Docker Desktop is more than just a local application for containerized development — it’s your gateway to an integrated suite of cloud-native tools that streamline the entire development workflow. While Docker Desktop can be used without signing in, doing so unlocks the full potential of Docker’s powerful, interconnected ecosystem. By signing in, you gain access to advanced features and services across Docker Hub, Build Cloud, Scout, and Testcontainers Cloud, enabling deeper collaboration, enhanced security insights, and scalable cloud resources. 

This blog post explores the full range of capabilities unlocked by signing in to Docker Desktop, connecting you to Docker’s integrated suite of cloud-native development tools. From enhanced security insights with Docker Scout to scalable build and testing resources through Docker Build Cloud and Testcontainers Cloud, signing in allows developers and administrators to fully leverage Docker’s unified platform.

Note that the following sections refer to specific Docker subscription plans. With Docker’s newly streamlined subscription plans — Docker Personal, Docker Pro, Docker Team, and Docker Business — developers and organizations can access a scalable suite of tools, from individual productivity boosters to enterprise-grade governance and security. Visit the Docker pricing page to learn more about how these plans support different team sizes and workflows. 

Benefits for developers when logged in

Docker Personal

• Access to private repositories: Unlock secure collaboration through private repositories on Docker Hub, ensuring that your sensitive code and dependencies are managed securely across teams and projects.
• Increased pull rate: Boost your productivity with an increased pull rate from Docker Hub (40 pulls/hour per user), ensuring smoother, uninterrupted development workflows without waiting on rate limits. The rate limit without authentication is 10 pulls/hour per IP.
• Docker Scout CLI: Leverage Docker Scout to proactively secure your software supply chain with continuous security insights from code to production. By signing in, you gain access to powerful CLI commands that help prevent vulnerabilities before they reach production. 
• Build Cloud and Testcontainers Cloud: Experience the full power of Docker Build Cloud and Testcontainers Cloud with free trials (7-day for Build Cloud, 30-day for Testcontainers Cloud). These trials give you access to scalable cloud infrastructure that speeds up image builds and enables more reliable integration testing.

Docker Pro/Team/Business 

For users with a paid Docker subscription, additional features are unlocked.

• Unlimited pull rate: No Hub rate limit will be enforced for users with a paid subscription plan. 
• Docker Scout base image recommendations: Docker Scout offers continuous recommendations for base image updates, empowering developers to secure their applications at the foundational level and fix vulnerabilities early in the development lifecycle.

• Docker Debug: The docker debug CLI command can help you debug containers, while the images contain the minimum required to run your application.

Docker Debug functionalities have also been integrated into the container view of the Docker Desktop UI.

• Synchronized file shares: Host to Docker Desktop VM file sharing via bind mounts can be quite slow for large codebases. Speed up your development cycle with synchronized file shares, allowing you to sync large codebases into containers quickly and efficiently without performance bottlenecks—helping developers iterate faster on critical projects.

• Additional free minutes for Docker Build Cloud: Docker Build Cloud helps developer teams speed up image builds by offloading the build process to the cloud. The following benefits are available for users depending on the subscription plan. 
  • Docker Pro: 200 mins/month per org
  • Docker Team: 500 mins/month per org
  • Docker Business: 1500 mins/month per org

• Additional free minutes for Testcontainers Cloud: Testcontainers Cloud simplifies the process for developers to run reliable integration tests using real dependencies defined in code, whether on their laptops or within their team’s CI pipeline. Depending on the subscription plan, the following benefits are available for users:
  • Docker Pro: 100 mins/month per org
  • Docker Team: 500 mins/month per org
  • Docker Business: 1,500 mins/month per org

Benefits for administrators when your users are logged in

Docker Business

Security and governance

The Docker Business plan offers enterprise-grade security and governance controls, which are only applicable if users are signed in. As of Docker Desktop 4.35.0, these features include:

• Image Access Management and Registry Access Management: Enhance security governance by controlling exactly which images and registries your developers can access or push to, ensuring compliance with organizational security policies and minimizing risk exposure.
• Settings management: Enforce organization-wide security policies by locking default settings across Docker Desktop instances, ensuring consistent configuration and compliance throughout your team’s development environment.
• Enhanced Container Isolation: Protect developer environments from malicious containers with rootless containers and much more.
• Air-gapped containers: Restrict network activity originating from containers.
• Private extensions marketplace: Limit Docker Desktop extensions that developers can install.
• Kerberos and NTLM authentication for proxies: Centralize Docker Desktop authentication to network proxies without prompts.
• SOCKS5 proxy support: Enable the usage of SOCKS5 proxies for Docker Desktop.

License management

Tracking usage for licensing purposes can be challenging for administrators due to Docker Desktop not requiring authentication by default. By ensuring all users are signed in, administrators can use Docker Hub’s organization members list to manage licenses effectively.

This can be coupled with Docker Business’s Single Sign-On and SCIM capabilities to ease this process further. 

Insights

Administrators and other stakeholders (such as engineering managers) must comprehensively understand Docker Desktop usage within their organization. With developers signed into Docker Desktop, admins gain actionable insights into usage, from feature adoption to image usage trends and login activity, helping administrators optimize team performance and security. A dashboard offering insights is now available to simplify monitoring. Contact your account rep to enable the dashboard.

Enforce sign-in for Docker Desktop

Docker Desktop includes a feature that allows administrators to require authentication at start-up. Admins can ensure that all developers sign in to access Docker Desktop, enabling full integration with Docker’s security and productivity features. Sign-in enforcement helps maintain continuous compliance with governance policies across the organization.

Developers can then click on the sign-in button, which takes them through the authentication flow. 

More information on how to enforce sign-in can be found in the documentation. 

Unlock the full potential of Docker’s integrated suite

Signing into Docker Desktop unlocks significant benefits for both developers and administrators, enabling teams to fully leverage Docker’s integrated, cloud-native suite. Whether improving productivity, securing the software supply chain, or enforcing governance policies, signing in maximizes the value of Docker’s unified platform — especially for organizations using Docker’s paid subscription plans.

Note that new features are introduced with each new release, so keep an eye on our blog and subscribe to the Docker Newsletter for the latest product and feature updates.

Up next

• Read Announcing Upgraded Docker Plans: Simpler, More Value, Better Development and Productivity.
• Learn more about the new Docker subscriptions.
• Get started with Docker Desktop.
• Subscribe to the Docker Newsletter.

When testing complex systems, the more edge cases you can identify, the better your software performs in the real world. But how do you efficiently generate hundreds or thousands of meaningful tests that reveal hidden bugs? Enter model-based testing (MBT), a technique that automates test case generation by modeling your software’s expected behavior.

In this demo, we’ll explore the model-based testing technique to perform regression testing on a simple REST API.

We’ll use the jqwik test engine on JUnit 5 to run property and model-based tests. Additionally, we’ll use Testcontainers to spin up Docker containers with different versions of our application.

Model-based testing

Model-based testing is a method for testing stateful software by comparing the tested component with a model that represents the expected behavior of the system. Instead of manually writing test cases, we’ll use a testing tool that:

• Takes a list of possible actions supported by the application
• Automatically generates test sequences from these actions, targeting potential edge cases
• Executes these tests on the software and the model, comparing the results

In our case, the actions are simply the endpoints exposed by the application’s API. For the demo’s code examples, we’ll use a basic service with a CRUD REST API that allows us to:

• Find an employee by their unique employee number
• Update an employee’s name
• Get a list of all the employees from a department
• Register a new employee

Once everything is configured and we finally run the test, we can expect to see a rapid sequence of hundreds of requests being sent to the two stateful services:

Docker Compose

Let’s assume we need to switch the database from Postgres to MySQL and want to ensure the service’s behavior remains consistent. To test this, we can run both versions of the application, send identical requests to each, and compare the responses.

We can set up the environment using a Docker Compose that will run two versions of the app:

• Model (mbt-demo:postgres): The current live version and our source of truth.
• Tested version (mbt-demo:mysql): The new feature branch under test.

services:
  ## MODEL
  app-model:
      image: mbt-demo:postgres
      # ...
      depends_on:
          - postgres
  postgres:
      image: postgres:16-alpine
      # ...
      
  ## TESTED
  app-tested:
    image: mbt-demo:mysql
    # ...
    depends_on:
      - mysql
  mysql:
    image: mysql:8.0
    # ...

Testcontainers

At this point, we could start the application and databases manually for testing, but this would be tedious. Instead, let’s use Testcontainers’ ComposeContainer to automate this with our Docker Compose file during the testing phase.

In this example, we’ll use jqwik as our JUnit 5 test runner. First, let’s add the jqwik and Testcontainers and the jqwik-testcontainers dependencies to our pom.xml:

<dependency>
    <groupId>net.jqwik</groupId>
    <artifactId>jqwik</artifactId>
    <version>1.9.0</version>
    <scope>test</scope>
</dependency>
<dependency>
    <groupId>net.jqwik</groupId>
    <artifactId>jqwik-testcontainers</artifactId>
    <version>0.5.2</version>
    <scope>test</scope>
</dependency>
<dependency>
    <groupId>org.testcontainers</groupId>
    <artifactId>testcontainers</artifactId>
    <version>1.20.1</version>
    <scope>test</scope>
</dependency>

As a result, we can now instantiate a ComposeContainer and pass our test docker-compose file as argument:

@Testcontainers
class ModelBasedTest {

    @Container
    static ComposeContainer ENV = new ComposeContainer(new File("src/test/resources/docker-compose-test.yml"))
       .withExposedService("app-tested", 8080, Wait.forHttp("/api/employees").forStatusCode(200))
       .withExposedService("app-model", 8080, Wait.forHttp("/api/employees").forStatusCode(200));

    // tests
}

Test HTTP client

Now, let’s create a small test utility that will help us execute the HTTP requests against our services:

class TestHttpClient {
  ApiResponse<EmployeeDto> get(String employeeNo) { /* ... */ }
  
  ApiResponse<Void> put(String employeeNo, String newName) { /* ... */ }
  
  ApiResponse<List<EmployeeDto>> getByDepartment(String department) { /* ... */ }
  
  ApiResponse<EmployeeDto> post(String employeeNo, String name) { /* ... */ }

    
  record ApiResponse<T>(int statusCode, @Nullable T body) { }
    
  record EmployeeDto(String employeeNo, String name) { }
}

Additionally, in the test class, we can declare another method that helps us create TestHttpClients for the two services started by the ComposeContainer:

static TestHttpClient testClient(String service) {
  int port = ENV.getServicePort(service, 8080);
  String url = "http://localhost:%s/api/employees".formatted(port);
  return new TestHttpClient(service, url);
}

jqwik

Jqwik is a property-based testing framework for Java that integrates with JUnit 5, automatically generating test cases to validate properties of code across diverse inputs. By using generators to create varied and random test inputs, jqwik enhances test coverage and uncovers edge cases.

If you’re new to jqwik, you can explore their API in detail by reviewing the official user guide. While this tutorial won’t cover all the specifics of the API, it’s essential to know that jqwik allows us to define a set of actions we want to test.

To begin with, we’ll use jqwik’s @Property annotation — instead of the traditional @Test — to define a test:

@Property
void regressionTest() {
  TestHttpClient model = testClient("app-model");
  TestHttpClient tested = testClient("app-tested");
  // ...
}

Next, we’ll define the actions, which are the HTTP calls to our APIs and can also include assertions.

For instance, the GetOneEmployeeAction will try to fetch a specific employee from both services and compare the responses:

record ModelVsTested(TestHttpClient model, TestHttpClient tested) {}

record GetOneEmployeeAction(String empNo) implements Action<ModelVsTested> {
  @Override
  public ModelVsTested run(ModelVsTested apps) {
    ApiResponse<EmployeeDto> actual = apps.tested.get(empNo);
    ApiResponse<EmployeeDto> expected = apps.model.get(empNo);

    assertThat(actual)
      .satisfies(hasStatusCode(expected.statusCode()))
      .satisfies(hasBody(expected.body()));
    return apps;
  }
}

Additionally, we’ll need to wrap these actions within Arbitrary objects. We can think of Arbitraries as objects implementing the factory design pattern that can generate a wide variety of instances of a type, based on a set of configured rules.

For instance, the Arbitrary returned by employeeNos() can generate employee numbers by choosing a random department from the configured list and concatenating a number between 0 and 200:

static Arbitrary<String> employeeNos() {
  Arbitrary<String> departments = Arbitraries.of("Frontend", "Backend", "HR", "Creative", "DevOps");
  Arbitrary<Long> ids = Arbitraries.longs().between(1, 200);
  return Combinators.combine(departments, ids).as("%s-%s"::formatted);
}

Similarly, getOneEmployeeAction() returns an Aribtrary action based on a given Arbitrary employee number:

static Arbitrary<GetOneEmployeeAction> getOneEmployeeAction() {
  return employeeNos().map(GetOneEmployeeAction::new);
}

After declaring all the other Actions and Arbitraries, we’ll create an ActionSequence:

@Provide
Arbitrary<ActionSequence<ModelVsTested>> mbtJqwikActions() {
  return Arbitraries.sequences(
    Arbitraries.oneOf(
      MbtJqwikActions.getOneEmployeeAction(),
      MbtJqwikActions.getEmployeesByDepartmentAction(),
      MbtJqwikActions.createEmployeeAction(),
      MbtJqwikActions.updateEmployeeNameAction()
  ));
}


static Arbitrary<Action<ModelVsTested>> getOneEmployeeAction() { /* ... */ }
static Arbitrary<Action<ModelVsTested>> getEmployeesByDepartmentAction() { /* ... */ }
// same for the other actions

Now, we can write our test and leverage jqwik to use the provided actions to test various sequences. Let’s create the ModelVsTested tuple and use it to execute the sequence of actions against it:

@Property
void regressionTest(@ForAll("mbtJqwikActions") ActionSequence<ModelVsTested> actions) {
  ModelVsTested testVsModel = new ModelVsTested(
    testClient("app-model"),
    testClient("app-tested")
  );
  actions.run(testVsModel);
}

That’s it — we can finally run the test! The test will generate a sequence of thousands of requests trying to find inconsistencies between the model and the tested service:

INFO com.etr.demo.utils.TestHttpClient -- [app-tested] PUT /api/employeesFrontend-129?name=v
INFO com.etr.demo.utils.TestHttpClient -- [app-model] PUT /api/employeesFrontend-129?name=v
INFO com.etr.demo.utils.TestHttpClient -- [app-tested] GET /api/employees/Frontend-129
INFO com.etr.demo.utils.TestHttpClient -- [app-model] GET /api/employees/Frontend-129
INFO com.etr.demo.utils.TestHttpClient -- [app-tested] POST /api/employees { name=sdxToS, empNo=Frontend-91 }
INFO com.etr.demo.utils.TestHttpClient -- [app-model] POST /api/employees { name=sdxToS, empNo=Frontend-91 }
INFO com.etr.demo.utils.TestHttpClient -- [app-tested] PUT /api/employeesFrontend-4?name=PZbmodNLNwX
INFO com.etr.demo.utils.TestHttpClient -- [app-model] PUT /api/employeesFrontend-4?name=PZbmodNLNwX
INFO com.etr.demo.utils.TestHttpClient -- [app-tested] GET /api/employees/Frontend-4
INFO com.etr.demo.utils.TestHttpClient -- [app-model] GET /api/employees/Frontend-4
INFO com.etr.demo.utils.TestHttpClient -- [app-tested] GET /api/employees?department=ٺ⯟桸
INFO com.etr.demo.utils.TestHttpClient -- [app-model] GET /api/employees?department=ٺ⯟桸
        ...

Catching errors

If we run the test and check the logs, we’ll quickly spot a failure. It appears that when searching for employees by department with the argument ٺ⯟桸 the model produces an internal server error, while the test version returns 200 OK:

Original Sample
---------------
actions:
ActionSequence[FAILED]: 8 actions run [
    UpdateEmployeeAction[empNo=Creative-13, newName=uRhplM],
    CreateEmployeeAction[empNo=Backend-184, name=aGAYQ],
    UpdateEmployeeAction[empNo=Backend-3, newName=aWCxzg],
    UpdateEmployeeAction[empNo=Frontend-93, newName=SrJTVwMvpy],
    UpdateEmployeeAction[empNo=Frontend-129, newName=v],
    CreateEmployeeAction[empNo=Frontend-91, name=sdxToS],
    UpdateEmployeeAction[empNo=Frontend-4, newName=PZbmodNLNwX],
    GetEmployeesByDepartmentAction[department=ٺ⯟桸]
]
    final currentModel: ModelVsTested[model=com.etr.demo.utils.TestHttpClient@5dc0ff7d, tested=com.etr.demo.utils.TestHttpClient@64920dc2]
Multiple Failures (1 failure)
    -- failure 1 --
    expected: 200
    but was: 500

Upon investigation, we find that the issue arises from a native SQL query using Postgres-specific syntax to retrieve data. While this was a simple issue in our small application, model-based testing can help uncover unexpected behavior that may only surface after a specific sequence of repetitive steps pushes the system into a particular state.

Wrap up

In this post, we provided hands-on examples of how model-based testing works in practice. From defining models to generating test cases, we’ve seen a powerful approach to improving test coverage and reducing manual effort. Now that you’ve seen the potential of model-based testing to enhance software quality, it’s time to dive deeper and tailor it to your own projects.

Clone the repository to experiment further, customize the models, and integrate this methodology into your testing strategy. Start building more resilient software today!

Thank you to Emanuel Trandafir for contributing this post.

Learn more

• Clone the model-based testing practice repo.
• Subscribe to the Docker Newsletter. 
• Visit the Testcontainers website.
• Get started with Testcontainers Cloud by creating a free account.
• Have questions? The Docker community is here to help.
• New to Docker? Get started.

Computer vision is a field that enables machines to interpret and understand the visual world. Its applications are rapidly expanding, from healthcare and autonomous vehicles to security systems and retail.

In this article, we’ll go over ten essential tools that every computer vision developer, whether a beginner or an advanced user, should master. These tools range from libraries for image processing to platforms that help with machine learning workflows.

1. OpenCV

• Beginners:
  OpenCV is a popular open-source library designed for computer vision tasks. It’s a great starting point for beginners because it lets you easily perform tasks like image filtering, manipulation, and basic feature detection. With OpenCV, you can start by learning fundamental image processing techniques such as resizing, cropping, and edge detection, which form the foundation for more complex tasks.

Free OpenCV Bootcamp

• Advanced:
  Expert Users: As you progress, OpenCV offers various functionalities for real-time video processing, object detection, and camera calibration. Advanced users can leverage OpenCV for high-performance applications, including integrating it with machine learning models or using it in real-time systems for tasks like facial recognition or augmented reality.

2. TensorFlow

• Beginners:
  TensorFlow is a powerful framework developed by Google for building and training machine learning models, especially in deep learning. It’s beginner-friendly due to its extensive documentation and tutorials. As a new developer, you can start with pre-built models for tasks like image classification and object detection, allowing you to understand the basics of how models learn from data.

Free Tensorflow Bootcamp

• Advanced:
  For advanced users, TensorFlow’s flexibility allows you to build complex neural networks, including Convolutional Neural Networks (CNNs), Transformers  for advanced image recognition tasks. Its ability to scale from small models to large production-level applications makes it an essential tool for any computer vision expert. Additionally, TensorFlow supports distributed training, making it ideal for large-scale datasets and high-performance applications.

3. PyTorch

• Beginners:
  PyTorch, developed by Facebook, is another deep learning framework widely used for building neural networks. Its straightforward, Pythonic nature makes it easy for beginners to grasp the basics of model creation and training. Beginners will appreciate PyTorch’s flexibility in creating simple models for image classification without having to worry about too much technical overhead.
• Advanced:
  Advanced users can use PyTorch’s dynamic computation graph, allowing greater flexibility when building complex architectures, custom loss functions, and optimizers. It’s a great choice for researchers, as PyTorch offers seamless experimentation with cutting-edge models like Vision Language Models, Generative Adversarial Networks (GANs) and deep reinforcement learning. Thanks to its efficient memory management and GPU support, it also excels in handling large datasets.

4. Keras

• Beginners:
  Keras is a high-level neural network API that runs on top of TensorFlow. It’s perfect for beginners as it abstracts much of the complexity involved in building deep learning models. With Keras, you can quickly prototype models for tasks like image classification, object detection, or even more complex tasks like segmentation without needing extensive knowledge of deep learning algorithms.

Free Keras Bootcamp

• Advanced:
  Professional Users: For more experienced developers, Keras remains a useful tool for rapid prototyping of models before diving into deeper customization. While it simplifies the process, Keras also allows users to scale their projects by integrating directly with TensorFlow, giving advanced users the control to fine-tune models and manage performance optimization on large datasets.

5. PaddlePaddle (PaddleOCR for Optical Character Recognition)

• Beginners:
  PaddlePaddle, developed by Baidu, offers an easy way to work with Optical Character Recognition (OCR) tasks through its PaddleOCR module. Beginners can quickly set up OCR models to extract text from images with minimal code. The simplicity of the API makes it easy to apply pre-trained models to your own projects, such as scanning documents or reading text in real-time from images.
• Advanced:
  Professional Users can benefit from PaddleOCR’s flexibility by customizing architectures and training models on their own datasets. The tool allows fine-tuning for specific OCR tasks, such as multilingual text recognition or handwritten text extraction. 

PaddlePaddle also integrates well with other deep learning frameworks, providing room for advanced experimentation and development in complex pipelines.

6. Labeling Tools (e.g., Labelbox, Supervisely)

• Beginners:
  Labeling tools are essential for creating annotated datasets, especially for supervised learning tasks in computer vision. Tools like Labelbox and Supervisely simplify the process of annotating images by offering intuitive user interfaces, making it easier for beginners to create training datasets. Whether you’re working on simple object detection or more advanced segmentation tasks, these tools help you get started with proper data labeling.
• Advanced:
  Experienced professionals  working with large-scale datasets, labeling tools like Supervisely offer automation features, such as pre-annotation or AI-assisted labeling, which significantly speed up the process. These tools also support integration with your machine learning pipelines, enabling seamless collaboration across teams and managing annotations at scale. Professionals can also take advantage of cloud-based tools for distributed labeling, version control, and dataset management.

7. NVIDIA CUDA and cuDNN

• Beginners:
  CUDA is a parallel computing platform and programming model developed by NVIDIA, while cuDNN is a GPU-accelerated library for deep neural networks. For beginners, these tools may seem technical, but their primary purpose is to accelerate the training of deep learning models by utilizing GPU power. By setting up CUDA and cuDNN properly within the training environment, a significant boost in speed and optimization of model training can be achieved, especially when working with frameworks like TensorFlow and PyTorch.
• Advanced:
  Experts can harness the full power of CUDA and cuDNN to optimize performance in high-demand applications. This includes writing custom CUDA kernels for specific operations, managing GPU memory efficiently, and fine-tuning neural network training for maximum speed and scalability. These tools are essential for developers working with large datasets and needing top-tier performance from their models.

8. YOLO (You Only Look Once)

• Beginners:
  YOLO is a fast object detection algorithm that is especially popular for real-time applications. Beginners can use pre-trained YOLO models to quickly detect objects in images or videos with relatively simple code. The ease of use makes YOLO a great entry point for those looking to explore object detection without needing to build complex models from scratch.
• Advanced:
  YOLO provides opportunities for fine-tuning models on custom datasets to detect specific objects, improving detection speed and accuracy. YOLO’s lightweight nature allows it to be deployed in resource-constrained environments, like mobile devices, making it a go-to solution for real-time applications.  Professionals can also experiment with newer versions of YOLO, adjusting parameters to fit specific project needs.

9. DVC (Data Version Control)

• Beginners:
  DVC is a version control system for machine learning projects. For beginners, it helps manage and track datasets, model files, and experiments, making it easy to keep everything organized. Instead of versioning only code (as Git does), DVC ensures that the data and models you are working on are consistently tracked, reducing the hassle of manually managing data for machine learning projects.
• Advanced:
  Expert users can leverage DVC for large-scale projects, enabling reproducibility and collaboration across teams. DVC integrates well with existing workflows, making it easier to manage multiple experiments, track changes in large datasets, and optimize models based on previous runs. For complex machine learning pipelines, DVC helps streamline the workflow by keeping everything under version control, ensuring consistency from data collection to model deployment.

10. Git and GitHub

• Beginners:
  Git and GitHub are essential tools for version control and collaboration. Beginners will find Git useful for managing project history and tracking changes, while GitHub allows easy sharing of code with others. If you’re just starting out in computer vision, learning Git can help you maintain organized project workflows, collaborate on open-source projects, and get familiar with basic version control techniques.
• Advanced:
  Experienced professionals can utilize Git and GitHub to manage complex research projects, handle contributions from multiple developers, and ensure version consistency in large repositories. GitHub Actions allow automation of workflows, such as testing and deploying models, which is especially useful for continuous integration and deployment (CI/CD) in machine learning pipelines. Advanced users can also benefit from using Git LFS (Large File Storage) to manage large datasets within their Git projects.

Roundup

Tools like OpenCV and Keras provide easy entry points for beginners, while advanced options like PyTorch, TensorFlow, and DVC help experienced developers tackle more complex challenges. 

GPU acceleration with CUDA, advanced object detection with YOLO, and efficient data management with labeling tools ensure you can build, train, and deploy powerful models effectively.

The post 10 Essential Tools Every Beginner & Advanced Computer Vision Developer Should Master appeared first on OpenCV.

Welcome to the third annual Docker State of Application Development survey!

Please help us better understand and serve the application development  community with just 20-30 minutes of your time. We want to know where you’re focused, what you’re working on, and what is most important to you. Your thoughts and feedback will help us build the best products and experiences for you.

And, we don’t just keep this information for ourselves — we share with you¹! We hope you saw our recent report on the 2023 State of Application Development Survey. The engagement of our community allowed us to better understand where developers are facing challenges, what tools they like, and what they’re excited about. We’ve been using this information to give our community the tools and features they need.

Take the Docker State of Application Development survey now!

By participating in the survey, you can be entered into a raffle for a chance to win² one of the following prizes:

• 1 laptop computer (an Apple M3 Macbook Pro 16″)
• 2 game consoles with VR headsets: Playstation Virtual Reality Headset and a Playstation 5
• 5 $300 Amazon.com gift cards 
• 50 exclusive Docker swag sets 

Additionally, the first 200 respondents to complete the survey will receive an exclusive pair of Docker socks!

The survey is open from September 23rd, 2024 (7AM PST) to November 20, 2024 (11:59PM PST). 

We’ll choose the winners randomly in accordance with the promotion official rules.* Winners will be notified via email by January 10, 2025.

The Docker State of Application Development Survey only takes about 20-30 minutes to complete. We appreciate every contribution and opinion. Your voice counts!

1. Data will be reported publicly only in aggregate and without personally identifying information. ↩︎
2. Docker State of Application Development Promotion Official Rules. ↩︎

As artificial intelligence (AI) has erupted, Secure by Design for AI has emerged as a critical paradigm. AI is integrating into every aspect of our lives — from healthcare and finance to developers to autonomous vehicles and smart cities — and its integration into critical infrastructure has necessitated that we move quickly to understand and combat threats. 

Necessity of Secure by Design for AI

AI’s rapid integration into critical infrastructure has accelerated the need to understand and combat potential threats. Security measures must be embedded into AI products from the beginning and evolve as the model evolves. This proactive approach ensures that AI systems are resilient against emerging threats and can adapt to new challenges as they arise. In this article, we will explore two polarizing examples — the developer industry and the healthcare industry.

Complexities of threat modeling in AI

AI brings forth new challenges and conundrums when working on an accurate threat model. Before reaching a state in which the data has simple edit and validation checks that can be programmed systematically, AI validation checks need to learn with the system and focus on data manipulation, corruption, and extraction. 

• Data poisoning: Data poisoning is a significant risk in AI, where the integrity of the data used by the system can be compromised. This can happen intentionally or unintentionally and can lead to severe consequences. For example, bias and discrimination in AI systems have already led to issues, such as the wrongful arrest of a man in Detroit due to a false facial recognition match. Such incidents highlight the importance of unbiased models and diverse data sets. Testing for bias and involving a diverse workforce in the development process are critical steps in mitigating these risks.

In healthcare, for example, bias may be simpler to detect. You can examine data fields based on areas such as gender, race, etc. 

In development tools, bias is less clear-cut. Bias could result from the underrepresentation of certain development languages, such as Clojure. Bias may even result from code samples based on regional differences in coding preferences and teachings. In developer tools, you likely won’t have the information available to detect this bias. IP addresses may give you information about where a person is living currently, but not about where they grew up or learned to code. Therefore, detecting bias will be more difficult. 

• Data manipulation: Attackers can manipulate data sets with malicious intent, altering how AI systems behave. 
• Privacy violations: Without proper data controls, personal or sensitive information could unintentionally be introduced into the system, potentially leading to privacy violations. Establishing strong data management practices to prevent such scenarios is crucial.
• Evasion and abuse: Malicious actors may attempt to alter inputs to manipulate how an AI system responds, thereby compromising its integrity. There’s also the potential for AI systems to be abused in ways developers did not anticipate. For example, AI-driven impersonation scams have led to significant financial losses, such as the case where an employee transferred $26 million to scammers impersonating the company’s CFO.

These examples underscore the need for controls at various points in the AI data lifecycle to identify and mitigate “bad data” and ensure the security and reliability of AI systems.

Key areas for implementing Secure by Design in AI

To effectively secure AI systems, implementing controls in three major areas is essential (Figure 1):

1. Data management

The key to data management is to understand what data needs to be collected to train the model, to identify the sensitive data fields, and to prevent the collection of unnecessary data. Data management also involves ensuring you have the correct checks and balances to prevent the collection of unneeded data or bad data.

In healthcare, sensitive data fields are easy to identify. Doctors offices often collect national identifiers, such as drivers licenses, passports, and social security numbers. They also collect date of birth, race, and many other sensitive data fields. If the tool is aimed at helping doctors identify potential conditions faster based on symptoms, you would need anonymized data but would still need to collect certain factors such as age and race. You would not need to collect national identifiers.

In developer tools, sensitive data may not be as clearly defined. For example, an environment variable may be used to pass secrets or pass confidential information, such as the image name from the developer to the AI tool. There may be secrets in fields you would not suspect. Data management in this scenario involves blocking the collection of fields where sensitive data could exist and/or ensuring there are mechanisms to scrub sensitive data built into the tool so that data does not make it to the model. 

Data management should include the following:

• Implementing checks for unexpected data: In healthcare, this process may involve “allow” lists for certain data fields to prevent collecting irrelevant or harmful information. In developer tools, it’s about ensuring the model isn’t trained on malicious code, such as unsanitized inputs that could introduce vulnerabilities.
• Evaluating the legitimacy of users and their activities: In healthcare tools, this step could mean verifying that users are licensed professionals, while in developer tools, it might involve detecting and mitigating the impact of bot accounts or spam users.
• Continuous data auditing: This process ensures that unexpected data is not collected and that the data checks are updated as needed. 

2. Alerting and monitoring 

With AI, alerting and monitoring is imperative to ensuring the health of the data model. Controls must be both adaptive and configurable to detect anomalous and malicious activities. As AI systems grow and adapt, so too must the controls. Establish thresholds for data, automate adjustments where possible, and conduct manual reviews where necessary.

In a healthcare AI tool, you might set a threshold before new data is surfaced to ensure its accuracy. For example, if patients begin reporting a new symptom that is believed to be associated with diabetes, you may not report this to doctors until it is reported by a certain percentage (15%) of total patients. 

In a developer tool, this might involve determining when new code should be incorporated into the model as a prompt for other users. The model would need to be able to log and analyze user queries and feedback, track unhandled or poorly handled requests, and detect new patterns in usage. Data should be analyzed for high frequencies of unhandled prompts, and alerts should be generated to ensure that additional data sets are reviewed and added to the model.

3. Model tuning and maintenance

Producers of AI tools should regularly review and adjust AI models to ensure they remain secure. This includes monitoring for unexpected data, adjusting algorithms as needed, and ensuring that sensitive data is scrubbed or redacted appropriately.

For healthcare, model tuning may be more intensive. Results may be compared to published medical studies to ensure that patient conditions are in line with other baselines established across the world. Audits should also be conducted to ensure that doctors with reported malpractice claims or doctors whose medical license has been revoked are scrubbed from the system to ensure that potentially compromised data sets are not influencing the model. 

In a developer tool, model tuning will look very different. You may look at hyperparameter optimization using techniques such as grid search, random search, and Bayesian search. You may study subsets of data; for example, you may perform regular reviews of the most recent data looking for new programming languages, frameworks, or coding practices. 

Model tuning and maintenance should include the following:

• Perform data audits to ensure data integrity and that unnecessary data is not inadvertently being collected. 
• Review whether “allow” lists and “deny” lists need to be updated.
• Regularly audit and monitor alerts for algorithms to determine if adjustments need to be made; consider the population of your user base and how the model is being trained when adjusting these parameters.
• Ensure you have the controls in place to isolate data sets for removal if a source has become compromised; consider unique identifiers that allow you to identify a source without providing unnecessary sensitive data.
• Regularly back up data models so you can return to a previous version without heavy loss of data if the source becomes compromised.

AI security begins with design

Security must be a foundational aspect of AI development, not an afterthought. By identifying data fields upfront, conducting thorough AI threat modeling, implementing robust data management controls, and continuously tuning and maintaining models, organizations can build AI systems that are secure by design. 

This approach protects against potential threats and ensures that AI systems remain reliable, trustworthy, and compliant with regulatory requirements as they evolve alongside their user base.

Learn more

• Read our Security series.
• Subscribe to the Docker Newsletter. 
• Get the latest release of Docker Desktop.
• Vote on what’s next! Check out our public roadmap.
• Have questions? The Docker community is here to help.
• New to Docker? Get started.