Record and Replay Touchscreen Events on Android

How to record and replay touchscreen events on an Android device.

Table of contents

• Prerequisites
• Quick start
• Easy method
• Step-by-step method and Explanation
• Outputting Data
• Authors

Prerequisites

You need to:

• Install ADB in your computer (sudo apt install adb)
• Enable USB Debugging in your Android device (Settings > Developer Options > USB Debugging)

Now, connect your device to your computer, via a USB cable, and run the following bash script in a terminal:

./find_touchscreen_name.sh

if you get the output message "Touchscreen device found!", then you can use the Easy method. Otherwise, you need to use the Step-by-step method.

Optional:

• It is easier for you to visualize where you touched the screen by turning one of the following (1) Show touches or (2) Pointer location (In your phone go to Settings > Developer Options > Input > Show touches)

Quick start

To start using the mAP you need to clone the repo:

git clone https://github.com/Cartucho/android-touch-record-replay

Easy method

1. Record touch events with timestamps:

   ./record_touch_events.sh

2. Replay those touch event:

   ./replay_touch_events.sh

Step-by-step method and Explanation

Android uses the Linux kernel so it processes input events the same way as any other Linux system. With the android's getevent tool we can record the live dump of kernel input events and use that recording to replay the same touch events.

1. Connect your phone via a USB cable to your computer.
2. Find out the name of the touchscreen's device (should look something like /dev/input/event<your_index>). There a couple of ways of doing this:
   • Use the command adb shell getevent -lp and figure out which of the input devices contains events with the ABS_MT_TOUCH_MAJOR textual label;
   • Another option is to try the command adb shell getevent | grep event<your_index> with the different possible values for <your_index> until you see that the terminal is printing events exactly when you move your finger over the phone's touchscreen.

In my case, my touchscreen's name is /dev/input/event7 which I will be using from now forward. Don't forget to change this in the following step's commands!

3. Using the touchscreen's device name we can now record events using the command adb shell getevent -t /dev/input/event7 > recorded_touch_events.txt. Using the -t option we also record the timestamps (so that later we can replay the recorded input with the proper delay between events). The content on this file should look something like:

    ...
    [   53890.813990] 0000 0000 00000000
    [   53890.828065] 0003 0035 000001be
    [   53890.828065] 0003 0036 00000258
    [   53890.828065] 0003 0030 0000001f
    [   53890.828065] 0003 0032 0000001f
    [   53890.828065] 0003 0039 00000000
    ...
   

   , where each line represents an event's [ timestamp] type code value.

4. We now push those recorded events into the phone using the command adb push recorded_touch_events.txt /sdcard/

5. In order to replay those events we will use the mysendevent executable, which will read the recorded_touch_events.txt file line by line and inject those events in the kernel input. To do this we need to push this executable to the phone by using the following command adb push mysendevent /data/local/tmp/

   Note: In case you want to modify the file mysendevent.c you have to recompile it (for android, e.g., use arm-linux-gnueabi-gcc -static -march=armv7-a mysendevent.c -o mysendevent) and re-push mysendevent to the phone.

6. Finally, we can replay the previous recording by running the command adb shell /data/local/tmp/mysendevent /dev/input/event7 /sdcard/recorded_touch_events.txt

Outputting Touch Data

You may want to export and anyalse the touch screen data.

As mentioned in the previous section the touch data is in the following format

Timestamp	Type	Code	Value
[ 53890.813990]	0000	0000	000005b0
[ 53890.828065]	0003	0035	000001be
[ 53890.828065]	0003	0036	00000258
[ 53890.828065]	0003	0030	0000001f
[ 53890.828065]	0003	0032	0000001f
[ 53890.828065]	0003	0039	00000000

Timestamp

According to the android documentation the timestamp data is in the following format:

getevent timestamps use the format $SECONDS.$MICROSECONDS in the CLOCK_MONOTONIC timebase.

CLOCK_REALTIME vs. CLOCK_MONOTONIC

This dataformat is not ideal as it is relative to an arbitrary time in the system.

We will not use this data in our data collection script.

Type & Code

Currently we have output these as their codes, we can use the -l argument to output these codes as descriptive labels.

For example:

adb shell getevent -lt /dev/input/event7 > recorded_touch_events.txt

would now output:

Timestamp	Type	Code	Value
[ 46112.124651]	EV_ABS	ABS_MT_TRACKING_ID	000005b0
[ 46112.124651]	EV_ABS	ABS_MT_POSITION_X	0000010f
[ 46112.124651]	EV_ABS	ABS_MT_POSITION_Y	00000479
[ 46112.124651]	EV_ABS	ABS_MT_PRESSURE	0000005f
[ 46112.124651]	EV_ABS	ABS_MT_TOUCH_MAJOR	0000005f
[ 46112.124651]	EV_SYN	SYN_REPORT	00000000
[ 46112.173948]	EV_SYN	0004	0000b420
[ 46112.173948]	EV_SYN	0005	0a5db877

Value

The value is currently being output as hexadecimal. We will convert the value into an integer when we collect the data in our script later.

The following command will display the min and max values for each input:

adb shell getevent -lp /dev/input/event7

  name:     "Melfas MMSxxx Touchscreen"
  events:
    ABS (0003): ABS_MT_SLOT           : value 0, min 0, max 9, fuzz 0, flat 0, resolution 0
                ABS_MT_TOUCH_MAJOR    : value 0, min 0, max 30, fuzz 0, flat 0, resolution 0
                ABS_MT_POSITION_X     : value 0, min 0, max 720, fuzz 0, flat 0, resolution 0
                ABS_MT_POSITION_Y     : value 0, min 0, max 1280, fuzz 0, flat 0, resolution 0
                ABS_MT_TRACKING_ID    : value 0, min 0, max 65535, fuzz 0, flat 0, resolution 0
                ABS_MT_PRESSURE       : value 0, min 0, max 255, fuzz 0, flat 0, resolution 0
  input props:
    INPUT_PROP_DIRECT

Output data to CSV

The human_readable_data.py script will convert all of the touch data into a more readable format and output it to a csv file.

Usage: python human_readable_data.py arg1 arg2

Argument	Value	Example
arg1	touchscreen device	/dev/input/event7
arg2	csv output location	/Users/$user/Desktop/touch_info.csv

This will create a csv file containing the following information:

Timestamp	Type	Code	Value
2019-02-13 10:30:41.558271	EV_ABS	ABS_MT_TRACKING_ID	1804
2019-02-13 10:30:41.558557	EV_ABS	ABS_MT_POSITION_X	805
2019-02-13 10:30:41.558797	EV_ABS	ABS_MT_POSITION_Y	848
2019-02-13 10:30:41.559035	EV_ABS	ABS_MT_PRESSURE	51
2019-02-13 10:30:41.559245	EV_ABS	ABS_MT_TOUCH_MAJOR	51
2019-02-13 10:30:41.559453	EV_SYN	SYN_REPORT	0
2019-02-13 10:30:41.697711	EV_SYN	4	66179
2019-02-13 10:30:41.701748	EV_SYN	5	233329954

Authors:

• João Cartucho

  Feel free to contribute