Applications of computer science

5.1Internet of Things (IoT)Often examined

You already use devices that talk to the Internet: a smartphone, a smart watch or a smart TV. The Internet of Things (IoT) takes this idea to ordinary objects. Fans, lights, water meters, cars, farm sensors and hospital beds are fitted with sensors and a network connection, so they can collect data, share it and act on it without a person doing every step.

Internet of Things (IoT)

A network of physical devices having sensors (such as smartphones, smart watches, smart security systems, smart home appliances, smart TVs and smart doors) that are connected through the Internet so that they can communicate with each other.

Sensor

A device that measures a physical quantity (temperature, humidity, gas level, water level, motion, heart rate) and turns it into data.

Actuator

A device that acts on a command, for example a motor, a valve, a relay switch or an alarm.

The book says IoT has a big impact on modern society. In Pakistan it is at an early stage of planning and implementation. Using IoT in homes, offices, industries, the government and the private sector improves productivity and efficiency, and lets people do daily activities with convenience and comfort.

How an IoT system works

Every IoT application, whether for a farm or a hospital, follows the same four-part flow shown below. The sensors collect data, a network carries it, the cloud stores and analyses it, and an application shows the result to people or sends a command back to an actuator.

1. Devices (perception): sensors collect real-time data; actuators perform actions.
2. Network (connectivity): Wi-Fi, mobile data (4G/5G), Bluetooth or RFID carry the data to a gateway or central hub.
3. Cloud (processing): the data is stored and analysed, often with machine learning, to find patterns, predict events and raise alerts.
4. Application (user): a mobile app or web dashboard shows the results, sends alerts and lets the user control devices remotely.

Why Pakistan needs IoT

• Enhanced automation: machines and appliances work with little human effort.
• Better control: devices can be monitored and controlled from anywhere.
• Comfort and convenience in homes and offices.
• Economical use of energy: lights, fans and air conditioners run only when needed, which matters in a country with costly electricity.

5.2Designing IoT-based applicationsOften examined

There are billions of Internet-connected devices in the world and the number is rising quickly. IoT connects devices, machines and people through the Internet, and this has produced new ideas in many industries. Designing an IoT application means using IoT technology inside an application built for a specific purpose. The book gives two designs that combine three technologies: IoT (to collect data), cloud computing (to store and analyse it) and Blockchain (to keep the records safe and honest).

IoT-based air pollution monitoring

Big cities of Pakistan suffer serious air pollution from industrial emission, vehicle emission, brick kilns and seasonal crop burning. It causes respiratory and heart diseases. An IoT system joined with cloud computing and Blockchain works in three stages.

1. IoT-based real-time data collection using sensors. A network of IoT sensors is placed across a polluted region. They measure pollutants such as carbon monoxide ($\mathrm{CO}$) and nitrogen dioxide ($\mathrm{NO}{}_2$) and conditions such as temperature and humidity. The readings go by wireless communication (for example Wi-Fi) to a central network hub, which sends them to the cloud.
2. Cloud platform for data storage and processing. A cloud platform such as AWS or Google Cloud stores the massive amount of data, including historical data for trends and forecasts. Machine learning and data visualization calculate the Air Quality Index (AQI) in real time. When pollution crosses a set level the platform sends alerts and notifications, and it can predict future air quality from past patterns.
3. Blockchain for transparency and security. A Blockchain ledger stores the readings so they are tamper-proof and cannot be manipulated, which keeps the data honest and authentic.

Finally the three parts are joined by defining data formats, communication protocols and security measures, and a user-friendly web dashboard or mobile app shows air quality, trends and alerts.

IoT-based flood management

Floods are common in some regions of Pakistan because rivers and lakes overflow after heavy rain. Lives and property are lost. A plan that improves forecasting, monitoring and quick response can reduce the damage. Joining IoT, cloud computing and Blockchain gives real-time monitoring, data analysis and secure information sharing, so that flood risk can be judged early, warnings can be given and the response can be planned.

1. IoT-based real-time data collection. Sensors in flood-prone areas monitor water level, rainfall, temperature, humidity, wind speed, wind direction and water temperature. Data is sent wirelessly to the cloud. This real-time collection is the core of the system.
2. Cloud computing for storage and processing. The cloud stores and analyses large volumes of flood data. Machine learning finds patterns and predicts flood risk. Computer vision and image processing make real-time maps of flood zones. When a reading crosses its upper limit, the cloud sends an alarm to the administrator for timely evacuation. Different agencies can coordinate through the same platform.
3. Blockchain for transparency and security. Blockchain keeps an immutable and transparent record of flood events. It can also track the movement of relief materials and the progress of recovery, so that aid reaches the people who need it.

5.3.1–5.3.4IoT applications in Pakistan: agriculture, smart grid, traffic and healthcareOften examined

Application of IoT in agriculture

Agriculture is the backbone of Pakistan's economy. The population is growing, so the demand for food is growing too. We need to move from traditional farming to smart agriculture based on IoT.

• Sensors observe real-time weather, so farmers have proper knowledge of the climate.
• Data on temperature, humidity, rainfall and soil condition makes farming more controlled, precise and efficient (for example, water is given only when the soil is dry).
• Drones monitor crop planting and crop health, spray fertilizer and pesticides, and help in irrigation.
• IoT-based farming reduces production cost and contributes to economic development.

Application of IoT in smart grid

Electricity in Pakistan is costly, often short, and demand keeps rising. The ordinary grid (transmission lines, transformers, substations and power grids) can be turned into a smart grid: the same infrastructure, but with IoT devices that communicate with each other and with consumers.

• More reliable supply through better fault detection.
• Less wastage: consumers get feedback on their electricity use.
• It analyses production and predicts when a shortage may occur, so companies can act in advance.
• It can automatically reroute electricity when a fault occurs, avoiding blackout of a whole area, and helps restore power quickly.
• In peak hours the IoT application advises consumers to reduce consumption.
• It generates more accurate bills.

Application of IoT in traffic management system

Traffic congestion in big cities grows with the population. Rush-hour jams waste fuel and time. A smart traffic management system uses IoT.

• Sensors and IoT devices at traffic lights, roads and highways collect real-time traffic data wirelessly and send it to traffic control rooms.
• Traffic flow is optimized, so congestion falls, road safety improves and the risk of accidents drops.
• Drivers and control rooms get real-time information about road conditions, congested areas and road hazards.
• Radio Frequency Identification (RFID) tags on cars open gated entrances automatically and save time.
• Surveillance cameras capture images of traffic rule violators so that authorities can take action.

Application of IoT in healthcare industry

IoT combined with artificial intelligence (AI) is changing healthcare. This has produced the Internet of Medical Things (IoMT): medical devices and wearables connected to the Internet. Heart problems, cancer and other diseases are increasing in Pakistan, so IoT-based healthcare should be used in all hospitals. It improves the productivity of doctors and the workflow of hospitals.

• Patients' critical data is collected and saved in the cloud; machine learning (ML) algorithms analyse it for patterns to diagnose disease and recommend medicines or tests.
• IoT devices track blood pressure, heart rate and oxygen level in real time.
• Smart wearables (smart watches, wristbands, rings, patches, eyeglasses, footwear, garments) allow remote patient monitoring at home, which saves hospital visits, time and money.
• Smart beds with sensors observe vital signs and alert the nurse or doctor when urgent attention is needed.
• Smart ambulances with sensors send the patient's critical data to the hospital during transport, so diagnosis and treatment can start inside the ambulance.

5.3.5–5.3.8IoT applications in Pakistan: manufacturing, homes, air pollution and disasters

Application of IoT in manufacturing industry

IoT combined with AI is transforming manufacturing. Pakistan's industry can use it to improve efficiency, productivity and worker safety and to reduce production cost.

• Automation and optimization of manufacturing processes.
• A network of sensors communicates wirelessly, collects and shares data and coordinates production.
• Robots increase accuracy and speed; workflow runs with minimal human intervention.
• Robots, sensors, cameras and AI-based machine learning improve product quality and reduce defective products, so customers are more satisfied.
• Fewer workplace accidents, because robots do risky and heavy manual tasks.
• Lower product cost through automation, efficiency, less labour cost and lower energy use by machines.
• IoT combined with Blockchain gives an efficient supply chain of products.

Application of IoT in homes

Smart homes are residences equipped with IoT-enabled home appliances connected to a central hub. The hub receives data from sensors and processes it to take decisions. Appliances such as TVs, ovens, washing machines, refrigerators, lights, fans, air conditioners and security cameras communicate wirelessly with each other and with the owner. Some IT companies in Pakistan already convert ordinary homes into smart homes.

• Convenient control of devices through a smartphone or web interface.
• Remote control and monitoring when the owner is away.
• Lower electricity bills: sensors detect whether a person is in the room and switch the lights or air conditioner on or off.
• Security: smart cameras, smart door locks and motion sensors alert the owner in real time.

Application of IoT to control air pollution

Air pollution from industries, vehicles and homes is a serious issue in the metropolitan cities of Pakistan. It causes heart disease, lung cancer, eye irritation and allergic reactions of the throat and eyes. An IoT-based monitoring system collects data from many sensors in real time and processes it to forecast pollution, show trends and control it. This allows an intelligent and quick response for the health and safety of the community (for example warning people, closing polluting kilns or diverting traffic).

Application of IoT in natural disaster management

Earthquakes, thunderstorms, floods and landslides after heavy rain are common in Pakistan. IoT helps to anticipate disasters and gives emergency responders the information they need, so that damage to infrastructure and loss of life are reduced.

• Real-time data collection and processing: IoT devices collect environmental data; the analysed data helps emergency authorities make the right decisions.
• Monitoring infrastructure: sensors embedded in dams, bridges and buildings monitor their condition, so action can be taken to protect them and save lives.
• Search and rescue: IoT with rescue robots and camera drones helps responders find and rescue people quickly after a disaster.

Pakistan National Disaster Management Authority (NDMA)

The NDMA makes and enforces national disaster policies at the federal and provincial levels. It works with government ministries, the Pakistan Army and United Nations organizations to coordinate disaster management, search and rescue and humanitarian operations in Pakistan and abroad.

The Pak NDMA Disaster Alert mobile app gives early warnings and advisories directly on users' phones. Its menu includes Pak NDMA alerts, hazard projections, safety guidelines and public service messages, and it shows alerts such as heatwaves and flash floods for each province. Its aim is to empower individuals and communities with vital information so that they can prepare for and respond to disasters.

5.4BlockchainOften examined

Think of a register in which a class records who paid the fee. If one person keeps it, that person could change an entry. If every student keeps an identical copy and a new page can be added only when everyone agrees, cheating becomes almost impossible. This is the idea of Blockchain.

Blockchain

A distributed ledger (database) of records or transactions that is shared among many network users. The ledger keeps growing as new records, called blocks, are added. Blocks are chained together using cryptography to increase security against fraud and to increase transparency.

Blockchain application

A decentralized application developed on Blockchain technology.

Distributed ledger

A database whose identical copies are kept by many computers (nodes) in a network instead of one central server.

Immutability

The property that once data is recorded on the Blockchain it cannot be changed.

How blocks are chained

Each block holds some data (for example votes or payments), a time stamp, its own hash and the hash of the previous block. A hash is a short code calculated from the block's contents by a cryptographic function; if even one character of the data changes, the hash changes completely.

Suppose someone changes the data in Block 2. Its hash changes, so it no longer matches the \"previous hash\" stored in Block 3, and the chain is broken. Also, all other nodes still hold the correct copy, so the false copy is rejected. That is why data on a Blockchain is tamper-proof.

Features and benefits

• Decentralized: no single owner controls the records.
• Security: cryptography links the blocks, which protects against data fraud and reduces vulnerability to cyber attacks.
• Transparency: every user of the network can see the same records.
• Immutability: recorded data cannot be changed.
• Reliability: it gives reliable solutions across many industries (voting, health, banking, education, supply chain, land records).

5.5Blockchain applications in PakistanOften examined

Application of Blockchain in voting

Pakistan's paper ballot system has faced allegations of rigging and manipulation. Blockchain can make voting secure and accurate. Each vote is recorded as a transaction on the ledger; because the ledger is transparent and immutable, no vote can be added, deleted or changed without detection. Voters can be confident that their vote is counted and the result is accurate. Blockchain voting can also be more efficient and cheaper.

Application of Blockchain in healthcare

Hospitals in Pakistan and worldwide face data security problems. Blockchain can change the way healthcare data is stored. Sensitive medical data, diagnostic images and patient records are protected by cryptographic algorithms and immutability, which reduces the risk of data loss through unauthorized access and breaches. Patients get more privacy, and authorized doctors get transparent access to the records.

Application of Blockchain in banking industry

Online banking has made transfers and transactions easy and reduced trips to banks and ATMs, but it has also exposed banks to data breaches and cyber threats. Blockchain integrated with cyber security makes the operation of banks more secure, transparent, efficient and cost-effective and helps eliminate financial fraud, because every transaction is recorded permanently on a shared ledger that cannot be secretly altered.

Application of Blockchain in education

Degrees, diplomas and certificates are kept manually, so verification is slow and difficult and fake credentials are sometimes used. Keeping these records on a distributed ledger makes record-keeping efficient, secure and transparent, eliminates fake degrees and makes verification simple. Students' academic records and administrative processes also become more secure. Students gain complete control over their records, because they are saved on the Blockchain instead of on the servers of institutions.

5.6Cloud computingOften examined

When you save photos on Google Drive or watch a film on YouTube, the files are not on your phone. They are on powerful servers owned by a company and reached through the Internet. This is cloud computing.

Cloud computing

The use of hosted services such as servers, data storage, databases, networking and software over the Internet. The services are maintained by a cloud service provider and are available on demand.

Pay-as-you-go

A billing method in which the user pays only for the cloud resources actually used.

Users can save data, files, images, videos and software on the cloud and access them from anywhere with a laptop, desktop, mobile phone or tablet connected to the Internet.

Areas where cloud computing can be used in Pakistan

Advantages of cloud computing (summary)

• Access anywhere, any time with an Internet connection.
• Lower cost: no need to buy and maintain servers; pay only for what is used.
• Scalability: resources can be increased or reduced on demand.
• Collaboration: many people can work on the same files in real time.
• Backup and security are handled by the provider.

5.7Machine learning: the starting point

Neural networks and deep learning are different, but they are closely related. Both are subsets of machine learning (ML), so we start there.

Machine learning (ML)

A branch of artificial intelligence that teaches computers to learn from data inputs and experience, like humans, without direct (explicit) programming. ML algorithms train on data, recognize patterns in large databases and build models that automate decision-making and produce accurate output.

In ordinary programming a person writes the rules. In machine learning the computer is given many examples (data with the correct answers) and it finds the rules itself. For example, instead of writing rules for spam, we show the computer thousands of emails marked \"spam\" or \"not spam\" and it learns the pattern.

5.7.1Neural networksOften examined

The human brain has billions of neurons connected to each other. Each neuron receives signals, and if the signal is strong enough it passes a signal to the next neurons. A neural network copies this idea in a computer.

Neural network

A simulation of the working of the human brain, based on the functioning of neurons. It is a network of neurons that receive and send information throughout the network to make predictions and solve complex problems in a human-like manner.

Neuron (node)

A computational unit of a neural network.

Synapse

The connection (small gap) between layers of a neural network that allows information to pass from one layer to the next.

Layers of a neural network

A neural network processes information layer by layer. It has three kinds of layers of neurons (nodes).

• Input layer: one or more neurons. Each neuron receives raw data from the outside world (text, image, video, etc.) and represents one feature of the input data. The data is passed to the hidden layer.
• Hidden layers: one or more layers between the input and output layers. The number of hidden layers and the number of neurons in each depend on the complexity of the problem.
• Output layer: the final layer that produces the output (the final result). It can have one neuron or many. To classify an email as spam or not spam, one output neuron is enough; problems with several outputs need several output neurons.

What happens inside one neuron

$x_i$: inputs, $w_i$: weights, $b$: bias, $z$: weighted sum, $f$: activation function, $y$: output passed to the next layer.

Weights

Numerical values associated with the connections between neurons. A weight decides how much influence one neuron's output has on the next neuron's input. Weights act as coefficients that increase or decrease the importance of specific information. They are adjusted during training to reduce the difference between the prediction and the actual output, so that the network makes accurate predictions.

Biases

Numerical constants associated with each neuron and added to the neuron's output. They give adaptability (flexibility): a bias lets a neuron activate even when its input is below the usual threshold, which helps the network learn complex patterns and make better predictions.

Activation function

The function that decides whether a neuron should be activated, that is, whether its input is important for predicting the output. It applies a mathematical operation to the weighted inputs and gives a value. If the value reaches a threshold, the neuron is activated and its output goes to the next layer (or becomes the final output); otherwise the neuron is deactivated and passes nothing on.

Loss function

A function that measures how good the network is: how well the predicted outputs match the true outputs during learning. Training tries to minimize the loss, which means reducing the mistakes on the learning data.

5.7.1How a neural network is trainedOften examined

A neural network is not programmed with rules. It is an AI model that accepts a set of input data and learns to predict the output. It learns through two processes repeated many times: feedforward propagation and backpropagation.

Feedforward propagation

Information moves in the forward direction: from the input layer, through the hidden layers, to the output layer. Each neuron takes its inputs, processes them with the activation function, and passes its output to the next layer only if the value is above a certain threshold. Each hidden neuron repeats this until the output layer produces the final result (the prediction).

Backpropagation

The prediction is compared with the actual output. The loss function measures the difference. To reduce the loss, the weights are modified, working backward from the output layer, through the hidden layers, to the input layer. This makes the network perform better by reducing the difference between the actual and the predicted output.

Training shown in Python

This program trains a single neuron $y=wx+b$ on four examples that follow the rule $y=2x+1$. Each pass (epoch) does a feedforward step, measures the squared-error loss, and moves $w$ and $b$ a little in the direction that reduces the loss. This is backpropagation in its simplest form.

# Train one neuron y = w*x + b by feedforward and backpropagation
xs = [1, 2, 3, 4]
ys = [3, 5, 7, 9]          # true outputs (the rule is y = 2x + 1)
w, b = 0.0, 0.0            # start with random-like values
rate = 0.05                # learning rate

for epoch in range(1, 201):
    dw = db = loss = 0
    for x, y in zip(xs, ys):
        pred = w * x + b               # feedforward: make a prediction
        err = pred - y
        loss += err ** 2               # squared error
        dw += 2 * err * x              # backpropagation: how the loss
        db += 2 * err                  # changes with w and with b
    w -= rate * dw / len(xs)           # adjust weight and bias
    b -= rate * db / len(xs)           # to reduce the loss
    if epoch in (1, 50, 100, 200):
        print("epoch", epoch, "loss", round(loss / len(xs), 4))

print("w =", round(w, 2), " b =", round(b, 2))
print("prediction for x = 5:", round(w * 5 + b, 2))

epoch 1 loss 41.0
epoch 50 loss 0.0031
epoch 100 loss 0.0007
epoch 200 loss 0.0
w = 2.0  b = 0.99
prediction for x = 5: 11.01

The loss falls from $41$ to almost $0$, and the neuron discovers $w\approx 2$ and $b\approx 1$ by itself. Nobody told it the rule; it learned from the data.

5.7.2Real-life example: predicting the price of a houseOften examined

Many things affect the price of a house. To keep the example simple the book uses four inputs: area (square feet), number of bedrooms, distance to city centre (km) and house age. Each input has its own weight, and one output neuron adds the weighted inputs.

Training (feedforward and backpropagation on many houses whose prices are known) finds suitable weights. A large area and more bedrooms raise the price, so their weights are positive. More distance and more age lower the price, so their weights come out negative.

# One neuron: weighted sum model of Fig. 5.30
inputs = [1200, 3, 5, 10]            # area (sq ft), bedrooms, distance (km), age (years)
weights = [4000, 500000, -100000, -50000]   # learned by training (Rs per unit)

price = 0
for x, w in zip(inputs, weights):
    price = price + w * x           # add w1*x1 + w2*x2 + ...
print("Predicted price: Rs", price)

Predicted price: Rs 5300000

Check: $4000(1200)+500000(3)-100000(5)-50000(10)=4800000+1500000-500000-500000=5300000$ rupees.

5.7.3–5.7.5Deep learning and its applicationsOften examined

Deep learning and neural networks

Neural networks are the building blocks of machine learning; deep learning emphasizes the depth of the network, that is, many hidden layers. Deep learning models use the neural network architecture, so they are often called deep neural networks. Without neural networks there would be no deep learning.

Deep learning (DL)

An AI technology that uses deep neural networks (networks with multiple hidden layers) to enable computers to autonomously extract hidden patterns and features from huge data sets for making decisions. It gives human-like thinking ability for developing intelligent systems.

• Compared with ordinary machine learning, deep learning needs vast amounts of data, more training time and high-performance computers with a GPU (Graphical Processing Unit), which can process many data sets at the same time.
• Deep learning algorithms improve their decisions over time as more sample data is given.

Applications of deep learning

1. Natural Language Processing (NLP): lets machines understand human language, both text and speech, and give suitable replies. Examples: language translators, sentiment analysis, customer-service chatbots.
2. Healthcare: diagnosis of life-threatening diseases such as cancer and diabetes through medical imaging; acute disease detection, smart health monitoring, drug discovery and medical research.
3. Fraud detection: detects suspicious login attempts and informs the user. PayPal used DL-based predictive analytics to detect and prevent fraud.
4. Self-driving car: an autonomous vehicle that senses its environment with cameras, sensors and geo-mapping, calculates distances, detects traffic lights, recognizes signs and avoids pedestrians. Example: Tesla. Helpful for elderly and physically disabled people.
5. Robotics: robots with sensors, cameras and microphones, controlled by a computer, perform tasks with little or no human help: manufacturing, moving goods in warehouses, hospitality, packaging medicine, surgery, handling hazardous material, working in extreme temperature.
6. Entertainment: DL analyses browsing history, preferences, interests and behaviour to give personalized recommendations. Amazon, YouTube, Netflix and Spotify recommend movies, songs and videos this way.

5.7.6Relationship between AI, ML, neural network and DL

The four terms are distinct but closely related. Each one sits inside the one before it.

• Artificial intelligence (AI) is the core idea: creating machines that simulate human intelligence. Examples: robotics, computer vision, virtual assistants.
• Machine learning (ML) is a subset of AI that develops algorithms which learn from training data without explicit programming. Examples: email spam filtering, online fraud detection, product recommendation.
• Neural network is a subset of ML. It simulates the human brain to build algorithms that recognize patterns and make intelligent decisions. Examples: image recognition, stock market prediction, weather forecasting.
• Deep learning is a specialized form of ML that uses neural networks with multiple hidden layers, especially for large data sets. Examples: language translation, NLP, self-driving car, healthcare diagnostics, playing strategic games such as chess.

5.8.1–5.8.5Data sharing and data privacyOften examined

Data sharing

Data sharing

Making the same data available to multiple users, applications and organizations.

Benefits of data sharing

• Efficiency and collaboration within an organization, with business partners and with the wider community. It promotes transparency and research and lets businesses offer better services. For example, sharing health data helps medical research; sharing result data shows institutions' academic achievements.
• Better decision-making by policymakers, managers and teams, which is essential for good organizational performance.
• Customer relationships and marketing: products and services reach the right people with the right offer.
• Public policy: data shared by and among government agencies helps to make public policies, plan the delivery of services and improve citizens' lives.

Data privacy

Data privacy

The ability of a person to control his personal information and to know when, how and what personal information is shared with others.

Personal Identifiable Information (PII)

Information that identifies a person, such as name, ID (CNIC) number, email or postal address, birth date and contact number.

Personal Health Information (PHI)

Health records of a person, such as diseases, test reports and prescriptions.

Data privacy applies to sensitive data: PII, PHI and the financial information of businesses and organizations. Sensitive data about employees, customers and shareholders is important for the operations and finances of an organization.

Conflict between data sharing and data privacy

With growing Internet use, websites, apps and social media collect and store personal data to provide services, for example when we order food or shop online. Sharing has benefits but also risks. Shared data can face unauthorized access, misuse and theft, and it is hard to ensure it is used only for the right purpose. Illegal access can cause invasion of privacy, risk to personal safety and damage to reputation. So sharing needs openness while privacy needs limits: the two have conflicting requirements.

Misuse of personal data

Misuse of personal information can harm individuals, businesses and the whole society. Examples of misuse include identity theft and financial fraud, selling data to advertisers without consent, spam and scam messages, blackmail or harassment, and discrimination. Individuals, organizations and governments must maintain ethical standards and follow data protection regulations when collecting, sharing and processing personal information. Respecting others' privacy rights builds a trustworthy and safe digital environment.

Safeguards against information misuse

Governments have made data protection regulations.

• General Data Protection Regulation (GDPR): rules on how personal data of individuals in the European Union (EU) can be collected, stored and processed.
• California Consumer Privacy Act (CCPA): consumers must know what personal data is collected about them and get control over it, such as the right that it is not disclosed or sold to others.

More countries may pass such laws, because many people feel individuals do not have enough control over their personal data.

5.8.6Assessing policies to protect stakeholders in data privacy conflictsOften examined

A stakeholder is any person or group affected by a decision, for example customers, employees, managers, IT teams, legal advisors and the government. The book gives three steps to assess data privacy conflicts among stakeholders and find practical solutions.

Step 1: Identify the source of the conflict

• Inconsistencies in the organization's data governance policies.
• Conflicting standards and expectations about data quality (accuracy and reliability), security and accessibility between data users and providers.
• Ethical and social issues in collecting, using or sharing personal data.
• Different interpretations of privacy laws, unclear data-handling policies or conflicting interests of stakeholders.
• Misunderstanding of how data is being used.
• Different data protection laws and regulations in different countries and regions.

Data governance

How data is collected, stored and processed, and who can have access to what kind of data.

Step 2: Negotiate and compromise

• Stakeholders negotiate trade-offs and find common ground for mutual agreement.
• They hold dialogue to understand each party's priorities and limitations.
• They balance data access and usability with data protection and encryption.
• They balance data completeness and accuracy with data minimization and anonymization.

Data minimization

Limiting the collection of personal information to what is necessary to accomplish a specified purpose. Example: a pizza delivery app asks for name, phone and address, but not CNIC or date of birth.

Data anonymization

Altering or removing personally identifiable information from data sets to protect the privacy of individuals while the data can still be used for software testing and analytics. Example: removing house numbers from addresses so that people cannot be identified.

Step 3: Implement the solutions

• Adopt a data privacy framework that implements all aspects of the data governance strategy.
• Apply data privacy tools such as data encryption, data loss prevention and firewalls to control what is shared.
• Monitor data privacy compliance.
• Run training programs so employees at all levels support the policy.
• Take feedback from all stakeholders to check effectiveness and find areas for improvement.

5.8.7Developing data governance policies to reduce conflicts

Well-structured data governance policies are needed to handle data according to laws and regulations. Conflicting policies cause confusion, inefficiency and even legal problems. The book gives five steps to develop policies for data collection, availability, usability, integrity and security.

1. Assess context: find why and when the conflicting policies were made; check whether any are outdated; identify which parts can be revised and which are fixed by law; study each policy's objectives and compliance needs and how they match the organization's goals; involve stakeholders; assess the risks and benefits of each policy for privacy and security.
2. Identify the stakeholders: those affected by the policies and those with authority to change them (department heads, IT security teams, legal advisors and others). Their views build consensus and balanced policies.
3. Evaluate risks: consider security of sensitive information, the result of non-compliance with regulations and damage to reputation; consult stakeholders and legal advisors; document every decision and communicate it for transparency and accountability.
4. Propose solutions: analyse the strengths and weaknesses of each policy; suggest changing one policy, combining parts of two, or writing a new one; keep solutions specific and actionable, because unclear policies cause confusion.
5. Implement changes: update documents, communicate the changes to everyone, give training if needed, monitor the changes and collect feedback to improve data governance.

5.8.8OECD principles on data sharing and protection of privacy

The Organization for Economic Cooperation and Development (OECD) is an international organization of 37 member countries, established in Paris in 1961. It develops social and economic policies and international standards for better lives and works on trade, environment, technology and education. Pakistan is not a member but is regularly invited to its meetings.

The OECD principles on data sharing and privacy are:

1. Collection limitation: there should be limits to how much personal data is collected.
2. Data quality: personal data should be accurate and relevant to the purpose for which it is used.
3. Purpose specification: the use of personal data should be specified.
4. Use limitation: data should not be used for purposes other than those specified.
5. Security safeguards: data should be kept secure.
6. Openness: collection and use of personal data should not be kept secret from individuals.
7. Individual participation: individuals have the right to know who has their data, to have it communicated to them, to know why a request for their data is denied, and to have it corrected or erased.
8. Accountability: anyone who collects data should be held accountable for following these principles.

Formula sheet

Key terms

Internet of Things (IoT)

A network of physical devices with sensors (smartphones, smart watches, security systems, home appliances, smart TVs, smart doors) connected through the Internet to communicate with each other.

Sensor

A device that measures a physical quantity and converts it into data.

Air Quality Index (AQI)

A number calculated from pollutant readings that shows how polluted the air is.

Smart grid

An electricity grid that uses IoT devices communicating with each other and with consumers for reliable, efficient supply.

RFID (Radio Frequency Identification)

Tags that are read wirelessly; used on cars to open gated entrances automatically.

Internet of Medical Things (IoMT)

The use of IoT devices and wearables, combined with AI, in healthcare.

Smart home

A residence with IoT-enabled appliances connected to a central hub that receives sensor data and takes decisions.

NDMA

National Disaster Management Authority of Pakistan; formulates and enforces national disaster policies and runs the Pak NDMA Disaster Alert app.

Blockchain

A distributed ledger of records or transactions shared among many network users, in which blocks are chained together using cryptography.

Distributed ledger

A database whose identical copies are kept by many nodes of a network.

Immutability

The property that recorded data cannot be changed.

Cloud computing

The use of hosted services such as servers, storage, databases, networking and software over the Internet, maintained by a provider and available on demand.

Learning Management System (LMS)

A cloud-based platform that provides course material, assignments, assessments and announcements.

Telemedicine

Virtual consultation between patients and healthcare providers through video conferencing.

Machine learning (ML)

A branch of AI that teaches computers to learn from data and experience without direct programming.

Neural network

A simulation of the working of the human brain: a network of neurons that receive and send information to make predictions.

Neuron

A computational unit of a neural network.

Input layer

The layer whose neurons receive raw data; each neuron represents a feature of the input.

Hidden layer

A layer between input and output layers; the number of hidden layers depends on the complexity of the problem.

Output layer

The final layer that produces the output.

Synapse

The connection between layers of a neural network that allows information to pass from one layer to the next.

Weights

Numerical values on the connections between neurons that set the influence of one neuron's output on the next neuron's input.

Biases

Numerical constants associated with each neuron, added to its output to give flexibility.

Activation function

Decides whether a neuron is activated, that is, whether its input is important for predicting the output.

Loss function

Measures how well the network's predicted outputs match the true outputs; training minimizes it.

Feedforward propagation

Passing information forward from the input layer through hidden layers to the output layer.

Backpropagation

Comparing predicted and actual output with the loss function and adjusting weights from the output layer back to the input layer.

Deep learning (DL)

AI technology that uses deep neural networks (many hidden layers) to autonomously extract hidden patterns from huge data sets.

GPU

Graphical Processing Unit; processes many data sets at the same time, used for training deep learning models.

Natural Language Processing (NLP)

A deep learning application that lets machines understand and respond to human language in text and speech.

Data sharing

Making the same data available to multiple users, applications and organizations.

Data privacy

The ability of a person to control his personal information and know when, how and what is shared with others.

PII

Personal Identifiable Information: name, ID number, email or postal address, birth date, contact number.

PHI

Personal Health Information: health records of a person.

GDPR

General Data Protection Regulation: EU rules on collecting, storing and processing personal data.

CCPA

California Consumer Privacy Act: gives consumers knowledge and control of the personal data collected about them.

Data governance

How data is collected, stored and processed and who can access what kind of data.

Data quality

The accuracy and reliability of data.

Data minimization

Limiting the collection of personal information to what is necessary for a specified purpose.

Data anonymization

Altering or removing PII from data sets to protect privacy while allowing use in testing and analytics.

OECD

Organization for Economic Cooperation and Development: 37 member countries, established in Paris in 1961; Pakistan is not a member but is invited to meetings.

FBISE 2026 questions

Questions from the FBISE HSSC-II 2026 (1st Annual) papers that belong to this unit, answered the way an examiner expects.

Exam focus

1. How is a neural network trained? Explain feedforward and backpropagation. (long)
   Asked in both 2026 sets (Set B Q.6 OR as 5 marks, Set D Q.2 (xii) OR as 3 marks).
   Go to answer
2. What is Blockchain? Explain its applications in banking and education (or give one application). (long)
   Asked in both 2026 sets (Set D Q.5, Set B Q.2 (xiii)); the definition words also give MCQs.
   Go to answer
3. How IoT helps in healthcare and in traffic management. (long)
   Set B Q.5 and Set D Q.2 (xii) in 2026; learn smart wearables, smart beds, smart ambulances, RFID and cameras.
   Go to answer
4. How IoT can help control air pollution in Pakistan. (short)
   Book short Q3 was repeated as Set B Q.2 (xiv); the three-stage design can also come as a long question.
   Go to answer
5. What is deep learning? Describe two applications (NLP, self-driving car, robotics, healthcare). (long)
   Set D Q.4 OR and Set B Q.2 (xii) OR (robotics) in 2026.
   Go to answer
6. Cloud computing: definition and applications (business, education, healthcare). (short)
   Set D Q.2 (xiv) in 2026 and book short Q9.
   Go to answer
7. Layers of a neural network, weights, biases, activation and loss function. (short)
   Source of several MCQs (layers in both 2026 sets) and likely short questions.
   Go to answer
8. Neural network model to predict the price of a house or mobile phone. (long)
   Book long Q3; tests layers, weighted sum and training together.
   Go to answer
9. Data privacy, PII, data minimization and anonymization. (short)
   2026 MCQs on PII and on why sharing and privacy conflict; book short Q10.
   Go to answer
10. Reasons for data privacy conflicts and data governance steps to resolve them. (long)
    Three of the book's seven long questions come from 5.8.6 to 5.8.8.
    Go to answer
11. OECD principles on data sharing and privacy. (long)
    Book long Q7; a list of eight principles is easy to score.
    Go to answer