Adversarial Learning challenges for Cybersecurity

According to the European Union Agency for Network and Information Security (ENISA) Threat Landscape report 2017, firms face millions of cyber threats including malware, web-based attacks, phishing, ransomware, botnets, etc. Detecting advanced persistent threats is a tough task since the real goals of these attacks stay undetected for a long period of time.

To overcome the ever-increasing new threats being designed by attackers, firms are increasingly relying on augmenting their security systems with advanced Machine Learning and Deep Learning techniques to protect their data and network from malicious attacks. However there is a growing realization that to defend your systems you need to learn how to attack them first and Adversarial Machine learning algorithms are finding use in this space.

Adversarial Machine learning is the study of machine learning vulnerabilities in adversarial environments. Much like how a hacker might take advantages of a firewall vulnerability to gain access to a web server, a machine learning system can itself be targeted to serve the goals of an attacker. Hence before putting such solutions in production, it is crucial that machine learning system designers build safeguards to preempt these attacks.

In this talk we start with a quick overview of the overall landscape of Cyber threats, understand some of the commonly used threat hunting methodologies and focus on some real world uses of Machine learning solutions to augment security. The second part of the talk focuses on developing an understanding of Adversarial Machine Learning algorithms, how they can be used to bypass security solutions build using Machine Learning algorithms. In the third section, we will demonstrate how Adversarial techniques can be developed to subvert solutions built using Machine Learning algorithms. We will also focus on some of the countermeasures to adopt which can help in protecting Machine learning based security systems


Outline/Structure of the Talk

  1. Overview of Cyber Threat Landscape and Real world use of Machine Learning in Security -10 minutes
  2. Conceptual understanding of how Adversarial Machine Learning algorithms works and ways in which they can be used for subverting security systems -10 minutes
  3. Demonstration of how we can apply Adversarial Machine Learning techniques to subvert security solutions built using Machine Learning algorithms -20 minutes
  4. Q&A - 5 minutes

Learning Outcome

  • A better understanding of Cyber Threat landscape and how Machine Learning solutions are used in the space.
  • Conceptual understanding of Adversarial Machine Learning techniques and ways they can be used to subvert security solutions
  • Develop an understanding of how to develop adversarial algorithms that can bypass security systems

Target Audience

Machine learning enthusiasts, Network Security, Developers

Prerequisites for Attendees

Awareness of Machine Learning concepts

General Interest in the field of Cybersecurity and Artificial Intelligence

schedule Submitted 9 months ago

Public Feedback

comment Suggest improvements to the Speaker
  • Vishal Gokhale
    By Vishal Gokhale  ~  8 months ago
    reply Reply
    Hi Gaurav, Thanks for submitting the proposal. Would you also discuss the use-cases you implemented at Fidelity/ Otherwise ? ODSC is a practitioner conference and it would be important for the audience to learn from your experiences/challenges faced when implementing these ideas. Thanks, Vishal
    • Gaurav Shekhar
      By Gaurav Shekhar  ~  8 months ago
      reply Reply

      Hi Vishal,

      Thanks for reviewing my proposal. As you are aware Adversarial learning is an area where a lot of research is currently underway. At present we primarily look at Adversarial learning as one of the ways to enhance AI safety by increasing the robustness of the Machine Learning models developed. We have developed machine learning models in the areas of prediction of critical events that can cause the failure of important services, network latency, detecting correlations between events, early warning detection solutions etc. Recently on some of the above use cases, we have experimented with leveraging Adversarial learning concepts to generate test data that can challenge the model's output; the larger objective here is to explicitly train the models on these adversarial examples and make them more robust. I will be happy to give an overview of how we used Adversarial Learning in one of the use cases listed above.


  • Liked Gaurav Shekhar

    Gaurav Shekhar - AIOps - Prediction of Critical Events

    45 Mins
    Case Study

    With the rise of cloud, distributed architectures, containers, and microservices, a rise in data overload is visible. With growing amounts of DevOps processes; alerts, repeated mundane jobs etc. have put new demands to both synthesize meaning from this influx of information and connect it to broader business objectives.

    AIOps is the application of artificial intelligence for IT operations. AIOps uses machine learning and data science to give IT operations teams a real-time understanding of any issues affecting the availability or performance of the systems under their care. Rather than reacting to issues as they arise in the application environment, AIOps platforms allow IT operations teams to proactively manage performance challenges faster, and in real-time

    This case study focuses on solving the following business needs:

    1. With an ever-increasing rise in alerts, a large number of incidents were getting generated. There was a need to develop a framework that can generate correlations and identify correlated events, thereby reduce overall incidents volume.

    2. For many incidents a reactive strategy does not work and can lead to a loss of reputation; there was a need to develop predictive capabilities that can detect anomalous events and predict critical events well in advance.

    3. Given the pressures of reducing the Resolution time and short window of opportunity available to the analysts, there was a need to provide search capabilities so that the analysts can have a head start as to how similar incidents were solved in past.

    Data from multiple systems sending alerts, including traditional IT monitoring, log events in text format, application and network performance data etc were made available for the PoC.

    The solution framework developed had a discovery phase where the base data was visualized and explored, a NLP driven text mining layer where log data in text format was pre-processed, clustered and correlations were developed to identify related events using Machine Learning algorithms. Topic Mining was used to get a quick overview of a large number of event data. Next, a temporal mining layer explored the temporal relationship between nodes and cluster groups, necessary features were developed on top of the associations generated from temporal layers. Advanced Machine learning algorithms were then developed on these features to predict critical events almost 12 hours in advance. Last but not the least a search layer that computed the similarity of any incident with those in Service Now database was developed that provided analysts insights readily available information on similar incidents and how they were solved in past so that the analysts do not have to reinvent the wheel.