Cyber security has always been challenging, both in detection and prevention. The field is a source of many machine learning challenges, critical for solving security problems. In addition to that, the constant arms race between the attacker and defender, make the field rapidly advance. Recent advancement in machine learning and deep neural network research have enabled novel ways to approach these problems.
We will present some of the methods and products developed at Palo Alto Networks to build better security products.
18:00-18:30 Mingling, Snacks & Drinks
18:30 - 19:10 Detecting Targeted Attacks with Machine Learning:
Overview and Real-World Examples
19:10-19:40 Machine Learning Challenges in Cybersecurity
19:50-20:30 Using Deep Neural Networks for Malware Detection
Lecture 1: Detecting Targeted Attacks with Machine Learning: Overview and Real-World Examples
Speaker: Yinnon Meshi
We will present some real world applications of machine learning in the field of cyber security.
We will start with a high level overview of cyber attacks conducted by advanced attackers (namely APT).
Then we'll talk about the difficulties of common ML approaches (such as anomaly detection) in this use case, and what is our approach to these problems.
Lastly, we'll show a real-world example for one of our research projects, of using co-training and concept decomposition.
No prior knowledge in cybersecurity is required, though it will be helpful.
Lecture 2: Machine Learning Challenges in Cybersecurity
Speaker : Idan Amit
* In this talk we will discuss our recently published paper "Machine Learning in Cyber-Security-Problems, Challenges and Data Sets" that was accepted to EDSMLS workshop of the AAAI conference
Some machine learning challenges reappear in various cyber problems. These challenges might prevent successful detection. On the other hand, coping with them can lead to a huge performance advantage. We will present the machine learning challenges, the security problems in which the challenges appear and ways to cope with these challenges.
Examples of these challenges are: Attacker-Defender game, Imbalanced datasets, Lack of labeled samples and ground truth, Redundancy , Failure of usual metrics, Domain adaptation and Concept drift.
Lecture 3: Using Deep Neural networks for malware detection
Speaker: Yigal Weinberger
Static malware detection is the act of analyzing file’s binaries in order to determine whether it is a malware or not. In this lecture we will show several architectures that were tested for optimal performance for this task. The session will include a technical overview of this specific task and the lessons learned from developing them.
Yinnon is a Senior Data Scientist at Palo Alto Networks cyber analytics research team.
Prior to that, Yinnon served as an officer at IDF's cyber security and cryptography unit.
Yinnon holds a B.Sc. (summa cum laude) and a M.Sc. in Electrical Engineering from the Technion.
Yigal is the Lead Data Scientist of Palo Alto Networks endpoint security research team. Having over 8 years of experience as a machine learning researcher he now consider himself mostly as a Data Scientist Evangelist, and enjoy spending most of his free time as a Senior lecturer of the course: Data Science Professional