CLOUDFLARE: USING MACHINE LEARNING TO SECURE AGAINST CYBER ATTACKS

Cyber attacks got you down? Read how CloudFlare is building better security on the internet.

Today, it’s almost impossible to succeed in business without holding some aspects of your processes online. Whether it is storing company data on the backend, or offering the convenience of e-commerce transactions, each process is vulnerable to penetration by hackers or a bug that exposes key information. That is why detecting vulnerabilities and redirecting illegitimate traffic on a network before it has a chance to affect your systems is key to business performance. However, with the changing landscape and increasing complexity of network infrastructure, now more than ever it’s difficult to find these weaknesses. There is just too much information to synthesize for humans to monitor traffic events, and a rule-based algorithm does not sufficiently recognize patterns to secure Internet systems from penetration.5 That is where the startup CloudFlare comes in. CloudFlare’s mission is to build a better Internet, and integral to its success is using machine learning to prevent attacks on the network.7

CloudFlare offers Internet services by sitting between Internet servers, and the customer’s browsing activity to monitor traffic flows between the nodes. Founded in June 2017 and based out of San Francisco, CA, they offer a range of services such as content delivery, web-application firewalls, and denial-of-service (DDoS) Protection (when the cyber attack tries to interrupt the service being provided).6 Beneath these services is the use of machine learning. CloudFlare learns from harmful traffic trying to get onto the network, and continuously updates its algorithm to spread those learnings across the network.2 In this way, the CloudFlare system builds an understanding about what expected behavior should be on the system, so that when unexpected traffic hits the network, it will both stop the intrusion and integrate that pattern into its internal algorithm. This addresses the increasing complexity of today’s environment since the machine-learning algorithm doesn’t need to be told what to look for; instead, it follows an “unsupervised” approach. The developer doesn’t prescribe the algorithm with what an attack will look like but instead the algorithm learns on its own. Machine learning is an important step in the cyber security space as it moves to be proactive against vulnerabilities instead of reactive. That is, we detect malicious actors within the system before they are able to create harm.

In the next two years, the company is focusing on building out its mobile platform. Just seven months ago, CloudFlare announced their internet service provider platform (otherwise known as a DNS service) that offers privacy of the user’s browsing history. Now they are bringing that to mobile products.3 What makes this product different from its existing products, and other items on the market, is its ability to offer users privacy on their iPhone as well as an extremely fast browsing experience.6 This is extremely important today with so many transactions occurring on mobile phones, and the need for both speed and protection.

In in the long term, CloudFlare is building relationships with governments to help protect against potential hacks during elections. As of now, ten states implemented CloudFlare services.6 In the next two to ten years, it is likely to see more CloudFlare integration with electoral systems and other high-profile internet connected-systems.4 This comes at a crucial time when it is important to instill trust and security in public systems.

In addition to the steps mentioned above, the company should also focus its attention on protection for IoT devices. Given that so much of its growth comes from businesses expanding their portfolio of networked devices like IoT and other internet-enabled sensors, I recommend that CloudFlare consider how they can provide protection for them.1 This way, they are not just preventing attacks coming into the system, but attacks that can exist within the system as well.

Despite these recommendations, CloudFlare is making great progress to mend the problems that exist in the cyber security space. However, machine learning is just one aspect of the solution. Human error such as weak passwords, can mitigate even the strongest network defense. Thus, if CloudFlare wants to be all encompassing solution for business’ security, it should also address: how are you promoting other security best practices? Another key thing to consider since this is a service that can be purchased, is how are you preventing a hacker from acquiring the product and reverse engineering it? While CloudFlare makes the product easy to acquire, it is important they don’t compromise their product by letting it get into the wrong hands.

Word Count: 744

____________________________________________________________________________________________________

Endnotes

1 “Advanced DDoS Protection and Mitigation.” Cloudflare, www.cloudflare.com/ddos/.

2 Council, Young Entrepreneur. “How Machine Learning Will Impact Online Security This Year.” Forbes, Forbes Magazine, 8 May 2017, www.forbes.com/sites/theyec/2017/05/08/how-machine-learning-will-impact-online-security-this-year/#1cf11c2c18c3.

3 Fagioli, Brian. “Cloudflare Launches 1.1.1.1 Consumer DNS Service with a Focus on Privacy.” BetaNews, 1 Apr. 2018, betanews.com/2018/04/01/cloudflare-dns-privacy-four-ones/.

4 Hatmaker, Taylor. “Cloudflare Recruits State and Local Governments for Free Election Site Security Program.” TechCrunch, TechCrunch, 19 July 2018, techcrunch.com/2018/07/19/cloudflare-athenian-project/.

5 Mytton, David. “Have the Big Cloud Providers Won the Machine Learning Advantage?” Medium, Medium, 4 Dec. 2017, medium.com/@davidmytton/have-the-big-cloud-providers-won-the-machine-learning-advantage-c91e8da44f44.

6 Whittaker, Zack. “Cloudflare Rolls out Its 1.1.1.1 Privacy Service to IOS, Android.” TechCrunch, TechCrunch, 11 Nov. 2018, techcrunch.com/2018/11/11/cloudflare-privacy-dns-service-ios-android/.

7 Zatlyn, Michelle. “That’s Freaking Awesome: CloudFlare Automatically Learns How to Stop New Attacks.” The Cloudflare Blog, The Cloudflare Blog, 26 Jan. 2018, blog.cloudflare.com/thats-freaking-awesome-cloudflare-automatical/.

 

 

Previous:

Using machine learning to improve lending in the emerging markets

Next:

Twine Health: Pioneering Machine Learning in Chronic Disease Management

4 thoughts on “CLOUDFLARE: USING MACHINE LEARNING TO SECURE AGAINST CYBER ATTACKS

  1. Great article! This is such a needed technology today as we are seeing the increase of security attacks across the internet. I think the CloudFlare has great potential to make a big impact with its work on government projects as the increase of state sponsored cyber attacks continues to skyrocket. Security has been top of mind for many consumers over the past two years since the issues with the Facebook Ads. I think that with machine learning, CloudFlare has the upper hand in defending target devices from malicious actors. I think that it is smart for them to be moving into the mobile space as that it where the majority of internet communications is taking place and where it is least protected. I’ll have to see what I can do to get CloudFlare services on my phone!

  2. Nice piece! As the CloudFlare product becomes more prevalent does that introduce some level of vulnerability in itself? It will be interesting to see the application in IoT devices given that one of the main vulnerabilities is out of date software on devices. Perhaps if one part of CloudFlare’s system is deemed vulnerable, does that make out of date devices a target? Alternatively, if the machine learning is at the network level and local hardware is irrelevant, that seems like a great opportunity for addressing that very criticism of IoT!

  3. Great read! It’s interesting to see the growth of companies like CloudFlare as the internet has becoming increasingly complex. I don’t see the need for companies like this diminishing anytime in the future, although I wonder how the future of privacy will affect them – some individuals (e.g. https://www.theguardian.com/world/2014/aug/03/internet-death-privacy-google-facebook-alex-preston) believe that personal privacy should/will be dissolved significantly over time (potentially to a point where personal privacy will no longer exist in a meaningful way), so I wonder how CloudFlare thinks about that in terms of 1) the likelihood of this being a possibility for the future (and how far into the future) and 2) what it could mean for their business. One the second point, how does the “death of privacy” change the needs of CloudFlare’s consumers? Would CloudFlare offer increasingly sophisticated personal privacy solutions? Or would they accept the limitations of personal privacy and pivot to “making the internet a better place” initiatives?

    Beyond privacy concerns, I agree that proactive counter measures against cyber attacks are crucial and that these will only become more and more proactive. Perhaps it will even get to a point where a reactive counter measures are considered no longer effective. The IoT devices are another fascinating topic for cyber security – it’s interesting to see how “breach-able” the current IoT devices on the market already are. I’m a little surprised CloudFlare hasn’t moved more quickly around IoT device protection considering the level of media attention and press those breaches get; is this a case of IoT device breaches being blown out of proportion by the media or the case of CloudFlare as a startup being too small to react as quickly as we would want/expect?

  4. Mollie – this is a great overview of machine learning and a very exciting application to a very suitable field (cyber security)! You did a great job of highlighting how pervasive cyber security threats have become – basically anyone connected to the Internet (either within or outside of any organization) can pose a threat to an organization’s network. Also, as you mentioned, I completely agree that human monitoring of potential threats is impractical given the explosive growth and complexity of recent threats. I personally think that the edge for these machine learning players is scale – you need the most amount of data to detect the most patterns and that ultimately will drive a virtuous cycle / self-reinforcing network effect (better algorithms will attract more customers that bring in more data and further improves the algorithms). Getting that flywheel into motion is the tough part, but it seems like CloudFlare has scaled very nicely. (There are recent rumors of a potential IPO that would value the company at $3.5Bn+.) Will be interested to see how the company continues to grow and benefits from scale advantages!

Leave a comment