biometric identification is carried out without human interference in the passenger line as they are moving at a walking speed. When the system identifies a passenger whose face matches the face of a target, a signal is sent to the Customs officer on duty, who will then approach the target and begin an inspection.

The system enables Customs officers to easily identify the targets pinpointed by the API/PNR risk assessment system without disrupting the general flow of passengers. After the implementation of the IRIS system in September 2016, it became possible to reduce the number of passengers subjected to Customs inspection upon arrival while increasing the overall success rates.

The following table compares the results before and after the system was implemented at São Paulo International Airport (GRU), where 65% of all the flow of international passengers takes place. The implementation of the IRIS system enabled the airport terminal’s overall arrival capacity to be increased as the lines at Customs were significantly reduced. While less people were inspected, the level of declarations increased as well as the quantity and value of seized goods.

The intelligence-based and risk assessment approach allows Customs border control to identify air passengers which are on watch lists as well as suspicious passengers, and is as such an effective measure to counter the movement of foreign terrorist fighters (FTFs) too.

To address security related risks, a special list of persons being targeted as a result of their association with a possible security threat was created. Customs, the Federal Police Department and the Brazilian Intelligence Agency are all responsible for the data included in the list. If the system identifies an individual, the Federal Police proceed to formally identify the suspect before taking appropriate measures.

The use of facial recognition tools is also particularly effective in cases involving fraudulent documents since the facial characteristics of the target remain mostly unchanged even if he/she travels on false documents. In a broader sense, Customs acts as a “second barrier” for national immigration control in order to further safeguard the country’s security.

More information
felipe.moraes@receita.fazenda.gov.br