MOSID (Microcontroller On-chip Sensor IDentification): A dataset of readings from the internal monitoring sensors of STM32L152RTXX microcontrollers during the stimulation of their electronic activity

The MOSID (Microcontroller On-chip Sensor IDentification) dataset consists of 5 acquired data subsets (6,72 GB total, compressed into 560 MB), each collected during different experiments and periods using various equipment (HMP4040, DF1731SB & HM305) and acquisition strategies. These subsets contain readings from the temperature and voltage sensors embedded in 20 STM32L-DISCOVERY devices. The data was captured during the execution of 5 different workloads as stimuli, repeated over 20 iterations. The stimuli employed are as follows:

1. 20x20 Long-type matrix product.
2. 20x20 Float-type matrix product.
3. Algorithm for ascending sorting, Bubble Sort.
4. Algorithm for 2D-point clustering, Convex Hull.
5. Encryption algorithm AES 128-bit.

The subsets are structured according to the folder format "X_Y," where X is the manually assigned number to the board, and Y is the corresponding number for the executed algorithm. Within each of these folders, files are present in the format "data_Z.txt," where Z represents the iteration number to which the file belongs. In total, the dataset comprises 9600 files with a final size of approximately 7 GB. The different presented subsets are as follows:

• ACQ1: Derived from the experiment named "Automatic Acquisition 1 (HMP4040)" conducted using a daisy-chain topology (20 out of 20 boards, 2000 files).
• ACQ2: Derived from the experiment named "Automatic Acquisition 2 (HMP4040)" conducted using a daisy-chain topology (20 out of 20 boards, 2000 files).
• ACQ3: Derived from the experiment named "Individual Acquisitions (HMP4040)", performed board by board from idle conditions (20 out of 20 boards, 2000 files).
• ACQ4: Derived from the experiment named "GOLD SOURCE DF1731SB Acquisitions" conducted using a partial daisy-chain setup (2 devices at a time, 18 out of 20 boards excluding boards , 1800 files).
• ACQ5: Derived from the experiment named "HANMATEK HM305 Acquisitions" conducted using a partial daisy-chain setup (2 devices at a time, 18 out of 20 boards, 1800 files).

In each "data_Z.txt" file, starting from the 5th line, temperature and voltage raw ADC conversions from the sensors are provided, captured during the execution of the stimulus in successive lines. Additionally, a table (Table_UIDS.csv) with metadata for each of the boards used in the experiments is included, which is needed in order to normalize the data in terms of ºC and Volts.

• BOARD_NUM, which contains the manually assigned board number.
• UID, which contains the Unique Identifier of the board assigned by the manufacturer.
• T_CAL_1, which holds the calibration value of the board's temperature sensor at 30ºC.
• T_CAL_2, which holds the calibration value of the board's temperature sensor at 100ºC.
• VREFINT_CAL, which contains the calibration value of the board's voltage sensor.