CIS- 597: Machine Learning in Cyber Security

Total Marks: 50

Project marks: 25 (Based on previous project evaluation). However, students will need to answer questions related to projects in the final exam to be given these marks.

Course exam marks: 25 (10 for Neural Networks, 15 for other).

Time: 1.5 hours

Good luck.

The teaching staff has posted a new general resource.

Title: Neural Networks (upto sld. 63)
http://www.piazza.com/class_profile/get_resource/jr04f20wksv6ca/jw090jeyz6p51p


You can view it on the course page: https://piazza.com/pieas.edu.pk/spring2019/cis597/resources

I have been looking for commercial cybsec software that utilize ML. Here is one. You can View the talk at: https://www.brighttalk.com/webcast/13507/247091/battling-unknown-malware-with-machine-learning

Please provide your feedback at: https://forms.gle/sVKfrN9xTbspNHje8

Thank you.

-F

1. What factors control the amount of weight update of a neural network?
2. What factors can forestall the optimization of a neural network? How can you remedy it?
3. What is the vanishing gradient problem?
4. How is a deeper model more efficient for function approximation in comparison to a shallow network?
5. If deep models are efficient, why do we need GPUs to learn from them?
6. Why is the optimization of a neural network with a non-convex loss activation function more difficult?
7. What are the hyper-parameters associated with a neural network? What is the role of each?
8. How is the addition of multiple layers in a neural network related to the following concepts:
   • Feature transformations
   • Folding of the feature space
   • change of distance metric
   • change of kernel

Here are your grades for the sessional. If there is any discrepancy in your marks, please discuss it with me ASAP and no later than 1100HRS today (Thursday, May 16, 2019). You may want to check the posted grades for different assignments and quizzes etc. to make sure that they have been correctly added in this list. The project will constitute 25% marks in the final as well for a total of 40% of the course as outlined in the initial marks distribution. Please note that these marks are only tentative at this time.

These questions may be included in the final examination for project grade assignment as discussed in the class.

1. What is Homomorphic Encryption? How is it useful for machine learning?
2. What is super-resolution imaging? What are its applications?
3. What is neural voice cloning?
4. What is meant by tf-idf?
5. What is "Learning with Rejection"?
6. What is Reinforcement Learning?
7. How can a pre-trained neural network be used for feature extraction?
8. How can you classify attacks in network traffic?
9. How can template matching be used for object detection in images?
10. How can you classify software into benign and malware?
11. What is the role of iid assumption in cross-validation?
12. What is meant by static and dynamic analysis?
13. What is a quantum computer?
14. How can one extract features from textual data?

Great work everyone! It was difficult to assign top grades. Please complete your outstanding work if any so we can conclude the project. The grades are pretty much final. However, if you feel there is a major discrepancy, please discuss it with me by 1100HRS today (Thursday May 16, 2019).