Initial Process Flow

[FumiyukiKato]

:note:0cb5ceb6-d4b3-4804-9bb0-8e7e78fd96e1

Process Flow

1. Generate and hold random m queries with several elements(100-10000?) respectively, inside SGX.
2. Merge queries and mapping elements into PCT space.
3. Place central data into external storage.
4. First, try a simple Bloom filter(?) where the key is geohash, and
5. Sequential reading as a set of chunks.
6. Process PCT logic. and construct all responses.
7. Measure processing time. average latency(=throughput),

options

• (option) Prametize query accuracies
• (option) Encryption
• (option) Response with the result data

[FumiyukiKato]

(Bloom filter with attached data reference?)

[FumiyukiKato]

Similar Data Structure

Sun, Jingchao, Rui Zhang, and Yanchao Zhang. "Privacy-preserving spatiotemporal matching." 2013 Proceedings IEEE INFOCOM. IEEE, 2013.

https://www.eecis.udel.edu/~ruizhang/paper/sun-INFOCOM13.pdf

[FumiyukiKato]

Base Idea

Query multiplexing and sequential all data flow

Tamrakar, Sandeep, et al. "The circle game: Scalable private membership test using trusted hardware." Proceedings of the 2017 ACM on Asia Conference on Computer and Communications Security. 2017.

https://arxiv.org/pdf/1802.09085.pdf

[FumiyukiKato]

• [x] Generate and hold random m queries with several elements(100-10000?) respectively, inside SGX.
• [x] Merge queries and mapping elements into PCT space.
• [x] Place central data into external storage.
• [ ] First, try a simple Bloom filter(?) where the key is geohash, and
• [ ] Sequential reading as a set of chunks.
• [ ] Process PCT logic. and construct all responses.
• [ ] Measure processing time. average latency(=throughput),

[FumiyukiKato]

$ bin/app 100000 data/query/generated-client-query-qs-4500-cs-2000-20200614170915.json data/central/generated-central-data-1000000-20200614022055.json true