-
Remove separate process for experiment as it is obsolete now.
-
Clean up the repo.
-
Use CPU times instead of CPU percentage, per step instead of over the total process
-
Improve memory profiler. Maybe it helps if we store everything in variables. Current zero-measurement could be explained by the usage of the stack instead of memory.
-
Only measure network traffic which is not directly related to storage size
- Data update
- Policy update
- Not relevant: Attribute authority information request (name + attributes)
- Location/meta share with doctor
- Update keys (for TAAC)
- Key update (ciphertext update keys, key update keys)
-
Update keys in file storage -> are in network
-
Update keys: over all keys or just a few? -> all keys
-
User secret keys in storage are about twice the size of the keygen result. -> There are two authorities :P
-
CPU times is not over the same code, but also includes file io. That is why it is higher than profile times
-
secret keys request (keygen) + response -
public keys request + response -
create record + response -
update record -
update policy - location/meta share with 2nd user
-
fetch record + response - update keys (from authorities to users, if possible)
- Contribution in introduction
- Measurements: setup vs encryption/decryption
-
Export all measurement in a helpful format - Differentiate between time required for symmetric and ab encryption
-
Run each experiment in a subprocess to enable CPU usage determination -
<> Store all keys in order to measure storage costs-
Client: owner keys -
Client: registration data -
Client: secret keys -
Attribute authority: attribute public/private -
Central authority: Global parameters
-
- [?] Encrypt as a stream, instead of all in memory (see issue 2 of charm).
- Test experiments
-
RD13Implementation seems to have an authority attributes storage size of 6 bytes. This is most probably incorrect :( -
Measure one factor at a time(Better testing required) -
<> Fix pickle_serializer.public_keys for TAAC12 - Only do encryption and decryption on raspberry: mobile is not going to be an authority
- in dacmacs: make decryption_keys part of decryption as it is required for each decryption?
- This leaves only TAAC to have decryption keys (per time period)