Building web apps that respect a user’s privacy and security

What Are Ethics?

If we’re going to discuss the ethics of web development, we first need to establish a common understanding of how we apply the term ethics. The study of ethics falls into four categories:

Meta-ethics

An attempt to understand the underlying questions of ethics and morality

Descriptive ethics

The study and research of people’s beliefs

Normative ethics

The study of ethical action and creation of standards of right and wrong

Applied ethics

The analysis of ethical issues, such as business ethics, environmental ethics, and social morality

For our purposes, we will do our best to determine a normative set of ethical standards as applied to web development, and then take an applied ethics approach.

Within normative ethical theory, there is the idea of consequentialism, which argues that the ethical value of an action is based on its result. In short, the consequences of doing something become the standard of right or wrong. One form of consequentialism, utilitarianism, states that an action is right if it leads to the most happiness, or well-being, for the greatest number of people. This utilitarian approach is the framework I’ve chosen to use as we explore the ethics of web development.

Whew! We fell down a deep, dark hole of philosophical terminology, but I think it all boils down to this:

Make choices that have the most positive effect for the largest number of people.

Professional Ethics

Many professions have a standard expectation of behavior. These may be legally mandated or a social norm, but often take the form of a code of ethics that details conventions, standards, and expectations of those who practice the profession. The idea of a professional code of ethics can be traced back to the Hippocratic oath, which was written for medical professionals during the fifth century BC (see Figure 1-1). Today, medical schools continue to administer the Hippocratic or a similar professional oath.

In the book Thinking Like an Engineer (Princeton University Press), Michael Davis says a code of conduct for professionals:

[P]rescribes how professionals are to pursue their common ideal so that each may do the best she can at a minimal cost to herself and those she cares about…The code is to protect each professional from certain pressures (for example, the pressure to cut corners to save money) by making it reasonably likely (and more likely then otherwise) that most other members of the profession will not take advantage of her good conduct. A code is a solution to a coordination problem.

My hope is that this report will help inspire a code of ethics for web developers, guiding our work in a way that is professional and inclusive.

The approaches I’ve laid out are merely my take on how web development can provide the greatest happiness for the greatest number of people. These approaches are likely to evolve as technology changes and may be unique for many development situations. I invite you to read my practical application of these ideas and hope that you apply them in some fashion to your own work.

This series is a work in progress, and I invite you to contribute. To learn more, visit the Ethical Web Development website.

Intended Audience

This title, like others in the Ethical Web Development series, is intended for web developers and web development team decision makers who are interested in exploring the ethical boundaries of web development. I assume a basic understanding of fundamental web development topics such as HTML, CSS, JavaScript, and HTTP. Despite this assumption, I’ve done my best to describe these topics in a way that is approachable and understandable.

How Users Are Tracked

As users browse the web, they are being tracked; and as web developers, we are often enabling and supporting that surveillance. This isn’t a case of tinfoil hat paranoia: we’re introducing the code of ad networks to support our work, adding social media share buttons that allow users to easily share our sites’ content, or using analytics software to help us better understand the user experience. Websites track users’ behavior with the intention of providing them with a more unique experience. While this may seem harmless or well intentioned, it is typically done without the knowledge or permission of the end user.

The simplest way that web tracking works is that a user visits a site that installs a cookie from a third party. When the user then visits another site with the same third-party tracker, the tracker is notified (see Figure 2-1). This allows the third party to build a unique user profile.

The intention of this tracking is typically to provide more targeted services, advertising, or products. However, the things we buy, the news we read, the politics we support, and our religious beliefs are often embedded into our browsing history. To many, gathering this knowledge without explicit permission feels intrusive.

What Does Your Browser Know About You?

Those aware of user tracking may take a few steps to beat trackers at their own game. Ad blockers such as uBlock Origin block advertisements and third-party advertising trackers. Other browser extensions such as Privacy Badger and Ghostery attempt to block all third-party beacons from any source. However, even with tools like these, sites may be able to track users based on the unique footprint their browser leaves behind. In fact, according to the W3C slide deck “Is Preventing Browser Fingerprinting a Lost Cause?” the irony of using these tools is that “fine-grained settings or incomplete tools used by a limited population can make users of these settings and tools easier to track.”

Browsers can easily detect the user’s IP address, user agent, location, browser plug-ins, hardware, and even battery level. Web developer Robin Linus developed the site What Every Browser Knows About You to show off the level of detail available to developers and site owners. Additionally, the tools Am I Unique? and Panopticlick offer quick overviews of how unique your browser fingerprint is.

Do Not Track

With this information about the ways in which users can be tracked in mind, how can we, as web developers, advocate for our users’ privacy? My belief is that the first step is to respect the Do Not Track (DNT) browser setting, which allows users to specify a preference to not be tracked by the sites they visit. When a user has enabled the Do Not Track setting in her browser, the browser responds with the HTTP header field DNT.

According to the Electronic Frontier Foundation, Do Not Track boils down to sites agreeing not to collect personally identifiable information through methods such as cookies and fingerprinting, as well as agreeing not to retain individual user browser data beyond 10 days. The noted exceptions to this policy are when a site is legally responsible for maintaining this information, when the information is needed to complete a transaction, or if a user has given explicit consent.

With Do Not Track enabled, browsers send an HTTP header response with a DNT value of 1. The following is a sample header response, which includes a DNT value:

Host: "www.example.com"
Accept: "text/html,application/xhtml+xml,
         application/xml;q=0.9,*/*;q=0.8"
Accept-Language: "en-US,en;q=0.5"
Accept-Encoding: "gzip, deflate, br"
DNT: "1"

Do Not Track does not automatically disable tracking in a user’s browser. Instead, as developers, we are responsible for appropriately handling this user request in our applications.

// store user Do Not Track setting as a variable
var dnt = navigator.doNotTrack;

if (dnt !== 1) {
  // set cookie only if DNT not enabled
  document.cookie = 'example';
}

DoNotTrackHeader = "DNT"
DoNotTrackValue = "1"

pyHeader = "HTTP_" + DoNotTrackHeader.replace("-", "_").upper()

# request is an HttpRequest
if (pyHeader in request.META) and 
   (request.META[pyHeader] == DoNotTrackValue):
    # Do Not Track is enabled
else:
    # Do Not Track is not enabled

var http = require('http');

http.createServer(function (req, res) {

  var dnt = req.headers.dnt === '1' || false;

  if (dnt) {
    // Do Not Track is enabled
  } else {;
    // Do Not Track is not enabled
  }

    res.end();
}).listen(3000);

Web Analytics

One of the biggest challenges of handling user privacy is determining best practices for web analytics. By definition, the goal of web analytics is to track users, though the aim is typically to better understand how our sites are used so that we can continually improve them and adapt them to user needs.

To protect user privacy, when using analytics we should ensure that our analytics provider anonymizes our users, limits tracking cookies to our domain, and does not share user information with third parties. The US Government’s digital analytics program has taken this approach, ensuring that Google Analytics does not track individuals or share information with third parties and that it anonymizes all user IP addresses.

As an additional example, the analytics provider Piwik actively seeks to maintain user privacy while working with user analytics through:

• Providing an analytics opt-out mechanism
• Deleting logs older than a few months
• Anonymizing IP addresses
• Respecting Do Not Track
• Setting a short expiration date for cookies

These examples provide a good baseline for how we should aim to handle analytics on our sites with any provider. By taking this extra care with user information, we may continue to use analytics to provide greater insights into the use of our sites while maintaining user privacy.

De-identification

Though it is preferable to avoid the tracking of users completely, there may be instances where this choice is outside of the control of web developers. In these cases, we may be able to guide the decision to de-identify collected user data, ensuring that user privacy remains intact. The goal of de-identification is to ensure that any collected data cannot be used to identify the person who created the data in any way.

However, de-identification is not without its limitations, as de-identified data sets can be paired with other data sets to identify an individual. In the paper “No Silver Bullet: De-Identification Still Doesn’t Work,” Arvind Narayanan and Edward W. Felten explore the limits of de-identification. Cryptographic techniques such as differential privacy can be used as another layer to help limit the identification of individual users within collected data sets.

User Consent and Awareness

In 2011 the European Union passed legislation requiring user consent before using tracking technology. Specifically, the privacy directive specifies:

Member States shall ensure that the use of electronic communications networks to store information or to gain access to information stored in the terminal equipment of a subscriber or user is only allowed on condition that the subscriber or user concerned is provided with clear and comprehensive information in accordance with Directive 95/46/EC, inter alia about the purposes of the processing, and is offered the right to refuse such processing by the data controller.

This means that any site using cookies, web beacons, or similar technology must inform the user and receive explicit permission from her before tracking. If you live in Europe or have visited a European website, you are likely familiar with the common “request to track” banner. This law is not without controversy, as many feel that these banners are ignored, viewed as a nuisance, or otherwise not taken seriously.

In the UK, the guidance has been to simply inform users that they are being tracked, providing no option to opt out. For example, the website of the Information Commissioner’s Office, the “UK’s independent authority set up to uphold information rights in the public interest, promoting openness by public bodies and data privacy for individuals,” opts users in, but clicking the “Information and Settings” link provides information about browser settings and disabling cookies on the site (see Figure 2-2).

Though based in the United States, the site Medium.com alerts users with DNT enabled how their information will be used and assumes tracking consent only when users log in to their accounts (see Figure 2-3).

Further Reading

• “The Emerging Ethical Standards for Studying Corporate Data” by Jules Polonetsky and Dennis Hirsch
• “Do Not Track Is No Threat to Ad-Supported Businesses” by Jonathan Mayer
• The Electronic Frontier Foundation’s guide to Do Not Track
• Mozilla: Developer Network’s DNT header reference
• W3C: Working Draft “Tracking Compliance and Scope”

How HTTPS Works

At the most basic level, the HTTP request and response cycle is when a web-connected computer requests a specific resource through a URL and a server responds with that resource, such as an HTML page (see Figure 3-1).

When this information is requested, not only are the files sent over the wire, but so is user information, such as the user’s IP address, location, browser information, system information, and so on. More importantly, all of this information is sent as unencrypted plain text over the public internet, meaning that any network sitting between the user’s browser and the server has access to that information. This means that when I request a website like in Figure 3-1, what I’m really saying is, “Hello, I’m user 192.00.000.001 in the United States using Mozilla Firefox 48.0.1 on an Intel Macintosh 10.11.6 and would like the /page.html resource from http://ethicalweb.org.” The server in turn responds by returning the unencrypted resource to my browser.

HTTPS works similarly to HTTP, but adds a layer of Secure Sockets Layer/Transport Layer Security (SSL/TLS) encryption. This means that requests and responses are made over a secure encrypted connection. These requests include only the user’s IP address and the domain of the requested resource, so in this case my request would appear as “Hello, I’m user 192.00.000.001 and would like a resource from https://ethicalweb.org.” The server would then respond with an encrypted version of the resource.

The United States government’s HTTPS-Only Standard helpfully demonstrates the difference between these two requests. The standard unencrypted HTTP request includes a number of headers about the client and the request, as seen in Figure 3-2.

By contrast, the encrypted HTTPS request limits this information (Figure 3-3).

Why Use HTTPS

Now that we’ve looked at what HTTPS is and how it works, we can begin to see some of the value it provides both to our users and to ourselves as site owners and maintainers. Specifically, reasons to use HTTPS include the following:

• Protecting users’ privacy and security
• Proving a site’s authenticity and integrity
• Browser deprecated HTTP
• Potential search ranking improvements

Let’s take a closer look at each of these.

Browsers Deprecating HTTP

Currently, browsers display an indication whenever a site is being served securely using HTTPS. This appears as a green padlock next to the site’s URL (Figure 3-4).

However, there is no indicator for sites that are not using HTTPS (Figure 3-5).

Recently the Chromium team pointed out that “people do not generally perceive the absence of a warning sign” and consequently suggested that browsers instead mark HTTP as insecure, alerting users when sites are being served over HTTP. Calling attention to sites served over plain HTTP would send a clear signal that HTTPS is preferred, promoting the deprecation of HTTP.

The second way that browsers are beginning to deprecate HTTP is by making new browser APIs available only to sites served over HTTPS. These include offline capabilities with service workers (covered in Building Web Apps that Work Everywhere), the ability to access users’ camera and audio inputs with getUserMedia, and the ability to access user location information with the geolocation API. Looking at the types of information these APIs have access to, I’m thankful that browser vendors have decided that they should only be accessed over a secure connection. An added benefit of developing forward-thinking applications is that HTTPS will quickly become a requirement.

Implementing HTTPS

Now that we have examined how HTTPS works and explored why we should use it, let’s take a look at implementing HTTPS for our own sites.

$ sudo apt-get install python-letsencrypt-apache

$ letsencrypt --apache

$ letsencrypt renew

$ letsencrypt renew --dry-run

$ crontab -e

17 05, 17 17 * * * letsencrypt renew

Other Considerations

Once you have implemented HTTPS, there are a few site-wide changes to take into consideration:

• Redirecting HTTP to HTTPS
• Enabling HTTP Strict Transport Security (HSTS)
• Preventing mixed content and using relative URLs
• Using secure cookies

ServerName www.example.com
Redirect "/" "https://www.example.com/"

Strict-Transport-Security: max-age=31536000; includeSubDomains

Header always set Strict-Transport-Security "max-age=63072000; 
       includeSubdomains; preload"

res.cookie('user', 'adam', { secure: true });

Conclusion

HTTPS provides numerous benefits to both site owners and users, helping make the web more secure. Whatever method you choose to implement HTTPS for your sites, you are taking important steps to improve the security and privacy of your users.

Further Reading

• The United States Government’s HTTPS-Only Standard
• GOV.UK’s “Using HTTPS”
• “How Does HTTPS Actually Work?” by Rob Heaton
• Is TLS Fast Yet?
• W3C report “Securing the Web”
• Google Developers resource “Enabling HTTPS on Your Servers”
• “We’re Deprecating HTTP and It’s Going to Be Okay” by Eric Mill

Building on a Strong Foundation

Being web developers means that we are constantly learning about and using new tools. It’s an exciting perk of the job. That said, when building secure applications, we are often best served to use established frameworks that have been thoroughly vetted and that provide baked-in security support. As an example, let’s look at the security options when building a web application with Python or Node.js.

The Python environment is relatively stable, and most web applications are built using either the Django or Flask web frameworks. Django provides many security features out of the box, such as cross-site scripting (XSS), SQL injection, and clickjacking protection. As Flask is an intentionally more lightweight framework, it comes with relatively few built-in security features, such as manageable XSS protection. Additional security features can be added with the Flask-Security extension.

Node.js is notorious for its rate of change and the number of available frameworks and libraries. This can be both something to love about the platform and a frustration point for many developers. The site Node Frameworks attempts to catalog them all. Despite there being dozens of Node.js web framework options, when considering security we are likely to be best served by choosing an established framework that is used in production by other web applications, such as Express.

Similar to Flask, Express is a lightweight application framework, but there are several plug-ins that enhance its security features. The two most common are Lusca, which was developed by PayPal, and Helmet. These both add sensible defaults for features such as XSS protection, cross-site request forgery (CSRF) protection, and content security policy settings, among others.

In addition to using security-focused libraries, we should also work to ensure that our dependencies are up to date and free of known vulnerabilities.

In Python we can check for outdated pip-installed packages with:

$ pip list --outdated

Similarly, in Node.js we can list outdated packages with npm:

$ npm outdated

In addition to these manual methods, there are tools that will help maintain dependency updates and scan for security vulnerabilities. Greenkeeper will scan your project for outdated Node.js dependencies and create a pull request with the updates. Greenkeeper will also run your application’s test suite, ensuring that the updated dependencies do not break the build. Snyk is a tool that will scan Node.js packages for known vulnerabilities and alert you to insecure dependencies. The site also provides a command-line means to fixing these vulnerabilities.

Though these examples are limited to Python and Node.js, I hope that you can see how the concepts map to your web stack of choice. When we use established technologies and take advantage of built-in or plug-in-based security features, we are creating solid security foundations for our sites.

OWASP Top 10

Now that we’ve started our application off on a strong foundation, it’s worth exploring the common security vulnerabilities that we should be aware of. Every few years the Open Web Application Security Project (OWASP) publishes a list of the most critical web application security flaws. As of the most recent publication, the OWASP Top 10 is comprised of:

1. Injection
2. Broken authentication and session management
3. Cross-site scripting
4. Insecure direct object references
5. Security misconfiguration
6. Sensitive data exposure
7. Missing function-level access control
8. Cross-site request forgery
9. Using components with known vulnerabilities
10. Unvalidated redirects and forwards

This list of common vulnerabilities can provide us with an awareness of potential weaknesses in our own applications.

Secure User Authentication

When a user creates an account with our site, she is placing her trust in us. Often in this process the user may agree to terms of service about how she may interact with our site and services and how the site owners will use the data and information users provide within the application. One crucial step in upholding our end of this agreement is to ensure that user login information is kept secure and private. Let’s explore how we can do so.

$ npm install bcrypt --save

// require the module
var bcrypt = require('bcrypt');

// the cost of processing the salting data - 10 is the default
var saltRounds = 10;

// function for hashing and salting
function passwordEncrypt(username, password) {
  // generate the salt
  bcrypt.genSalt(saltRounds, function(err, salt) {
    // generate the hash
    bcrypt.hash(password, salt, function(err, hash) {
      // store username, hash, and salt in your password DB
    });
  });
}

// password is a value provided by the user
// hash is retrieved from our DB
bcrypt.compare(password, hash, function(err, res) {
  // res is either true or false
});

OAuth 2.0

An alternative option to providing our own login system is to use OAuth 2.0. OAuth 2.0 is a user authorization system that lets us provide a user login option through popular third-party sites such as Google, Facebook, Twitter, LinkedIn, and more. This allows us to both rely on large and trusted third parties for providing authentication and pull in useful user information, as authorized by the user, from the chosen service.

Even if you have never worked with OAuth as a developer, you are likely familiar with the flow from the perspective of a user.

First, a user clicks a login link from our application (Figure 4-1).

The user is then directed to an access request for the specified service provider, like the one in Figure 4-2. This request details the level of access to user information that the accessing application will have.

If the user grants the authorization, the service redirects the user back to our site with an authorization code. Our server will then exchange the authorization code for an access token. Once the access token has been granted, that will be used to access the user information from the third-party service.

Links to libraries for popular programming languages and web frameworks, tutorials, and documentation can be found on the OAuth website. Additionally, Aaron Parecki, the maintainer of OAuth, has written a fantastic guide called “OAuth 2 Simplified.”

Multifactor Authentication

One way we can provide a more secure authentication system to our users is by making multifactor authentication available. Multifactor authentication is done by combining two more or more of the following:

1. A secret known to the user, such as a password or PIN
2. A physical object in the user’s possession, such as a mobile phone or a USB FIDO U2F Security Key (Figure 4-3)
3. A physical characteristic of the user, such as a fingerprint or typing speed

In web applications, the most common pattern is to make two-factor authentication available by providing a physical authentication in addition to the standard username/password flow. Often, users will receive a text message on their mobile phone or install a multifactor authentication application that will provide the appropriate code for this additional verification step. Adding a physical dimension reduces the possibility of password theft providing access to a user’s account. Though many users may opt not to enable two-factor authentication, providing this option is a good step toward better security than standard username and password authentication.

Encrypting User Data

Depending on the types of applications we work on, they may contain sensitive user information beyond user credentials. Our applications may store user locations, journal entries, Social Security numbers, health records, or any number of private bits of information that users have entrusted us with. When this is the case, it becomes important to encrypt sensitive user information, in addition to passwords. Doing this acknowledges that we are willing to take extra steps and security precautions with our users’ information.

We’ve recently seen a rise in the popularity of services that encrypt and secure user data. The email provider ProtonMail offers secure and encrypted email accounts, and the mobile messaging application Signal is fully encrypted. We’ve also seen encryption become a selling point for mainstream applications. For instance, the popular messaging app WhatsApp now provides end-to-end encryption for user conversations. Even if a user is unaware of this feature, it is provided as an additional layer of security and privacy.

Encrypting user data can be useful for much more than messaging and email applications, however. For example, in the case of the OPM hack mentioned at the beginning of this chapter, government employees would have greatly benefited from having their records encrypted rather than stored as plain text in the database.

In Node.js we can use the built-in crypto library to encrypt and decrypt user data. Here’s a very basic example of what that might look like with a pair of functions that encrypt and decrypt some plain text using a provided password:

var crypto = require('crypto');

function dataEncrypt(password, text) {
  var cipher = crypto.createCipher('aes192', password);
  var encrypted = cipher.update(text, 'utf8', 'hex');
  encrypted += cipher.final('hex');
  return encrypted;
}

function dataDecrypt(password, encrypted) {
  var decipher = crypto.createDecipher('aes192', password);
  var decrypted = decipher.update(encrypted, 'hex', 'utf8');
  decrypted += decipher.final('utf8');
  return decrypted;
}

// encrypt some data
var encrypt = dataEncrypt('Password', 'This is encrypted!');
// returns f53a6a423a11be8f27ff86effa5ace548995866009190a90...
var decrypt = dataDecrypt('Password', encrypt);
// returns This is encrypted!

By storing user data in an encrypted format we are taking an extra step toward securing that data for our users.

Sanitizing and Validating User Input

Interactive form fields and text input are often the differentiator between a website and a web application. Introducing this type of interactivity opens a site up to both database injections and cross-site scripting attacks, two of the top three security vulnerabilities on the OWASP Top 10 list. Database injections occur when an attacker injects code or database commands (such as SQL statements) into the database. Cross-site scripting can occur when an attacker is able to inject malicious scripts into a site. Steps can be taken to prevent both of these potential attacks by sanitizing and validating user input.

To sanitize user-submitted content, we should whitelist the HTML input that our application will accept. Whitelisting is preferred to blacklisting user input as it gives us fine-grained control over the type of content being entered and stored. If users are able to add HTML to a field, we can choose the tags that should be available to the user and whitelist those. We should be sure to avoid giving users the ability to execute JavaScript or <script> tags within our applications.

In Node.js we can use the sanitize-html module to do this. First, we install the module as a project dependency:

$ npm install sanitize-html --save

Now in our project code we can include the module and sanitize using a whitelist of accepted tags:

var sanitizeHtml = require('sanitize-html');

var dirty = 'HTML entered from the client';
var clean = sanitizeHtml(dirty, {
  allowedTags: [ 'b', 'i', 'em', 'strong', 'a' ],
  allowedAttributes: {
    'a': [ 'href' ]
  }
});

To avoid database injection, we should further sanitize our user input. When using an SQL database it is important to prevent characters being entered into the database so that SQL statements cannot be injected. By contrast, NoSQL injections may be executed at either the database or application layer. To prevent attacks when using a NoSQL database, we should again ensure that executable code or special characters used by the database are not entered into it.

Cross-Site Request Forgery Attacks

Cross-site request forgery (CSRF) is a type of attack where a site utilizes a user’s browser to manipulate a web application. Through CSRF, an attacker can forge login requests or complete actions that are typically done by a logged-in user, such as posting comments, transferring money, or changing user account details. These attacks can be carried out by utilizing browser cookies or user IP address information. Whereas cross-site scripting attacks exploit a user’s trust in a site, CSRF attacks exploit the trust a site places in the user’s browser.

Wikipedia defines the following common CSRF characteristics:

• They involve sites that rely on a user’s identity.
• They exploit the site’s trust in that identity.
• They trick the user’s browser into sending HTTP requests to a target site.
• They involve HTTP requests that have side effects.

Two possible steps we can take to prevent CSRF attacks are to include a secret token in our forms and to validate the Referer header in requests.

When dealing with form submission, most web frameworks provide CSRF protection or have available plug-ins for generating and validating the tokens. The Django web framework includes default middleware for creating posts with CSRF tokens. The Node module csurf provides the same functionality for applications built using the Express framework.

The second protective measure we can take is to verify the Referer header and, if it is not present or comes from an incorrect URL, deny the request. It should be noted that this header can be spoofed, so this is not a failsafe measure, but it can add a layer of protection for users. Additionally, be aware that some users may disable this header in their browsers due to privacy concerns and thus will not benefit from Referer header validation.

Security Headers

To further harden our application’s security, we can set a number of HTTP headers that give our users’ browsers information about the types of requests possible on our site. Enabling each of these headers will provide further protection for our users against potential threats such as cross-site scripting and clickjacking.

header always set Content-Security-Policy "default-src 'self';"

default-src 'none'; script-src 'self'; connect-src 'self'; 
            img-src 'self'; style-src 'self';

script-src 'self' cdn.example.com;

X-Frame-Options: DENY
X-Frame-Options: SAMEORIGIN
X-Frame-Options: ALLOW-FROM https://example.com/

header always set x-frame-options "SAMEORIGIN"

header always set x-xss-protection "1; mode=block"

header always set X-Content-Type-Options "nosniff"

Checking Security Headers

Once our security headers have been set, we can use securityheaders.io to scan our site. The tool analyzes the site’s response headers and produces a grade indicating the level of protection. Scanning the tool’s own site results in an A+ score (Figure 4-4).

Security Disclosures and Bug Bounty Programs

No matter how diligent we are about security, there may be flaws in our application. To improve security and the user experience, we should acknowledge this potential by having a strong security disclosure plan and consider implementing a bug bounty program.

Developer Jonathan Rudenberg’s post “Security Disclosure Policy Best Practices” provides a succinct strategy for handling security disclosures. In it, he outlines the following key points for having an effective security program:

1. Have a security page with an email address and PGP key for submitting security disclosures.
2. Have a clear, concise, and friendly security policy.
3. Disclose any reported vulnerability.
4. Respond to the vulnerability quickly.
5. Don’t place blame on teammates or employees.
6. Alert customers and inform them of the remediation steps.

As part of this process, you may want to offer a bug bounty for security researchers who discover vulnerabilities. The site Bugcrowd has compiled a list of bug bounty programs that can serve as exemplars. Some well-known sites that offer bug bounties include Facebook, Google, GitHub, and Mozilla. Recently the United States Department of Defense has even gotten in on the action, launching the Hack the Pentagon program.

By providing clear steps for reporting security vulnerabilities and transparent communication about remediation steps, we can work to build additional trust in our users.

Conclusion

There are a dizzying number of possibilities when it comes to web application security, but by building on a solid foundation, following best practices, and providing clear security information to our users, we can work to build a more secure web. I hope that this chapter serves as a strong jumping-off point for your efforts to build and maintain secure web applications.

Further Reading

• Identity and Data Security for Web Development by Jonathan LeBlanc and Tim Messerschmidt (O’Reilly)
• Security for Web Developers by John Paul Mueller (O’Reilly)
• Awesome AppSec
• “A Practical Security Guide for Web Developers” by FallibleInc
• OWASP Testing Guide
• “Python & Django Security on a Shoestring: Resources” by Kelsey Gilmore-Innis
• “Security Tips for Web Developers” by Jesse Ruderman
• “The Password Manifesto” by Andrew A. Gill
• “Mozilla Cybersecurity Delphi 1.0: Towards a User-Centric Policy Framework”
• XATO: Security
• xkcd: Password Strength

Data Ownership

Who owns the data generated within our applications? Though it may be easiest to say “the user,” this can become an increasingly complicated question when we consider things such as collaborative documents, online discussions, and shared calendars, which may have an initial creator but ultimately may also have multiple maintainers. What about the sites themselves? Sometimes the terms of service may insist on ownership or exclusive rights to a user’s created content. As part of Facebook’s terms of service, the company enforces exclusive rights to any content created within or posted to the site:

For content that is covered by intellectual property rights, like photos and videos (IP content), you specifically give us the following permission, subject to your privacy and application settings: you grant us a non-exclusive, transferable, sub-licensable, royalty-free, worldwide license to use any IP content that you post on or in connection with Facebook (IP License).

In doing this, we take the power away from the user and assert ownership over the content they have created. Though there is a business case for this, it comes at a potential cost to our users. The creator of the World Wide Web, Tim Berners-Lee, has spoken out in favor of user-owned data, stating that “the data that [firms] have about you isn’t valuable to them as it is to you.”

If we take this perspective, we should aim to open user data to our users and provide a means of exporting it from our sites in an open format.

In his article “Rights to Your Data and Your Own Uber ‘God’ View,” Miles Grimshaw suggests adapting a Creative Commons-style license for personal data, which would be adopted by services collecting this data:

You are free to:

Download—free access to your raw data in standard file formats

Share—copy and redistribute the data in any medium or format

Adapt —remix, transform, and build upon the data

Under the following terms:

Attribution—You must provide a sign-up link to the application

The (since acquired) start-up Kifi had a forward-thinking approach to user data, stating in a blog post that:

Any service that manages your data has an implicit contract with users: you give us your data and we’ll organize it, but it’s still your data; we are just stewards for it. At Kifi, one way we try to fulfill our end of this contract is by making sure users can export their data for offline use (or so they can import it into another service).

These ideas are not limited to start-ups or small services. In 2012 Twitter introduced the ability to download an archive of your Tweets, giving users permanant access to their Twitter content as well as the potential ability to import it into another service. Google also allows users to download an archive of the data created with any of its services, including the ability to easily store the archive in common file-sharing applications such as Dropbox, Google Drive, and Microsoft OneDrive.

By giving our users access to their data, we can be better stewards of that information. This aids us in creating long-lasting user content and opens up the potential for users to adapt and use their data in novel and interesting ways. Most importantly, by providing access to user data we are able to give ownership of the data our users create directly to the users.

Deleting User Data

An inevitable reality is that some users will want to stop using the services we build. In many cases, these users may simply allow their accounts to decay, but other users will explicitly seek to delete their accounts and associated information. When a user does delete his account, we should also delete it from our databases, rather than simply hiding the user’s content within our site or application. Doing so will be more in line with user expectations and and ensures that in the case of a data breach previously deleted accounts won’t be at risk.

Archiving and Graceful Shutdown

At the beginning of this chapter, we looked at a few web application shutdowns and the resulting loss of user data. According to the United States Small Business Administration, nearly 40% of small businesses fail after three years. In the world of tech start-ups, that number is significantly higher, as reportedly 9 out of 10 start-ups fail. And this doesn’t take into account web applications that are acquired or owned and closed by large companies.

The group Archive Team works to catalog and preserve digital history, but also keeps a Deathwatch of sites risking shutdown and provides advice for individuals on backing up our data. Though this is a wonderful project, we cannot assume that users will back up their data. When our services are closing down, we can do so gracefully. For example, the music streaming service Rdio closed its doors in 2015, but in doing so offered a farewell that included the ability for users to download CSV files of things such as their playlists and saved music to be imported into another service. As the site Hi.co shuttered, its founder Craig Mod committed to keeping the archive on the web for the next 10 years, making individual contributions exportable and producing five nickel-plated books of the site to be preserved. In an article about the shutdown, Mod wrote:

At the same time we understand the moral duty we took on in creating Hi.co — in opening it up to submissions and user generated content. There was an implicit pact: You give us your stories about place, and we’ll give you a place to put your stories. This was not an ephemeral pact.

Though we may not choose to nickel-plate our own services’ contents, providing exports will ensure that users are able to preserve their data if they choose to do so.

Further Reading

• “With Great Data Comes Great Responsibility” by Pascal Raabe
• “Archiving a Website for Ten Thousand Years” by Glenn Fleishman
• “Preserving Digital History” by Daniel J. Cohen and Roy Rosenzweig