Skip to content

Latest commit

 

History

History
164 lines (115 loc) · 4.3 KB

readme.md

File metadata and controls

164 lines (115 loc) · 4.3 KB

File tree

server
├── bat_generator.py # used for data collection
├── collect_data.py # used for data collection
├── evaluation # reproduce the results in paper
├── raw_data # raw data collected from devices
├── readme.md
├── requirements.txt
└── tasks # used for data collection

Data collection

Before you start collecting data, make sure you have installed the programs in "../../Device-porting" and connected the device to your PC.

If you only want to reproduce the results in the paper, go to the next section.

1.Connect PINs with dupont wire as below:

esp32s2

P5 -- P1

P16 -- P17

P18 -- P19

STM32F103

PA2 -- PB5

STM32F429

PA5 -- PC7

2.Bat generate

$ pip install -r requirements.txt
$ python bat_generator.py --device_number n1 n2 n3 --port COMx1 COMx2 COMx3 --only_sram 1

Replace COMxi with your serial port number,and ni is the device number.

3.Run the shell

$ run_get_dataset_*.bat

If you want to collect all types of fingerprints, replace --only_sram 1 with --only_sram 0.

Reproduce results in the paper

Python: 3.8

1.Install the requirements.

$ pip install -r requirements.txt
$ cd evaluation

The file tree of "./evaluation" is as below:

evaluation
├── plot # generate tables and figures in the paper
├── picture
├── result # logs used in the paper
├── result_cmp # logs generated by you
├── *.bat # shells for generating logs
├── ...
├── *.py
└── sh_bat_convert.py # a tool for converting shell types

2.Repeat the results in the paper.

We provide the logs used to generate the results in the paper and you can repeat the results without re-running the programs.

Run the python files in "./plot", such as

$ cd plot
$ python Fig-2_plot.py

For table VI, the raw log is an excel and you should read it directly.

3.Generate the logs by yourself.

We provide shells to generate logs, you can run these shells and check if their output is similar or the same as the logs we provide.

For simplify, you can run the shell 0_generate_ref_log.bat as

# windows
$ call 0_generate_ref_log.bat
# linux
$ sh 0_generate_ref_log.sh

Also, you can generate results as you want, here is an example

# convert shell types
$ python sh_bat_convert.py
# windows
$ call 1_Fig-2.bat
# linux
$ sh 1_Fig-2.sh
# The output is similar to "./result/Fig-2_log"

In the file directory ./result_cmp there are logs that you have generated, which correspond exactly to those in ./result.

Some tips:

  • To save your testing time, we reduce our data set, which may result in results that are not exactly the same as the original protocol, but are acceptable. (important)

    • The 7_Fig-9* && 8_Fig-10* shells are very time consuming, you can use the shells we provide to check if the format of the data is correct or not. However, the results may be distorted. You can replace --reduction 1 with --reduction 0 to get results similar to those in the paper and choose --reduction 1 to save your test time. Here we provide 8_Fig-10a_partdata.bat (whose log is Fig-10a_log) as an example.

  • There are some special shells, explained below.

    • 5_Tab-4.bat : The filename in ./result/Tab-4_log corresponds to the source_device parameter in the shell. esp32_* is the result if source_device==esp32, f429_* corresponds to stm32 and f103_* corresponds to stm32f103.
    • 10_Fig-11.bat: The filename in ./result/Fig-11_log is formatted as result_{out_len}_{test_type}.
    • 8_Fig-10a_partdata.bat : shells for generating Fig-10a_log using part of the results (which can reduce the testing time but impact the results).
    • 2_Tab-3_*_ensemble : display the results of the last two rows of Table III.

  • If you want to use the full data set for testing, there are parameters we have used.

    1_Fig-2: 
	--repeat_time 100
	
2_Tab-3_*_*:
	--reduction 0
	--repeat_time 400

4_Fig-7*:
	--repeat_time 400

5_Tab-4:
	--repeat_time 400

6_Fig-8:
	--reduction 0
	--repeat_time 400

7_Fig-9* && 8_Fig-10*:
	--reduction 0
	--device_number 18
	--test_time 1000

9_Tab-5:
	--reduction 0

10_Fig-11:
	--test_type 1 # 1/3/12/123
	--out_len 5000 # 1000/5000/15000/20000