Fix issue #12 with clean extracted databases.

ATMEGA_AES_v1/ATM_AES_v1_variable_key/Readme.md

@@ -8,17 +8,17 @@ $ mkdir -p ASCAD_data/ASCAD_databases
 $ cd ASCAD_data/ASCAD_databases
 $ wget https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190730-071646/atmega8515-raw-traces.h5
 $ mv atmega8515-raw-traces.h5 ATMega8515_raw_traces.h5
-$ wget https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190730-071619/ascad-variable.h5
+$ wget https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190903-083349/ascad-variable.h5
 $ mv ascad-variable.h5 ASCAD.h5
-$ wget https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190730-071628/ascad-variable-desync50.h5
+$ wget https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190903-084119/ascad-variable-desync50.h5
 $ mv ascad-variable-desync50.h5 ASCAD_desync50.h5
-$ wget https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190730-071636/ascad-variable-desync100.h5
+$ wget https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190903-084306/ascad-variable-desync100.h5
 $ mv ascad-variable-desync100.h5 ASCAD_desync100.h5
 </pre>
 
 Please be aware that all these steps should **download around 71 GB** of data.
-You can selectively download only the extracted databases (`https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190730-071619/ascad-variable.h5`
-and so on) that weight a more reasonable 281 MB each.
+You can selectively download only the extracted databases (`https://static.data.gouv.fr/resources/ascad-atmega-8515-variable-key/20190903-083349/ascad-variable.h5`
+and so on) that weight a more reasonable 418 MB each.
 
 ### Raw data files hashes
 
@@ -27,11 +27,11 @@ The data files SHA-256 hash values are:
 <hr>
 
 **ASCAD/ATMEGA_AES_v1/ATM_AES_v1_variable_key/ASCAD_data/ASCAD_databases/ASCAD.h5:**
-`515dcd3e11a6ece66415800df53db35a0ddb8106527b6ae75a370e33b66cea8c`
+`d834da6ca5a288c4ba5add8e336845270a055d6eaf854dcf2f325a2eb6d7de06`
 **ASCAD/ATMEGA_AES_v1/ATM_AES_v1_variable_key/ASCAD_data/ASCAD_databases/ASCAD_desync50.h5:**
-`72545013176518c64292ce84d28b274fbcb8d5354e813080e554b4323e5945bb`
+`0fa048bf42b9d8bbf9514770072c070175637e1b6fb6da370e2b020b1ecca673`
 **ASCAD/ATMEGA_AES_v1/ATM_AES_v1_variable_key/ASCAD_data/ASCAD_databases/ASCAD_desync100.h5:**
-`b992356f3eb16b75fb7c7197951aa4f670cfaaad4b9920ab7a85e7089bfc5fec`
+`cb82c553b84c29454ea23ec043730ed845e17c2a0e261853afc11e028fdf2710`
 **ASCAD/ATMEGA_AES_v1/ATM_AES_v1_variable_key/ASCAD_data/ASCAD_databases/ATMega8515_raw_traces.h5:**
 `6f13d7c380c937892c09b439910c4313d551adf011d2f4d76ad8b9b3de27b852`
 
@@ -42,16 +42,16 @@ downloaded and decompressed the raw data file as explained in [the previous sect
 
 ## The ATMega8515 SCA traces databases
 
-This database contains 150,000 traces from the acquisition campaign compiled in a [HDF5](https://en.wikipedia.org/wiki/Hierarchical_Data_Format) file of 71 GB named `ATMega8515_raw_traces.h5`. The structure of this HDF5 file is described in the article ["Study of Deep Learning Techniques for Side-Channel Analysis and Introduction to ASCAD Database"](https://eprint.iacr.org/2018/053.pdf).
+This database contains 300,000 traces from the acquisition campaign compiled in a [HDF5](https://en.wikipedia.org/wiki/Hierarchical_Data_Format) file of 71 GB named `ATMega8515_raw_traces.h5`. The structure of this HDF5 file is described in the article ["Study of Deep Learning Techniques for Side-Channel Analysis and Introduction to ASCAD Database"](https://eprint.iacr.org/2018/053.pdf).
 
 We emphasize that these traces are **not synchronized**. The key is **variable** in the following fashion: one acquisition every three uses a **fixed** key and the other two out of three use a **random** key.
 
 
 ## The ASCAD databases
 
 The databases, which are HDF5 files, basically contain two labeled datasets:
-  * A 100,000 traces **profiling dataset** that is used to train the (deep) Neural Networks models.
-  * A 50,000  traces **attack dataset** that is used to check the performance of the trained models after the
+  * A 200,000 traces **profiling dataset** that is used to train the (deep) Neural Networks models.
+  * A 100,000  traces **attack dataset** that is used to check the performance of the trained models after the
 profiling phase. 
 
 The ASCAD database is in fact extracted from the `ATMega8515_raw_traces.h5` file containing raw traces: in order to avoid useless heavy data processing, a window of 1400 points of

ATMEGA_AES_v1/ATM_AES_v1_variable_key/example_generate_params

@@ -2,8 +2,8 @@
 "traces_file" : "ATMEGA_AES_v1/ATM_AES_v1_variable_key/ASCAD_data/ASCAD_databases/ATMega8515_raw_traces.h5",
 "labeled_traces_file" : "ATMEGA_AES_v1/ATM_AES_v1_variable_key/ASCAD_data/ASCAD_databases/ASCAD_variable_desync100.h5",
 "profiling_index" : [n for n in range(0, 300000) if n%3 !=2],
-"attack_index" : [n for n in range(2,3000,3)],
+"attack_index" : [n for n in range(2,300000,3)],
 "target_points" : [n for n in range(80945, 80945+1400)],
-"profiling_desync" : 0,
+"profiling_desync" : 100,
 "attack_desync" : 100
 }