Physical Layer Security: Link Signatures


Vulnerabilities in Link Signatures and Proposed Countermeasures

It is a common belief that link signatures between two different receivers and a transmitter are uncorrelated as long as the receivers are at least a quarter of a wavelength away from one another. However, this statement has not been proven in a mathematically rigorous manner so its validity is debatable. We propose that link signatures as they currently stand are vulnerable to attackers because it is possible for them to estimate the channel gain between a transmitter and receiver in an indoors, relatively static environment by using their own link signatures with the receiver. In our research, we will explore how feasible/plausible these attacks are in a real world setting and what some countermeasures may be to protect against them.

About Us


Week One

Introduction to UnCoRe

General overview to the program and introduction of how to do a research.

Project Assigned

We were assigned to the project of Link Signature, working with Dr. Shu

Literature Review

Read papers and learned more about link signature and key transmissions.

Week Two

Literature Review

Read over and learned more about wireless communication, transmitters, receivers, RF signals, channel gains, link signatures obtained from using channel state information and how these things can be exploited, etc. We also began looking into using machine learning algorithms with wireless networks.

Began Coding

Took some time getting familiar with using Matlab We will be implementing algorithms in the upcoming weeks to see if we can make better predictions.

Week Three


Looked into different machine learning algorithms and features that could be implemented.


Continued implementing algorithms in matlab.

Week Four


Look into SVM, Decision Tree Algorithms and other ML methods. Researched papers to get better ideas of what could be used as features


Worked on creating a script to run probabilistic methods and on scripts to run ML methods. Tested ML methods to see how efficient they were.

Week Five


Continued looking into Support Vector Machines, different kernel methods and other machine learning techniques that we had been previously looking into.


Tested multiple links with a different number of machine learning algorithms, changing the number of transmitter , receiver pairs used, and other features trying to improve results.

Week Six

Continued work from last week and even introduced work with Multivariate Linear Regression. Continued to try and use different kernel methods for SVM.

Week Seven

Run more test this time using the Multivariate Linear Rregrression and compraring the result with the Support Vector Machine.

Week Eight

Read papers to suggest new ideas for future works.

Week Nine

Prepare the poster presentation for Mid-Michigan Symposium for Undergraduate Research Experiences (Mid-SURE). Also we start to write the Deliverables for the last week.

Week Ten

Finish all the deliverables and turn them in.


First Presentation
Second Presentation
Third Presentation
Fourth Presentation
Midterm Presentation
Sixth Presentation
Seven Presentation
Final Poster Presentation
NSF Final Report