Day 12: Local differential policy
As suggested by @Aniket Thomas, I move on to Lesson 5 today. Yesterday I watched the introduction video which explained that Global differential policy is adding noise on the database as a whole while Local differential policy is adding noise to each individual data.
The database owner called Trusted Curator.
Today, I am moving to the next video. The example given about jaywalking with flipping coin "noise" is very clearly explained and I am better at understanding the sensitivity of data. The dataset without noise would be not too sensitive or very sensitive depending on the query. We can attack the dataset using differencing attack to violate the privacy of dataset. One way to preserve the privacy of the dataset, we can add noise to each of the data. The example is by adding the flipping coin twice technique. If the first flip is head, answer honestly, if it is tail, answer depends on the second flip (head is true and tail is false). Hence, half of the time they will answer honestly.
If 70% of the people do jaywalking in the true dataset. It means The result in the 'noise' dataset will show as 60% of people do jaywalking (it is from (70% + 50%) / 2, the average).
The good thing is in real scenario, we did not know the 70% yet, but we can get the 70% from (x% + 50%) / 2 = 60%. We can get the approximate percentage of x % without knowing the true dataset.
The drawback of adding noise is the accuracy will be lower. Depending on the size of the dataset, the more of the data or in infinite dataset, the noise will be diminished. However, in the small dataset, the noise will heavily effect the accuracy. See my working on notebook for the project,
What we want to achieve in implementing differential policy are
- Reach most accurate query with greatest amount of privacy
- Greatest fit with trust models in the actual world. If we can trust each other we do not need to add noise that will lead to reduced accuracy. We will only add noise in the untrusted environment. For example, if we asked that people ever commited to crime, then we need to add noise to get an honest answer.
The database owner called Trusted Curator.
Today, I am moving to the next video. The example given about jaywalking with flipping coin "noise" is very clearly explained and I am better at understanding the sensitivity of data. The dataset without noise would be not too sensitive or very sensitive depending on the query. We can attack the dataset using differencing attack to violate the privacy of dataset. One way to preserve the privacy of the dataset, we can add noise to each of the data. The example is by adding the flipping coin twice technique. If the first flip is head, answer honestly, if it is tail, answer depends on the second flip (head is true and tail is false). Hence, half of the time they will answer honestly.
If 70% of the people do jaywalking in the true dataset. It means The result in the 'noise' dataset will show as 60% of people do jaywalking (it is from (70% + 50%) / 2, the average).
The good thing is in real scenario, we did not know the 70% yet, but we can get the 70% from (x% + 50%) / 2 = 60%. We can get the approximate percentage of x % without knowing the true dataset.
The drawback of adding noise is the accuracy will be lower. Depending on the size of the dataset, the more of the data or in infinite dataset, the noise will be diminished. However, in the small dataset, the noise will heavily effect the accuracy. See my working on notebook for the project,
What we want to achieve in implementing differential policy are
- Reach most accurate query with greatest amount of privacy
- Greatest fit with trust models in the actual world. If we can trust each other we do not need to add noise that will lead to reduced accuracy. We will only add noise in the untrusted environment. For example, if we asked that people ever commited to crime, then we need to add noise to get an honest answer.
Comments
Post a Comment