Executive summary
Several technologies associated with biometrics get many reactions; some are negative and others positive, just like facial recognition. Face recognition is considered very accurate when combined with artificial intelligence, and at the same time, it can be trespassing. This report focuses on applying artificial intelligence in facial recognition by looking at and examining two case studies examples. The first case study is on Unconstrained Facial Recognition Using Boston Marathon Bombing Suspects, used to identify suspects using photographs in the public domains matched against s background set mug shots from the facial recognition system (Amoore, 2019). The other case study example is on face recognition by the Us Government in Business cases. The cases involved in various businesses using face recognition include; facilitation, forensic examination, national security, and investigatory.
In conjunction with artificial intelligence, face recognition can verify or identify the focus by using images, videos, or audiovisual aspects of the observed face. Artificial intelligence has brought many developments in technology in industries. The industries have constantly been developing many innovations in the new digital technologies to use in different ways and their impacts on society. In the same way, it gave way to the development of face recognition that is now used for various reasons.
Introduction
Technology, especially artificial intelligence, has developed many valuable services used worldwide. Face recognition technology is mainly used for identification and recognition, primarily applying biometric identification, which uses body parts. This face recognition shall consider the head and face principally to confirm a person’s identity by matching their facial patterns and data (White, 2018). The face recognition technology collects unique biometric designs and data from individuals associated with their facial appearances. It uses the data to identify and authenticate if the individual is being scanned in the real one. The procedure for identification requires a digital photographic technology device to produce and obtain images and data that can create and save the facula patterns of an individual biometrically, which will be identified (Sarpatwar, Vaculin, Min, Su, Heath, Ganapavarapu and Dillenberger, 2019). Facial recognition technology plays a vital role in today’s lives, so we have to discuss it and get more insight into the technology. This report will look at the importance of facial recognition technology and its challenges. The recommendations based on the findings will also be provided to reduce the obstacles faced to better the technology. The field of facial recognition has wide chances that should be considered, especially in the automation of the system. Facial recognition has a challenge with the database because not everyone has an image in the system to be matched with the actual photo that has been taken. This makes it very difficult sometimes to recognize individuals or, in most cases, causes errors in the recognition.
Background
How facial recognition operates
Facial recognition technology is generally a set of algorithms that operate as one in identifying individuals in an image or a video. This technology has become a common one today and is still under innovations to develop better facial recognition in the future. The creation of facial recognition technology involves the integration of artificial intelligence, which is incorporated with face recognition systems (Andrejevic and Selwyn, 2020). Artificial intelligence is developed in software, comparing many faces in the database gathered on a specific scene (Bai, S.A., 2011). The technology provides highly accurate results. The artificial intelligence software has advantages: it allows for anti-spoofing measures, many cameras also use it, it has a real-time identification at the instance of the happenings, and it has no racial or gender bias, which is made possible by the many trials done across the world. Artificial intelligence is a machine that almost resembles human intelligence in performing operations (White, 2018). The artificial intelligence facial recognition technology applies a function called deep learning. Deep learning copies power processing and capabilities to create patterns in making very critical decisions (Murison, 2018). Deep learning is machine learning associated with artificial intelligence having a web that connects and learn from unlabeled and unstructured data with no supervision from other machines or human beings.
The operation of facial recognition starts with a tagged feature set of existing photos that have a hand-matched correlation with the involved individuals. For the face recognition process, there is a necessity to have the manual correlation of the face of the individual and the other identity of the person. ADoingso makes it easier for identification to be made to a person’s face who is being monitored. In some cases, the pictures are not always clear, and therefore, the data set is matched to the photo being studied (Anyoha, 2017). Artificial intelligence recognizes the face of individuals by searching on the data points that exist In the faces of people. The data points in people’s faces are the distance between the eyes, the distance between the mouth and the eye, the distance between a look and the other, and the distance between the nose and the mouth (Jeon, Jeong, Jee, Huang, Kim, Park, Kim, Wufuer, Jin, Kim and Choi, 2019). The artificial intelligence racial recognition technology can sometimes be fooled by individuals whenever the training is specific. The face recognition system cannot identify individuals with a mask or wearing a cap because the training level of the system was not established until that level (Marres, Cain, Gross, Kimbell, and Ulahannan, 2017). It means that for the system to be applied in the real-world context like humans, there must be intense training on the neural network to avoid being fooled.
Information about the area
Facial recognition systems are used in the modern world, especially in airports, railways stations, industrial parks, and the police, to help identify individuals passing through CCTV cameras. The cameras take pictures of individuals in either a video or a photo. The concept of individuals might appear alone or in a group; depending on the camera’s position, the image may appear straight or look nearly within the profile. The picture is then taken to facial recognition software, which will record the facial data of the face captured (Ezrachi and Stucke, 2017). The software can identify the facial data and the facial signature’s landmark. The facial signature is then compared to the facial signatures that exist in the database containing the faces of many individuals. Lastly, matching is done to recognize the beginning of the individual whose face was observed.
The operation of facial recognition technology becomes very difficult with its complexity. The legality of the use of face recognition must also be gotten by complying with the data protection law. Facial recognition is mainly used to identify criminal suspects from the public domain by law enforcement officers. It primarily enables the officers and the investigating officers to identify and take the suspects for court hearings (Racek, 2020). Using automated facial recognition systems is complex, and it requires skilled professionals who can operate and match the images to come up with the actual suspect.
Facial recognition is also used to find missing children reunited with their families. The police often trace the child who got lost from their last location or where they were last seen then use a reference photo or image to match their identity. The facial recognition in the missing children is used to confirm the disappeared child then used to trace the location and recover the missing child.
The central part I: Case Study of Examples
Example 1: Unconstrained facial recognition using the Boston Marathon Bombings Suspects
In 2013 two bombs killed three people at the finish line of the Boston Marathon, and other individuals who were 264 got injured. The bomb leads to an abrupt stoppage of the race to give the police an opportunity to corner the 12-block crime scene. The F.B.I.’s investigations showed that the explosive was made of a pressure cooker compressed with nails and N.N.s that were probably sealed in a dark nylon backpack (Vishnoi, Bagga, Sharma, and Wani, 2018). The law enforcement officers gave directions to business premises to release videos, which were provided to the police in large numbers as evidence from the public domain. The photos and the videos were reviewed together with other evidence then the F.B.I. gave out the images and the videos of two suspects to help the public identify them. In the identification event, their aunt revealed brothers, which again provoked them to get involved in more criminal activities (Suchman, 2011). The media mention that the challenge in identifying criminals resulted from the failure of the facial recognition system. Even after the government had a vast database of images in their designs (Bucher, 2018), this case study focused on examining unconstrained facial recognition in the public images of the suspects that took part in the Boston Marathon bombing. The photographs of the suspects were matched with the existing set background of mugshots with the modern face recognition systems.
The case in the Boston Marathon Bombing was made to simulate the automated facial recognition using their state-of-the-art face recognition system and images published by the ministry of immigration in their database. The following parts explained how the dataset and the matching were obtained.
Five probe images were cropped from the original photos of the suspects in the scene. There was no prior processing done to the pictures before their enrollment, though the photographs seemed similar and gotten from one image, with one appearing to be modified from the other (Mordvintsev, Olah and Tyka, 2015). This variation in the appearance of the photos gives a challenge in that automation of the face detection, tracking, and estimation of quality and recognition of activities in an environment that is not controlled.
In the Boston marathon case study, images of the two suspects were provided with different locations and times for the photos we captured earlier. The gallery images were then added to background images of the mugshot photographs in the Pinellas County Sherriff’s Office (PCSO) database. The mugshots were gotten from the public through Florida’s Sunshine Laws.
The two state-of-art matches and Cognitec FaceVACS were identified depending on their top operations in the National Institute of Standards and Technology (N.I.S.T.). Matchers were then run in the accepted settings to admit the unconstrained images under investigation with no other parameter.
Face matching results were done using various methods, including; blind search, filtered search, and fused search. In the blind search, every image is compared against all the other photos in the gallery without considering the demographic information like gender and race, which are found in the gallery. The retrieval rankings were then recorded and displayed in a table.
The filtered search was done on every probe compared to images of the same demographic data. The number of pictures in the gallery reduced from a million ideas to less than 200 000 images. Lastly, the fused search had the match scores of images from different sources of the same suspect gathered and summed without measuring before positioning the gallery images (Vizoso, Vaz-Álvarez, and López-García, 2021). The fusion helps in giving the best retrieval ranking for the fused probes without being affected by the demographic filtering.
It was challenging to locate the suspects from the crowd in the video clip used to trace them. It took the officers much time to identify the suspects and match them with the existing photos in the database. There is a need to automate the facial recognition system so that the identification of the suspects is made instantly, which was also a challenge in the case at the Boston marathon.
Example 2: U.S. Government Face Recognition Business Use Cases
Businesses in the United States apply cases with specific definitions and are on point. Enterprises use various matters ranging from a single type to a transverse number in their respective mission types. The use cases focus on several mission’s including; national security, investigatory, child exploitation cases, facilitation, and forensic examination. (Lauterbach, 2019).
The face recognition technology is highly satisfied with every business case’s missions, mainly whenever the case needs facilitation; the program will depend on face recognition. It is noted that prohibiting face recognition will negatively affect various solutions, especially when the person under investigation has agreed to use face recognition technology.
Some businesses use the cases required of them by the government agencies, and other companies use the technology based on the order given by the executive agencies (Makridakis, 2017). No business is required to apply the face recognition system without permission from a legal organization in the U.S. Doing tests and evaluating the use of the methods is made possible by using components verified and accepted by the government, with most departments in the government using biometrics in the specific labs.
Some programs work based on the Privacy Act of 1974, or requirements by the statutory authorities have been taken through a Privacy Impact Assessment, Systems of Records Notice, Privacy Threshold Analysis, or any other transparent review. In every program that requires facilitation as a mission use case, face recognition is applied with the consent of those individuals involved in the use case.
In the United States, face recognition algorithms are used, and their performance is evaluated by the Commerce Department’s National Institute of Science and Technology (N.I.S.T.). Face recognition applies to matching and searches of people by using biometrics. The biometrics searches are either done based on an individual image against the stored biometric of the individual, an image of the person against a small group, or an individual against the gallery of images in the database. The face recognition technology compares images to give results aimed at investigative leads.
The first case ly different from the second one because in the Boston marathon, the authorities used the facial recognition system without getting consent from the suspects o be sure of the identity of the suspects. In the second case, face recognition was used in businesses, but permission must be sought from the investigated individual.
Main Part II: Ethical issues arising
In the Main part I, the report focused and discussed two cases studies on the application of facial recognition technology. The first case study is on Unconstrained Facial Recognition Using Boston Marathon Bombing Suspects, used to identify suspects using photographs in the public domains matched against s background set mugshots from the facial recognition system. The other case study example is on face recognition by the Us Government in Business cases. Both the instances faced their challenges because they are associated with ethnicity directly. I aim to discuss the ethical issues in this Main part II, linking both examined and debated cases.
Artificial intelligence has always been an engine to research around S.T.E.M for many years. Many consumers in the technological fields like Facebook and Google have become aware of the capabilities of technology. Artificial intelligence has been used and applied in industries, banking, health care, and manufacturing companies in the current world.
Data without consent
The operation of face recognition systems requires images used for training and testing the system. The photos must be taken from different backgrounds, colors, and angles so that the facial recognition system can identify ideas from many other sources and experiences (Sierra, 2011). Unfortunately, it is not clear to get photos used in facial recognition and biometric use in Europe without the consent of the individual being investigated. In the Boston marathon bomb, the images were retrieved from the dataset without the suspect’s permission, maybe because it was a security issue that had to be addressed differently (Lippi, Contissa, Jablonowska, Lagioia, Micklitz, Palka, Sartor, and Torroni, 2020). In the business use in the second case study, it is evident by the laws of the U.S. that consent must be given by the individual being investigated for the facial recognition to be used.
Vulnerable population
Many accounts exist on the face recognition issues ranging from the developments of traditional ways of discussing ethics and face recognition in the literature, but having a rapid increase of ethical knowledge developed from a policy perspective.
Most machine learning focusing on face recognition uses techniques that have manifested the current success of the award based on the arterially neutral networks. The characteristics of the artificial neural network give rise to ethical issues. They are opacity, not easy to predict, and need enormous datasets to be applied in facial recognition. They are using face recognition for biomedical clinical and biometric data groups. If, in any case, the system has to be used in these vulnerable groups, then consent must be given by the individuals to ensure that their rights are not violated or their images are used for the wrong purposes (Meissner, G., 2020). In the second case, this restriction is witnessed in the business area. The government has provided the regulations to protect the vulnerable group from exploitation by researchers and the business community. It is only suitable to use individuals’ biometrics if they accept being investigated or have their images matched with the ones in the database (Bhavani, 2020). It is challenging to identify the vulnerable group in the first case study because the suspects investigating are terrorists. Still, their actions were uncouth and unaccepted, so consent was not necessary for that case.
Privacy and data protection are the most frequently observed and studied ethical issues. Privacy and data protection are never the same, but considering Artificial Intelligence, the most critical privacy in the privacy of information and data protection is the prevention of information privacy from being exposed.
Aspects of Regulation
The integrity of facial recognition systems security aspects or business is advanced to have high standards. Face recognition is made not to be used for unnecessary activities like cyberbullying and manipulation of information or business stocks. The images in the dataset are not accessible by any other individual apart from the authority who has the access and the permission to access the dataset and conduct the necessary matching to identify an individual.
The face recognition system has been made transparent. The manipulation of any image and information is highly condemned because it may lead to the wrongful conviction or identification of a wrong person who got lost (Anyoha and Rockwell, 2017). The Boston marathon case was regulated because only the investigating authorities had access to the dataset to do the matching, not any other individual from the public. This makes it easy to do investigation, and manipulation is also tricky. In the second case, regulation is made so that only those given permission and licensed by the U.S. government can use face recognition and biometrics. This makes it known to the public the businesses are allowed to use the honor and not any other business case.
Conclusion
Technology and especially artificial intelligence have become part of human livelihood and should be taken and accepted by the world. Ai makes life very easy, and many innovations are still underway, especially in robotics (Reynoso and Rebecca,2019). Much attention needs to be taken by the technology industries to curb the challenges associated with artificial intelligence. This report focuses the face recognition systems. The two findings from the case studies were mainly on face recognition and its applications (Bunz, Mercedes, and Graham, 2018). The Boston marathon case has been supported with much literature, especially from the classwork, and it followed the proper criteria for identifying the suspects. The dataset was made clear from this case study to be very vulnerable and essential because it helped identify the suspects with the help of the relative who knew the suspects. The second case study is generally based on the regulation of face recognition and biometrics, which has a necessity to permission by the U.S. government.
Recommendations
There is a need ta o automate the face recognition system to identify suspects that were seen in the first case in the Boston marathon bomb. The automation would make it take less time to remember the suspects and bring them on board. The automation would also make it easier for the system to verify the images taken at the streets with those in the dataset without requiring much confirmation.
I would also recommend that few regulation bodies be incorporated in the U.S. so that permission or license to use face recognition is easily gotten but upholding the regulatory statutes.
There is also a need to do more innovations to improve the face recognition systems by using small hidden cameras so that the public is not aware of the cameras’ location to avoid manipulating the images from the public or the streets.
Reflection
This project and the report that I wrote have provided me with the necessary skills and knowledge that will help me during my practice as a professional. The whole of the report and the project, more so considering the case studies, provides me with skills in the research methods and develops critical thinking that can help me present reports during my professional journey.
Being that the project was being done as a group and the members had various roles in the group work, it provided me with skills on working with a team and collaborating to develop an excellent output in the final project and the report the collaboration helped me, especially in the report because. I was able to find various ideas and information that I gathered to help me develop a report that is up to standards. Research is always done in groups, and therefore, it means the teamwork prepared me for future research during my working life or even having to do personal research.
Finding the case studies, the facial recognition was complex, especially in the second case, because it was difficult to relate to contemporary artificial intelligence (Adams, Wurzburg, and Kerr, 2021). Discussions that I applied we applied as a group in this paper. I got various ideas and information from facilitative others by providing a collaborative environment. The collaborative environment allowed me to get the necessary resources and knowledge in the research paper.
The findings of my report were obtained from the relevant methodology that I developed without any cooked information or data (Zylinska, J., 2020). I am pleased to mention that, together with the team that I worked with, we could create definitive evidence that was in line with the research requirements. I believe that with the proofs provided, it came out very clearly that technology, especially face recognition, has a vast opportunity to develop a good livelihood for humans worldwide.
Being the team leader, I faced several challenges while conducting the research. The first challenge was that the study tended to be slow and expensive since the methodology we put in place demanded a lot from us. Similarly, searching for and finding previous work was a challenge that I encountered during the whole process. Some institution models prevented this new research since most stakeholders made decisions according to their own experiences. An encounter in the time of operation meant less time for the study, making it time-consuming to execute it.
I devised tactics to overcome these challenges that I faced during my research in that I made an exploration of the new opportunities for different ways of working (Adams, Wurzburg, and Kerr, 2021). I attained this by participating in capacity-building activities and enhancing technological and environmental aspects of the research. I also improved the communication strategies with my team members, making them stay motivated and goal-oriented towards the study. What I made differently was carrying out the impact of my research and trying to pull my findings into a more critical area. I also leveraged for more enormous opportunities as I looked at the bigger picture, figuring out the essential questions and neglecting the uncertainties that existed as I proceeded with my research.
On the other hand, I can provide the links I worked on during the examination. Additionally, I can relate all that I read in class and published in the academic documents and connect it to my research. This is simply because the content of my study lay between the publications and the research work that I gathered from some other sources.
List of references
Adams, E., Wurzburg, E. and Kerr, S., 2021. The tip of the iceberg: Immaterial labor, technoskepticism, and the teaching profession. Contemporary Issues in Technology and Teacher Education, 21(1), pp.126-154.
Amoore, L., 2019. Introduction: Thinking with Algorithms: Cognition and Computation in the Work of N. Katherine Hayles. Theory, Culture & Society, 36(2), pp.3-16.
Andrejevic, M. and Selwyn, N., 2020. Facial recognition technology in schools: Critical questions and concerns. Learning, Media and Technology, 45(2), pp.115-128.
Anyone, R., 2017. The history of artificial intelligence. Science in the News, 28.
Anyoha, Rockwell (2017) ‘The History of Artificial Intelligence,’ Science in the News, Special Edition on Artificial Intelligence, Harvard University [blogpost 28 August].
Bai, S.A., 2011. Artificial intelligence technologies in business and engineering.
Bhavani, G., 2020, July. Artificial Intelligence: Simulations in Audit Education. In 2020 11th International Conference on Computing, Communication and Networking Technologies (I.C.C.C.N.T.) (pp. 1-5). IEEE.
Bhavani, G., 2020, July. Artificial Intelligence: Simulations in Audit Education. In 2020 11th International Conference on Computing, Communication and Networking Technologies (I.C.C.C.N.T.) (pp. 1-5). IEEE.
Bucher, T., 2018. If… then: Algorithmic power and politics. Oxford University Press.
Bunz, Mercedes and Graham Meikle. 2018. ‘Seeing Things’. In Bunz, M. and G. Meikle, The Internet of Things. Cambridge: Polity, p. 68-91.
Ezrachi, A. and Stucke, M.E., 2017. Artificial intelligence & collusion: When computers inhibit competition. U. Ill. L. Rev., p.1775.
Jeon, B., Jeong, B., Jee, S., Huang, Y., Kim, Y., Park, G.H., Kim, J., Wufuer, M., Jin, X., Kim, S.W. and Choi, T.H., 2019. A facial recognition mobile app for patient safety and biometric identification: Design, development, and validation. J.M.I.R. mHealth and uHealth, 7(4), p.e11472.
Lauterbach, A., 2019. Artificial intelligence and policy: quo Vadis?. Digital Policy, Regulation, and Governance.
Lippi, M., Contissa, G., Jablonowska, A., Lagioia, F., Micklitz, H.W., Palka, P., Sartor, G. and Torroni, P., 2020. The force awakens Artificial intelligence for consumer law. Journal of artificial intelligence research, 67, pp.169-190.
Lin, T.C., 2019. Artificial intelligence, finance, and the law. Fordham L. Rev., 88, p.531.
Makridakis, S., 2017. The forthcoming Artificial Intelligence (A.I.) revolution: Its impact on society and firms. Futures, 90, pp.46-60.
Marres, N., Cain, R., Gross, A., Kimbell, L., and Ulahannan, A., 2017. Surfacing social aspects of driverless cars with creative methods. University of Warwick, Workshop Report, April.
Meissner, G., 2020. Artificial intelligence: consciousness and conscience. AI & SOCIETY, 35(1), pp.225-235.
Mordvintsev, A., Olah, C. and Tyka, M., 2015. Inceptionism: Going deeper into neural networks.
Murison, M. (2018) Emotional A.I.: Mei messaging assistant warns of personality changes |
Internet of Business [online]. Available from: https://internetofbusiness.com/mei-anomalydetection- ai-messaging/ (Accessed 28 December 2019).
Racek, L., 2020. Trestní odpovědnost umělé inteligence.
Reynoso, Rebecca (2019) ‘A complete history of Artificial Intelligence. G2 Learning Hub. https://learn.g2.com/history-of-artificial-intelligence
Sarpatwar, K., Vaculin, R., Min, H., Su, G., Heath, T., Ganapavarapu, G. and Dillenberger, D., 2019. Towards enabling trusted artificial intelligence via blockchain. In Policy-based autonomic data governance (pp. 137-153). Springer, Cham.
Sierra, C. ed., 2011. Ramon Llull, from the Ars Magna to Artificial Intelligence. Barcelona: Artificial Intelligence Research Institute.Sobon, R., 2022. A.I. & Arms Control: The Cold War Continued.
Suchman, L., 2011. Anthropological relocations and the limits of design. Annual review of anthropology, 40, pp.1-18.
Vishnoi, S.K., Bagga, T.E.E.N.A., Sharma, A.A.R.U.S.H.I. and Wani, S.N., 2018. Artificial Intelligence enabled marketing solutions: A Review. Indian Journal Of Economics & Business, pp.167-177.
Vizoso, Á., Vaz-Álvarez, M. and López-García, X., 2021. Fighting Deepfakes: Media and Internet Giants’ Converging and Diverging Strategies Against Hi-Tech Misinformation. Media and Communication, 9(1), pp.291-300.
White, G. (2018) Child advice chatbots fail sex abuse test [online]. Available from: https://www.bbc.co.uk/news/technology-46507900 (Accessed 28 December 2019).
Zylinska, J., 2020. A.I. art: machine visions and warped dreams (p. 181). Open Humanities Press.
write