Detection: Identify the presence of objects or shapes in an image (e.g., detecting faces, vehicles, industrial defects).Recognition: Classify or name detected objects (e.g., recognize a person, an animal, or a road sign).Tracking: Follow the movement of an object or person in a video sequence.Scene Analysis: Understand the overall context of an image (e.g., recognizing activities, analyzing behaviors).

Computer vision relies on several methods derived from:

Machine learning, and especially deep learning (convolutional neural networks or CNNs).Image processing, to improve quality, detect edges, or extract visual features.Geometric analysis and 3D modeling, to estimate the depth, position, or shape of objects.

Facial recognition and biometrics.Autonomous vehicles and driver assistance systems.Intelligent surveillance and security.Image-assisted medical diagnosis.Industry 4.0 (automated inspection, robotics).Augmented reality and human-machine interaction.

Computer vision now occupies a central place in the development of artificial intelligence and the digital transformation of societies. Its importance lies in its ability to allow machines to understand the visual world automatically and intelligently, paving the way for countless applications in scientific, industrial, medical, and social fields.

Computer vision can replace or assist human perception in repetitive, dangerous, or highly precise tasks.

Examples: product inspection on production lines, infrastructure monitoring, automated quality control, etc.

Result: reduced human error, improved productivity, and lower operating costs.

It plays a key role in public safety and physical cybersecurity, thanks to facial recognition, intrusion detection, and the analysis of suspicious behavior in sensitive areas.

This strengthens risk prevention and responsiveness to threats.

In autonomous vehicles, computer vision enables pedestrian detection, sign reading, and understanding of the road environment.

In the medical field, it assists doctors in diagnosis using images (X-rays, MRIs, CT scans), enabling faster and more accurate analyses.

It saves lives and improves the quality of care.

Computer vision promotes social inclusion: it helps visually impaired people orient themselves through object and text recognition, and facilitates human-machine communication through gestures and expressions.

By combining computer vision, deep learning, and robotics, we are paving the way for autonomous systems capable of understanding, learning, and interacting with their environment. This represents a fundamental pillar for smart cities, Industry 4.0, and intelligent security systems.

In summary:

The importance of computer vision lies in its ability to give machines “sight,” improve decision-making, and transform industries by making systems smarter, safer, and more efficient.

The terms safety and security both refer to measures aimed at protecting people, property, and systems, but they differ in the nature of the threats they address.

Security

Security (from the Latin securitas) concerns protection against intentional acts, that is, threats caused by human beings.

It aims to prevent, detect, and counter malicious or deliberate actions.

Examples:

Cyberattacks, hacking, data theft;Terrorism, sabotage, burglary;Fraud, scams, malicious manipulation.Objective: to prevent or limit damage caused by intentional acts.

Safety (from the Latin securitas, but in the sense of “stability”) concerns protection against unintentional accidents, failures, or errors.

It addresses technical, human, or environmental risks that can threaten a system without any intent to harm.

Examples:

Failure of industrial equipment or software;Transportation accident;Human error or power outage;Natural hazards (fires, floods, earthquakes).Objective: to guarantee the reliability, robustness, and continuity of operations despite unforeseen events.

In modern organizations, these two dimensions are inseparable.

An infrastructure (digital or physical) must be:

Safe, to function correctly and prevent accidents;Secure, to withstand attacks and intrusions.

Example: a computer network must be safe (without failures, well-configured) and secure (protected against cyberattacks).

In summary:

Safety = protection against accidents or unintentional errors.Security = protection against malicious and intentional acts.

Both are essential to ensure the overall protection of people, property, and systems.

Operational effectiveness refers to an organization’s ability to achieve its objectives optimally by efficiently utilizing its human, technical, and organizational resources.

In the context of cognitive security, this translates to how an institution, such as a university or a company, manages to integrate, disseminate, and sustainably promote best practices in cybersecurity among its members.

Cognitive security relies on users’ understanding, vigilance, and behavior in the face of digital threats.

It aims to raise awareness of risks (phishing, social engineering, digital negligence, etc.) and encourage safe habits in the use of technology.

Operational effectiveness, therefore, depends on the ability to bring about lasting changes in behavior.

Computer vision relies on several successive technical steps.

5.1.1. Image or Video Acquisition

Cameras, sensors, or surveillance systems capture visual streams from the environment (laboratories, computer rooms, access areas, etc.).

5.1.2. Image Preprocessing

Images are cleaned and optimized (contrast enhancement, noise reduction, resizing) to facilitate analysis.

5.1.3. Detection and Recognition

Deep learning algorithms (such as convoluted neural networks—CNCs) identify specific objects, faces, gestures, or behaviors.

5.1.4. Behavioral and Decision Analysis

The system interprets visual data to detect risky behaviors (e.g., unauthorized access, leaving workstations locked, failure to comply with safety rules).

5.1.5. Feedback and Continuous Improvement

The information collected can be used to inform awareness programs, adjust training, or trigger alerts in case of suspicious behavior.

Computer vision supports the dissemination of the best practices in cognitive security through several concrete applications:

1) Intelligent monitoring: automatically detecting behaviors that violate cybersecurity rules in sensitive areas (for example, someone leaving a connected workstation unattended).

2) Immersive training: using image and gesture recognition in educational simulations to train users to adopt good security practices.

3) Objective assessment: non-intrusively measuring the implementation of digital security guidelines in work or study environments.

4) Proactive protection: combining computer vision with alert systems to prevent human error or intrusions before they cause an incident.

A digital image can be modeled as a matrix:

where each p_ij represents the light intensity of the pixel at that position.

Image transformations (blurring, edge detection, filtering) use matrix operations such as convolution:

where K is a convolution kernel (filter).

To detect behaviors or objects (e.g., detecting an unauthorized user), computer vision extracts features from the image.

These features (edges, textures, points of interest) are represented by feature vectors:

These vectors feed into machine learning models (neural networks, SVMs, k-means) that learn to distinguish between conforming and risky behaviors.

Convolutional neural networks (CNNs) are widely used in computer vision.

Each layer of the CNN applies a mathematical transformation:

where:

W^(l) is the weight matrix,b^(l) is the bias,f is the activation function (ReLU, sigmoid, etc.),a^(l) is the output of layer l.

These models allow for the recognition of faces, gestures, or behaviors related to digital security (e.g., a user failing to activate MFA or connecting to an unsecured network) [6].

In a university environment integrating secure Wi-Fi, MFA, and Zero Trust:

Smart cameras, coupled with computer vision models, can analyze physical and digital behaviors (access, posture, interactions with workstations).Statistical models (such as Bayesian probabilities) can estimate the level of compliance of a behavior based on visual and access data.These observations then feed into cognitive awareness-raising mechanisms: notifications, reminders of good practices, adjustment of the level of access according to the observed behavior.

The mathematical formalism supports the Zero Trust principle, where trust is never implicit:

Each access or behavior is evaluated according to a risk score (RRR) calculated by:

where:

Cv = vision score (observed visual behavior),Cn = network score (activity on the secure Wi-Fi network),Cb = biometric score (MFA, facial recognition),αi\alpha_iαi = weighting coefficients.

If R > Rthreshold, access is denied or subject to enhanced authentication.

In summary:

The mathematical formalisms of computer vision enable us to:

Model and analyze visual behaviors related to cybersecurity;Integrate these analyses into a Zero Trust approach to dynamically adjust trust levels;Enhance cognitive security through detection, awareness, and intelligent feedback. Thus, the combination of computer vision, MFA, secure Wi-Fi, and Zero Trust architecture provides Denis Sassou Nguesso University with an integrated, proactive, and intelligent approach to cybersecurity.

IoT can play a major role in disseminating best practices in cognitive security by making security more visible, smarter, and more interactive.

8.1.1. Intelligent Monitoring and Behavioral Learning

IoT sensors can detect risky behaviors (unauthorized access, connection to an unsecured network, failure to log out of an account, etc.) and send real-time alerts.

8.1.2. Cognitive Feedback and Awareness

Connected devices (interactive whiteboards, smartphones, information displays) can broadcast personalized awareness messages when non-compliant behaviors are detected.

8.1.3. Contextual Risk Analysis

By combining data from connected devices, secure Wi-Fi, and MFA systems, the university can establish a cognitive risk profile for each user, in order to adapt security measures (e.g., strengthening authentication if unusual behavior is observed).

8.2.1. Secure Wi-Fi

The IoT relies on a high-performance and secure wireless network (WPA3, network segmentation, stream encryption).

Secure Wi-Fi guarantees the confidentiality of data exchanged by connected devices and limits the risk of intrusion into the university ecosystem [7].

8.2.2. Multi-Factor Authentication (MFA)

The IoT facilitates the implementation of contextual MFA:

Connected objects (bracelets, smartphones, RFID badges) can serve as physical or biometric authentication factors, strengthening user identity verification while improving the cognitive experience (perceived security and user engagement).

8.2.3. Zero Trust Architecture

In a Zero Trust model, each connected object must be authenticated, verified, and authorized before accessing the network.

The IoT provides behavioral data to feed trust algorithms:

where:

Id = device or user identity,Cn = network context (location, time, connection type),Au = observed activity or behavior.

If the confidence score T is below a defined threshold, multifactor authentication (MFA) or access restriction is applied.

At Denis Sassou Nguesso University, the IoT can be integrated into a three-dimensional cognitive security strategy:

8.3.1. Technical

Deployment of a network of secure, interoperable, and monitored connected objects.

8.3.2. Behavioral

Training users to understand the risks associated with connected objects and to adopt best practices (do not share passwords, avoid open networks, keep up to date).

8.3.3. Cognitive

Using the IoT to strengthen digital awareness through intelligent alerts, security reminders, and interactive training scenarios.

Detection of Abnormal Behavior (IoT and Users)

The CNN analyzes Wi-Fi data streams.It learns normal behavior.It detects abnormal behavior that blocks access.The CNN continuously monitors Wi-Fi network connections and MFA authentication.It can detect suspicious patterns.It helps strengthen cognitive awareness: are you in a Zero Trust architecture?It is part of the behavioral analysis engine.It classifies requests according to their risk level.It provides an adaptive trust score.

It integrates with a cognitive Zero Trust gateway.

Figure 1: Simulation Results and Analysis

Simulations performed using the CNN application architecture within a cognitive security framework for smart campuses demonstrate how sensors from different objects communicate with each other to collect networked data. The integration of the CNN architecture ensures the extraction of behavioral patterns, dimensionality reduction and classification (normal and abnormal), and the zero-module Trust + cognitive AI for adaptive access control, alerts, and automated actions to train and raise user awareness through the dissemination of best practices via IoT screens, emails, and notifications.

Figure 1. First application architecture of the CNN in a cognitive security framework within the smart campus.

The CNN acts as the cognitive brain at the heart of the security system.

Step 1: Data Collection

R (with)M (or)IoT Objects (c)Record (a)

Step 2: Processing by the CNN

The CNN model learns the usual behaviors.It adapts new data to known patterns.If it detects unusual or risky behavior.

AND

The system provides an appropriate response:

Notification or reminder of best practices.Mandatory MFA re-authentication.Invitation to attend awareness training.

Table 1: Results obtained after functional analysis of the CNN in campus cognitive safety

The CNN analyzes traffic to detect abnormal connection patterns, identifies suspicious authentication attempts and triggers preventative measures, feeds the Zero Trust gateway with risk scores for each user or device and triggers the dissemination of personalized advice based on the profile and observed behavior.

Table 1. Functioning of the CNN in campus cognitive security.

Figure 2: The second application architecture of the CNN in the cognitive security of the smart campus comprises secure Wi-Fi with multi-factor authentication (MFA) and internet-connected devices for data collection and preprocessing. It is combined with convolutional neural networks and cognitive artificial intelligence for anomaly detection and classification of normal and abnormal behavior. A zero-trust application with dynamic access control is used to initiate preventative actions (MFA, blocking).

Figure 2. Functioning of the CNN in campus cognitive security.

Table 2: CNN Results in the Cognitive Security of an Intruder (Student)

A student logs in from an unknown device. The CNN detects the abnormal pattern, and the system sends a message to “enable MFA to protect your account”. An IoT camera sends unusual traffic. The CNN detects the IoT anomaly, and the system isolates this sensor using the Zero Trust application and notifies the administrator.

Table 2. Results obtained from the CNN regarding the cognitive security of an intruder (student) gaining abnormal access to this smart campus.

Table 3: Key element and function in the operation of the CNN in Security

The convolutional neural network’s main role is to learn and detect risky behaviors. Secure Wi-Fi protects the wireless network. MFA verifies identities and strengthens trust. Zero Trust enables dynamic access control. Cognitive diffusion allows for the automatic training and awareness-raising of users.

Table 3. Key element and function in the operation of the CNN in the cognitive security of the campus.

The Denis Sassou Nguesso University aims to implement an intelligent cognitive cybersecurity platform:

Analyzes the digital behaviors of users and IoT devices connected to secure Wi-Fi;Automatically detects behavioral anomalies (with Broadcasts cognitive awareness messages).

The CNN model is the core analytical tool, identifying patterns of normal and abnormal behavior. A multi-layered CNN comprises:

1) An input layer.

2) Several convolutional layers.

3) Pooling layers.

4) One or more fully connected layers.

5) An output layer:

Consider a set of input data.

Table 5: The 3rd application architecture of the CNN for cognitive security (Wi-Fi, MFA, Zero Trust) of the smart campus with input data

After inputting WPA3 logs, IP addresses, session duration, and location to the secure Wi-Fi network, the CNN analyzes network traffic and detects anomalous connections. After inputting failed login attempts, device type, and connection frequency in the MFA (Multi-Factor Authentication) domain, the CNN detects suspicious login behavior. After inputting access data, behavioral signatures, and IoT traffic in the Zero Trust domain, the CNN assigns a dynamic trust score to each user/device.

Table 5. The 3rd application architecture of the CNN for cognitive security (Wi-Fi, MFA, Zero Trust) of the smart campus.

Figure 3: Conceptual Architecture

This secure Wi-Fi (WPA3) collects data from the external environment, including multi-factor authentication and associated logs from internet-connected devices, as well as a Zero Trust system that preprocesses and normalizes the data. It is equipped with a multi-layer convolutional neural network (CNN) that extracts behavioral characteristics, performs dimensionality reduction, and classifies the data. This results in a risk score and a Zero Trust alert.

Figure 3. Conceptual architecture.

11.2.1. Cognitive Interpretation and Intelligent Dissemination

The CNN powers a cognitive dissemination engine, which:

1) Analyzes detected abnormal behaviors;

2) Assesses the user’s cognitive risk level (low, medium, high);

3) Automatically disseminates appropriate messages:

Low level → reminder of a best practice;Medium level → MFA recommendation.

High level → temporary block + short online training.

11.2.2. Advantages of This Formalism

Table 6 presents the advantages Mathematical rigor enables precise and measurable supervised training.

Adaptability allows the CNN to learn and adapt to the local behaviors of campus users.Interoperability ensures compatibility with Zero Trust policies and MFA authentication.Augmented human cognition enables students and staff to receive personalized and automated security training.

Table 6. Advantages of this formalism.

Implementation of a smart digital infrastructure hub.Creation of an experimental prototype for university-level cognitive cybersecurity.Strengthening of digital resilience.Development of a shared security culture.Replication.Contribution to digital sovereignty and training in cognitive cybersecurity.

Scientific basis for future research in AI applied to behavioral security.

Table 7: Impact on Denis Sassou Nguesso University

The implementation of this infrastructure offers the following types of contributions:

On the technical side, there is the integration of secure Wi-Fi + MFA + Zero Trust to achieve a resilient and intelligent network;On the scientific side, there is multi-layered CNN modeling for cognitive security to detect behavioral anomalies;On the educational side, there is a cognitive dissemination system for best practices to provide ongoing user training;On the institutional side, there is improved cybersecurity governance to achieve a sustainable digital policy outcome.

Table 7. Impact on Denis Sassou Nguesso University.

15.1.1. Implementation Objectives

The main objective is to create a secure and intelligent digital ecosystem where:

Users automatically adopt best security practices.The Wi-Fi network is secure and segmented.Access is dynamically controlled via MFA and the Zero Trust principle.Cognitive cybersecurity is delivered in a personalized way to students, faculty, and administrative staff.

15.1.2. Technical Infrastructure: [9]

a) Secure Wi-Fi:

Security Protocols: WPA3-Enterprise for encryption.Network Segmentation: Separate VLANs for students, teachers, IoT, and administration.Monitoring: IDS/IPS to detect network anomalies.Access Control: RADIUS server to centralize identities.

b) Multi-Factor Authentication (MFA):

Integration with the LDAP/Active Directory server.MFA Methods: OTP (mobile application), codes via email, NFC badges for certain services.MFA logs are collected for behavioral analysis.

c) Zero Trust Architecture:

Principle: “Never trust, always verify”.Every user, device, or service is verified before and during access.Access policies adapt in real time according to the trust score generated by the cognitive model.

15.1.3. Cognitive Dissemination of Best Practices

a) Behavior Collection and Analysis:

Collection of Wi-Fi, MFA, and IoT logs.Real-time analysis via an artificial intelligence model (e.g., CNN) to detect:Normal behavior.Suspicious behavior.High-risk behavior.

b) Personalized Dissemination:

Notifications via email, captive portal, or interactive IoT screens.Messages tailored to the user profile and the severity of the situation:Simple tips on MFA and passwords.Micro-lessons on cognitive security.Alerts in case of suspicious login attempts.

c) Feedback Loop:

User actions are fed back into the system to improve the model.

Allows for continuous improvement of recommendations and awareness.

15.1.4. Functional Implementation Diagram: [10]

1) User Login → Secure Wi-Fi.

2) Identity Verification → MFA.

3) Behavioral Analysis → CNN Model/Zero Trust Engine.

4) Access Decision → Allowed/Enhanced MFA/Blocked.

5) Best Practice Dissemination → Personalized Messages, Notifications, Microlearning.

Table 8: Component and expected outcome of the implementation

The implementation of this intelligent infrastructure includes various components and their results, namely:

Secure Wi-Fi for reduced intrusions and improved traceability;MFA for enhanced authentication and reduced impersonation;Zero Trust for adaptive access and dynamic control;Cognitive diffusion for proactive awareness and adoption of best practices.

Table 8. Component and expected result of the implementation.