Methodology

Data Collection

Strategy

Researchers affiliated with the Security Vision project (refer to Credits section for details) have conducted systematic internet queries utilizing pertinent keywords associated with the central theme of the investigation (facial recognition, movement recognition, etc.) across different intervals from 2021 to 2024. The searches were conducted in Arabic, Bosnian/Croatian/Serbian, English, French, Italian, Mandarin Chinese, Russian, Spanish. Starting with initial results, researchers progressively expanded the research corpus through snowball sampling.

Categorization Strategy

The data categorization, encompassing factors such as deployment purpose, institution sector, and product technology type, evolved through an iterative process. Initially, categories were established for data collection. However, as new deployments were logged into the wiki, fresh categories naturally surfaced. Upon completion of data collection, all emergent categories underwent a thorough review. This revision aimed to eliminate duplicates and ensure coherence across the spectrum of identified categories.

Participatory data collection

After the initial launch of the Security Vision Database, the data is collected through participatory contributions, similarly to Wikipedia. To ensure accuracy and reliability of the information, all submissions are reviewed by Security Vision.

Referencing

All the information contained in this wiki can be traced back to its original source. All references are indicated in the pages, and collected in a Zotero library accessible here.

Data Categorization

In this section, we explain some of the most important choices and definitions when categorizing information in our database. We skip self-evident fields such as "start date" and "city".

Deployments

• Information Certainty. We've organized the information we gathered into three tiers of certainty.
  • "Documented" denotes facts solidly supported by reliable sources.
  • "Rumored" encompasses information reported as potentially true but lacking confirmation.
  • Lastly, "Speculative" pertains to deployments inferred from multiple factual pieces but lacking explicit confirmation.
• Deployment Status. While the categories are self-explanatory, we distinguished "stopped" deployments, which have been interrupted due to the intervention of a court or a watchdog for example, and deployments that have naturally "concluded", either as part of an experimentation, or because they have for example been replaced with other systems.
• Deployment Purpose. We distinguish the following purposes for which security vision systems are implemented:
  • Access Control. Systems used to authenticate individuals who can access an area (whitelist) or who are banned from it (blacklist).
  • Automated Payments. Systems using computer vision to connect to payment services.
  • Border Control. Systems used to identify travelers at the border.
  • Commercial Surveillance. Systems used to collect information and analyze customer behavior with the purpose of impacting sales.
  • Crime Prevention. Systems used to prevent or discourage criminal behavior
  • Criminal Investigations. Systems used after a crime has taken place in the context of criminal investigation (for example to identify individuals).
  • Crowd Management. Systems to gain insights on crowds and collective behaviour of people.
  • Fraud Prevention. Systems aimed at preventing fraud.
  • Fugitive Detection. Systems aimed at identifying fugitives from prisons.
  • Health Surveillance. Systems aimed at detecting diseases (through thermal monitoring) and/or the application of health measures (covid-19 mask detection for example).
  • Labor Surveillance. Systems aimed at detecting absenteeism from work.
  • Missing Person Recovery. Systems aimed at identifying missing persons.
  • Political Surveillance. Systems aimed at exercising political control over specific individuals.
  • Refugee Identification. Systems aimed at identifying refugees.
  • Student Surveillance. Systems aimed at detecting absenteeism from school or university.
  • Surveillance. Systems generally aimed at identifying and tracking individuals.
  • Targeted Advertising. A specific type of commercial surveillance, aimed at serving targeted advertisement to identified individuals.
  • Traffic Surveillance. Systems aimed at detecting traffic issues.
  • Voter & ID Registration. Systems aimed at identifying citizens or voters as part of national registries.
  • War Operations. Systems aimed at identifying individuals or other forms of control in the context of conflicts.
• Managed by (custodian). The institution in charge of the maintenance of the system.
• Used by Institutions who have access to the system.
• Dataset used The dataset used by the system.
• Software used The specific products used by the system.

Datasets

• Information Certainty. See above
• Dataset category. Datasets are divided between datasets mainly used for civilian administration purposes (identity cards, drivers licenses) and law enforcement purposes (registers of known criminals, refugees, etc).
• Developed by. The institution that has collected the information contained in the dataset and/or has made it available.
• Contents. Specifies the type of contents available in the dataset (Facial images, fingerprints, etc.)
• Owned by. Specifies who is the owner of the dataset. Please note that it might be a different organization from the custodian, i.e. the organization in charge of maintaining and running the dataset.
• Funding. Additional information on the cost of creating the dataset
• Images. Number of images contained in the dataset
• Individuals. Number of individuals contained in the dataset. There can be more images from the same individuals, hence the difference.

Institutions

Institution is the generic term we use to define any legal entity that is involved in security vision systems.

• Institution Type. We distinguish companies (which produce, sell and provide services of security vision systems), public institutions, such as national and local governments, which can be both producers or consumers of such systems, law enforcement bodies who are generally the main users of the systems and NGOs that are involved in different degrees of contestation of the systems.
• Institution Sector. This defines the general area of operation of the institution. We define sectors as follows:
  • Academic. Universities, research centers, applied science institutions
  • Advocacy. Institutions involved in advocacy: NGOs involved in advocacy campaigns, governmental watchdogs.
  • Administration. Bureaucracies involved in the general administration of public life.
  • Art. Artistic projects or institutions
  • Education. Institutions related to the schooling system.
  • Energy. Institutions involved in the production or regulation of energy (electricity, oil, etc.)
  • Finance. Financial institutions
  • Health. Institutions involved in public health, ministries, hospitals, clinics.
  • Humanitarian. Institutions involved in providing humanitarian aid and support.
  • Retail. Institutions that sell goods and services directly to consumers for personal or household use.
  • Security. Institutions whose primary objective is the provision of security services, be them private or public. Security systems companies, law enforcement agencies, etc. They might sell security technologies, but as part of a broader involved in the security sector. For example: Thales, the Chicago Police Department.
  • Technology. Institutions primarily focused on developing software or hardware applications. While they may offer security technologies, these are typically part of a wider array of technological applications, which can span across several other sectors (health, agriculture, education, etc.). For example: Amazon, Google.
  • Telecom. Institutions involved in the provision or services in or the regulation of telecommunications.
  • Transportation. Institutions involved in the provision or services in or the regulation of transportation: Harbors, airports, etc.

Products

• Developed by: The institution (company, university, ministry) that developed the product.
• Technology Type:
  • ANPR. Automated Number Plate Recognition. A system to automatically read and process cars via their number plates.
  • Anomaly Detection. Identification of unusual behavior or events in data that deviate from expected patterns.
  • Audio Recognition. Identification of audio signals, commonly used for speech recognition or sound classification.
  • Control and Command Center. Centralized facility for monitoring and managing security operations, often equipped with technology for real-time decision-making.
  • Crowd Management. Automated analysis of sensor data in order to regulate the movement and behavior of large gatherings.
  • Database Software. Applications used for organizing, storing, and retrieving structured data.
  • Ear Acoustics. Identification technology based on the echo sound characteristics from the ear canal of individuals.
  • Emotion Recognition. Technologies aimed at analyzing facial expressions, gestures, or vocal cues to infer emotional states.
  • Face Mask Recognition. Identification of individuals wearing masks, used for enforcing safety protocols or security measures during public health crises.
  • Facial Recognition. Biometric technology for identifying or verifying individuals based on facial features.
  • Facial Recognition - Age Recognition. Identification of a person's age range based on facial characteristics, used for demographic analysis or age-restricted access control.
  • Facial Recognition - Authentication. Verification of a person's identity based on facial features, often used for secure access to devices or facilities.
  • Facial Recognition - Gender Recognition. Determination of a person's gender based on facial characteristics, used for demographic analysis or targeted advertising for example.
  • Facial Recognition - Live. Real-time facial recognition technology for instant identification or monitoring of individuals.
  • Facial Recognition - Retrospective. Analysis of recorded video footage to retrospectively identify individuals using facial recognition algorithms.
  • Finger Vein Recognition. Biometric method for identifying individuals based on the patterns of veins in their fingers.
  • Fingerprint Recognition. Biometric authentication technique based on the unique patterns of ridges and valleys on a person's fingertips.
  • Gunshot Detection. Automated systems for identifying and locating gunfire using acoustic sensors or analysis of audio signatures.
  • Iris Recognition. Biometric identification method based on the unique patterns in the iris of the eye.
  • Movement Recognition. Detection and analysis of human movement patterns, used for for categorizing behavior or tracking individuals.
  • Object Detection. Identification and location of objects within images or video streams.
  • Object Tracking. Monitoring the movement of objects over time within a scene.
  • Pattern Recognition. Identification and interpretation of patterns within data, used for example in anomaly detection or image analysis.
  • People Tracking. Monitoring and tracing the movements of individuals within a given area, often used for crowd surveillance or personnel management.
  • Predictive Policing: Use of data analysis and predictive algorithms to anticipate and prevent criminal activity, aiming to determine resource allocation and crime prevention strategies.
  • Smartphone App: Mobile application software designed for smartphones.
  • Social Distancing Monitor. Device or system for enforcing social distancing measures in public spaces, often utilizing sensors or cameras to detect proximity between individuals.
  • Social Media Monitoring. Surveillance of social media platforms to gather intelligence, monitor public sentiment, or detect potential security threats.
  • Thermal Camera: Imaging device capable of detecting and visualizing heat signatures, useful for surveillance, perimeter security, and search and rescue operations.
  • Video Analysis. Automated analysis of video footage to extract relevant information, detect anomalies, or identify objects or individuals of interest.
  • Video Summarization: Condensation of lengthy video footage into shorter summaries or key highlights, facilitating review and analysis by security personnel.
  • Violence Detection. Systems for automatically identifying violent behavior or incidents within video streams, aiming to enhance public safety and emergency response.
  • Voice Recognition. Technology for identifying and interpreting spoken words or phrases, commonly used for authentication or voice-controlled systems.

What links here

• About the Security Vision Database ‎ (← links)