Part II (50 points). Using the same dataset (UCI Machine Learning Adult data) to study differential privacy (centralized) . Laplace Mechanism: query the average age of the records with age greater than 25 (considering each record belonging to an individual adult). Inject Laplacian noise to the query result (average age) to ensure 0.5-differential privacy and 1-differential privacy. 1. in case of e0.5, generate 1,000 results for the query over the original dataset, and generate 1,000 results for the query over each of three other datasets (removing a record with the oldest age; removing any record with age 26; removing any record with the youngest age). (6 points) 2. in each of the above 4 groups of 1,000 results, round each number to two decimal places, define a measure and utilize it to validate that each of the last 3 groups of results and the original results are 0.5-indistinguishable (6 points) 3, repeat all the above for = 1, utilize the above measure to validate that each of the last 3 groups of results and the original results are indistinguishable. (6 points) 4. define another measure and utilize it to justify that the distortion of the 4,000 results for e-1 is less than that of e 0.5. (7 points) . Exponential Mechanism: query the most frequent "Education" result. Design an exponential mechanism (randomized) to ensure e-differential privacy for the query. Repeat a the procedures for Exponential mechanism (E0.5 and 5. in case of0.5, generate 1,000 results for the query over the original dataset, and generate 1,000 results for the query over each of three other datasets (removing a record with the most frequent "Education"; removing any record with the second most frequent "Education"; removing any record with the least frequent "Education". 6 points) 6. in each of the above 4 groups of 1,000 results, define a measure and utilize it to validate that each of the last 3 groups of results and the original results are 0.5-indistinguishable. (6 points) 7. repeat al the above for e-1, utilize the above measure to validate that each of the last 3 groups of results and the original results are 1 indistinguishable. (6 points) 8. define another measure and utize it to justify that the distortion of the 4,000 results for e- is less than that of e0.5. (7 points) For each task (8 in total), submit a source code file (can be only a few lines) and a result file, including the quantitative results and measure (if requested). The files are named with the prefix .hwl-II-" (e.g., hud-II-1.java, hw1-11-1.txt) Part II (50 points). Using the same dataset (UCI Machine Learning Adult data) to study differential privacy (centralized) . Laplace Mechanism: query the average age of the records with age greater than 25 (considering each record belonging to an individual adult). Inject Laplacian noise to the query result (average age) to ensure 0.5-differential privacy and 1-differential privacy. 1. in case of e0.5, generate 1,000 results for the query over the original dataset, and generate 1,000 results for the query over each of three other datasets (removing a record with the oldest age; removing any record with age 26; removing any record with the youngest age). (6 points) 2. in each of the above 4 groups of 1,000 results, round each number to two decimal places, define a measure and utilize it to validate that each of the last 3 groups of results and the original results are 0.5-indistinguishable (6 points) 3, repeat all the above for = 1, utilize the above measure to validate that each of the last 3 groups of results and the original results are indistinguishable. (6 points) 4. define another measure and utilize it to justify that the distortion of the 4,000 results for e-1 is less than that of e 0.5. (7 points) . Exponential Mechanism: query the most frequent "Education" result. Design an exponential mechanism (randomized) to ensure e-differential privacy for the query. Repeat a the procedures for Exponential mechanism (E0.5 and 5. in case of0.5, generate 1,000 results for the query over the original dataset, and generate 1,000 results for the query over each of three other datasets (removing a record with the most frequent "Education"; removing any record with the second most frequent "Education"; removing any record with the least frequent "Education". 6 points) 6. in each of the above 4 groups of 1,000 results, define a measure and utilize it to validate that each of the last 3 groups of results and the original results are 0.5-indistinguishable. (6 points) 7. repeat al the above for e-1, utilize the above measure to validate that each of the last 3 groups of results and the original results are 1 indistinguishable. (6 points) 8. define another measure and utize it to justify that the distortion of the 4,000 results for e- is less than that of e0.5. (7 points) For each task (8 in total), submit a source code file (can be only a few lines) and a result file, including the quantitative results and measure (if requested). The files are named with the prefix .hwl-II-" (e.g., hud-II-1.java, hw1-11-1.txt)
