​A large aspect of this project relied on back-end development. The back-end faced the following challenges:

• Determining if EMG signals are unique to each person

• Developing methods to capture and validate uniqueness in numerical values that can be used for authentication

• Validating methods used with theory and specifically designed experiments

​

Litmus tests were designed for back-end data collection and analysis. The following are the different scenarios regarding how authentication could occur. That is, the situations where the authenticator should allow the user to login and situations where the authenticator should not allow the user to login. 

 

(i) is for model (ii) is for experimental series, all experiments should be repeated multiple times. As a person’s name is something that the person is proficient in typing. Thus it makes it difficult for person B to imitate.

 

Scenario #1

​

i) Person A types in his/her name as proficiently as he/she can.

ii) Person A types in his/her name as proficiently as he/she can.

 

Manual inspection: Look for similarities in plots.

Computed analysis: compute correlation for all 8 sensors.

Expected: Correlation value to be greater than 0.9 (close to 1).

Authentication: Should be authenticated.

​

Scenario #2

​

i) Person A types in his/her name as proficiently as he/she can.

ii) Person A types in his/her name in a more relaxed manner.

​

 

Manual inspection: Look for similarities in plots.

Computed analysis: compute correlation for all 8 sensors.

Expected: Correlation value not known.

Authentication: Should be authenticated.

​

Scenario #3

​

i) Person A types in his/her name as proficiently as he/she can.

ii) Person B types in person A's name.

 

Manual inspection: Look for dissimilarities in plots.

Computed analysis: compute correlation for all 8 sensors.

Expected: Correlation value should not be close to 1.

Authentication: Should not be authenticated.

​

Scenario #4

​

i) Person A types in his/her name as proficiently as he/she can. (e.g. "sayak")

ii) Person A types in another word that has same number of characters as the name and has keystroke pattern similar. (e.g. "dsusl"  QWERTY keyboard)

 

Manual inspection: Look for similarities in plots.

Computed analysis: compute correlation for all 8 sensors.

Expected: Correlation value not known.

Authentication: Should not/Should be authenticated.

​

Scenario #5

​

i) Person A types in his/her name as proficiently as he/she can.

ii) Person B types in something that has same length as name and similar keystroke pattern.

 

Manual inspection: Look for dissimilarities in plots.

Computed analysis: compute correlation for all 8 sensors.

Expected: Correlation value not known.

Authentication: Should not authenticated.

​

​

​

​