---
tags: IsoPhase
---

# Data we have:
- Paper [link](https://ndss22-summer.hotcrp.com/paper/142#response)
- Rebuttal [link](https://docs.google.com/document/d/1LXK59O_O6saOXSgnhAOyHwsP2zMqRfgPUgbP_Q7rUFY/edit)
- Syscall categorization for each phase: [link](https://docs.google.com/spreadsheets/d/1_I-K1npgbJlMXQ_kpFBAhC_JEczcqEdSgA8Rruf_Xxc/edit#gid=710041261)
- API categorization of each framework:
- [OpenCV](https://docs.google.com/spreadsheets/d/1EXV_ASt9jQ6SQDGX7v6deaz6SGfN6rxgQ97xyXv_9x8/edit#gid=287301385)
- [Tensorflow]()
- [Caffe](https://docs.google.com/spreadsheets/d/1EXV_ASt9jQ6SQDGX7v6deaz6SGfN6rxgQ97xyXv_9x8/edit#gid=1247619853)
- [Torch](https://docs.google.com/spreadsheets/d/1EXV_ASt9jQ6SQDGX7v6deaz6SGfN6rxgQ97xyXv_9x8/edit#gid=984155587)

- Stateful APIs: [link](https://docs.google.com/spreadsheets/d/1uwtW_F0GBKvMa2Ls30bjSDJzLPjhATMDtlW4VBdvdoc/edit#gid=1510934009)
-
# Unclear:
- RBQ1: Intraphase attacks
- What experiment needed or what needs to be looked up?
- RBQ2: Indirect API calls
- We have coverage
- No statistics on the indirect API calls
- Do we need?

# Other work:
- RCQ1: List down isolation based techniques from related works
- Need 5

# Experiments needed:
- Run XXX number of applications for the following:
- RAQ1: Analysis of Number of API invocations
- We have APIs categorized in each phase
- Need to run for actual numbers
- Result: Distribution of invocation of APIs over different phases
- RAQ2: Overhead analysis
- Temporal analysis of invocations of APIs would give the following:
- Number of transitions
- Number of IPC calls
- Amount of data copied per call and max (***Not feasible***)
- RBQ8 and RBQ10: APIs not categorized:
- List down APIs in application that don't fall under our categorization.
- What are the frameworks?