**11. The ICMetric security system**

Current defensive mechanisms are not enough for preventing the internal attacks in device, since they require ICMetric security system as protection system to increase their security. ICMetric technology depends on measureable features, which have been achieved from the properties of a particular embedded system. Features are generated in the particular system that represents a unique feature for that system. The focus is on utilising a magnetometer, gyroscope, and accelerometer sensors that are provided in the new system.

In this chapter, an intelligent wheelchair is required where the bias readings extracted from sensors embedded in intelligent wheelchair are utilised in the ICMetric security system generation. These bias readings were employed to create an ICMetric basis that was used as identifications for device. In this chapter, ICMetric security system based on bias readings is extracted from gyroscope, accelerometer, and magnetometer sensors. The proposed algorithm is summarised in the following steps:


## **12. Sensors bias measurement**

In some cases, it is not feasible to collect the bias in the sensor. Sensors are required for generating the bias readings of a system which do not require user intervention and are not influenced by external factors. MEMS is a technology that combine mechanical and electrical components. There are many MEMS-based sensors that are being embedded into recent vehicles, wearable devices, laptops, and smartphones. The most commonly used and good examples of MEMS sensors are the accelerometer, the magnetometer, and the gyroscope. Accelerometers are intended to measure the acceleration of an object, whilst gyroscopes measure angular velocity. Magnetometer sensors are used for measuring and detecting magnetic fields.

MEMS sensors are used in this chapter because they are readily available and also the required stimulus is easy to create. Various numbers of bias reading are obtained from MEMS sensors to generate ICMetric number. The system needs to determine the optimal number of readings which are used in identification processes to control the stability of the statistical processes of the ICMetric generation. We need to calculate the population mean and compare the result with the mean value calculated for a smaller subset of readings to determine the best number of readings.

In order to extract reading from sensors, MEMS sensors are placed on the board, simulation is applied to produce a constant bias. The case is similar when a sensor is under operation where accuracy of sensor output is affected by unclear damages due to mistreating. For bias generation, the stimulus must be specified. One of the advantages of stimulus is that it does not need a specific device to evaluate it but is equipped by the user. The readings generated by the sensor must be equal to the stimulus applied to the sensor. Every axis owns a different bias, which is showed in the readings. Experiments prove that the bias in every sensor is unique and reproducible. These bias readings are utilised to provide identification of device and improve security.

#### **13. Statistical analysis for ICMetrics**

ICMetric is not stored but is created when needed this distinguishes the technology of ICMetric in protection from attacks. Since it is created when needed, it requires simple mathematical processes and a statistical analysis for the values of features. Below are some statistical analyses required for the generation process of the ICMetric number is utilising to apply identification of device.

If we assume that *X*¯ represents mean, *x* represents particular sample reading from accelerometer, magnetometer, and gyroscope, and *n* is total number of reading, then [7]:

$$\begin{array}{rcl}\overline{X}^{\overline{n}} & = & \frac{1}{\overline{n}}\sum\_{i=1}^{n} \mathcal{X}\_{i} \end{array} \tag{1}$$

In order to complete ICMetric generation process, we need to calculate σ<sup>2</sup> as explained below, where σ<sup>2</sup> is the standard deviation.

$$\sigma^2 = \begin{array}{c} \sum\_{i=1}^n p(\boldsymbol{\omega}\_i) \left( \boldsymbol{\omega}\_i - \overline{\boldsymbol{X}} \right)^2 \end{array} \tag{2}$$

Furthermore, other statistical and mathematical functions are utilised to analyse the generated reading. For example, we need to calculate variance (*s* 2) as explained in the following equation:

$$\left(\mathbf{s}^{2}\right)^{2} = \frac{1}{n-1}\sum\_{i=1}^{n} \left(\mathbf{x} - \overline{X}\right)^{2} \tag{3}$$

**103**

[5]: Let

*Embedded Devices Security Based on ICMetric Technology*

*CI* <sup>=</sup>¯

distribution of bias readings. It can be negative or positive:

is a measure of dispersion for extracting readings.

The skewness distribution (*S*) is a measure of asymmetry of the probability

*S* = \_ 3 (*X*¯<sup>−</sup> *<sup>m</sup>*) *s* 2

To prove the uniqueness of the bias generated from accelerometer, magnetometer, and gyroscope sensors, we use 95% confidence interval. If *X*¯ is the mean, σ is the standard deviation, and *n* is the total number of observations, then the confidence interval *CI* is given in the Eq. (5) where the numeric value *v* here equals to 1.96.

*X* ± *v*

In addition, other statistical and mathematical functions are utilised to analyse the generated reading such as inter quartile range (IQR) that represents difference between the third and the first quartile in offset data. IQR can be calculated according to the Eq. (6), where *Q*3 represents upper quartile and *Q*1 represents lower

*IQR* = *Q*3 − *Q*1 (6)

ICMetric security system has been evaluated using a number of ways based on trace file evaluation. The ICMetric security system can be generally evaluated from

1.Accuracy: this portion also termed effectiveness classification characterises the ability of the system to separate between intrusive and non-intrusive

2.Efficiency: this portion deals with the resources required to be allocated to the

The identification rate and four alarms are utilised as performance metrics to test the system. To measure and evaluate the system performance, four types of alarms are needed to calculate: true positive (TP), false positive (FP), true negative (TN), and false negative (FN). The measures will be calculated as follows

TN = attack connection record classified as attack

FP = normal connection record classified as attack

*T*P = normal connection record classified as normal

System features can be evaluated in terms of performance, correctness, and usability. Researchers used metrics to assess the performance of the system. Many performance measures are used to evaluate the system, which are based on the

system including CPU cycles and main memory

dataset extracted from the trace file created by ns-2.

\_σ √ \_

(4)

*<sup>n</sup>* (5)

*DOI: http://dx.doi.org/10.5772/intechopen.89240*

where *s* 2

quartile.

two views [8]:

activities.

**14. Performance metrics**

*Embedded Devices Security Based on ICMetric Technology DOI: http://dx.doi.org/10.5772/intechopen.89240*

*Service Robotics*

**12. Sensors bias measurement**

In some cases, it is not feasible to collect the bias in the sensor. Sensors are required for generating the bias readings of a system which do not require user intervention and are not influenced by external factors. MEMS is a technology that combine mechanical and electrical components. There are many MEMS-based sensors that are being embedded into recent vehicles, wearable devices, laptops, and smartphones. The most commonly used and good examples of MEMS sensors are the accelerometer, the magnetometer, and the gyroscope. Accelerometers are intended to measure the acceleration of an object, whilst gyroscopes measure angular velocity.

Magnetometer sensors are used for measuring and detecting magnetic fields.

for a smaller subset of readings to determine the best number of readings.

utilised to provide identification of device and improve security.

the ICMetric number is utilising to apply identification of device.

**13. Statistical analysis for ICMetrics**

MEMS sensors are used in this chapter because they are readily available and also the required stimulus is easy to create. Various numbers of bias reading are obtained from MEMS sensors to generate ICMetric number. The system needs to determine the optimal number of readings which are used in identification processes to control the stability of the statistical processes of the ICMetric generation. We need to calculate the population mean and compare the result with the mean value calculated

In order to extract reading from sensors, MEMS sensors are placed on the board, simulation is applied to produce a constant bias. The case is similar when a sensor is under operation where accuracy of sensor output is affected by unclear damages due to mistreating. For bias generation, the stimulus must be specified. One of the advantages of stimulus is that it does not need a specific device to evaluate it but is equipped by the user. The readings generated by the sensor must be equal to the stimulus applied to the sensor. Every axis owns a different bias, which is showed in the readings. Experiments prove that the bias in every sensor is unique and reproducible. These bias readings are

ICMetric is not stored but is created when needed this distinguishes the technology of ICMetric in protection from attacks. Since it is created when needed, it requires simple mathematical processes and a statistical analysis for the values of features. Below are some statistical analyses required for the generation process of

If we assume that *X*¯ represents mean, *x* represents particular sample reading from accelerometer, magnetometer, and gyroscope, and *n* is total number of read-

*xi* (1)

*X*)<sup>2</sup> (2)

*X*)2 (3)

as

2) as explained

¯ *<sup>X</sup>* = \_1 *<sup>n</sup>* ∑ *i*=1 *n*

σ<sup>2</sup> = ∑*i*=1

the generated reading. For example, we need to calculate variance (*s*

<sup>2</sup> = \_1 *n* − 1 ∑ *i*=1 *n* (*<sup>x</sup>* <sup>−</sup>¯

*s*

In order to complete ICMetric generation process, we need to calculate σ<sup>2</sup>

*<sup>n</sup> <sup>p</sup>*(*xi*) (*xi* <sup>−</sup>¯

Furthermore, other statistical and mathematical functions are utilised to analyse

is the standard deviation.

**102**

ing, then [7]:

explained below, where σ<sup>2</sup>

in the following equation:

where *s* 2 is a measure of dispersion for extracting readings.

The skewness distribution (*S*) is a measure of asymmetry of the probability distribution of bias readings. It can be negative or positive: *S* = \_

$$\mathbf{S}^{\top} = \frac{\mathbf{3} \left( \overline{\mathbf{X}} - m \right)}{s^2} \tag{4}$$

To prove the uniqueness of the bias generated from accelerometer, magnetometer, and gyroscope sensors, we use 95% confidence interval. If *X*¯ is the mean, σ is the standard deviation, and *n* is the total number of observations, then the confidence interval *CI* is given in the Eq. (5) where the numeric value *v* here equals to 1.96.

$$\text{CI} \quad = \begin{array}{c} \overline{X} \not\Rightarrow \upsilon \frac{\sigma}{\sqrt{n}} \end{array} \tag{5}$$

In addition, other statistical and mathematical functions are utilised to analyse the generated reading such as inter quartile range (IQR) that represents difference between the third and the first quartile in offset data. IQR can be calculated according to the Eq. (6), where *Q*3 represents upper quartile and *Q*1 represents lower quartile.

$$\text{IQR} \quad = \text{Q\\$} - \text{Q1} \tag{6}$$

#### **14. Performance metrics**

ICMetric security system has been evaluated using a number of ways based on trace file evaluation. The ICMetric security system can be generally evaluated from two views [8]:


System features can be evaluated in terms of performance, correctness, and usability. Researchers used metrics to assess the performance of the system. Many performance measures are used to evaluate the system, which are based on the dataset extracted from the trace file created by ns-2.

The identification rate and four alarms are utilised as performance metrics to test the system. To measure and evaluate the system performance, four types of alarms are needed to calculate: true positive (TP), false positive (FP), true negative (TN), and false negative (FN). The measures will be calculated as follows [5]: Let

*T*P = normal connection record classified as normal

TN = attack connection record classified as attack

FP = normal connection record classified as attack

FN = attack connection record classified as normal

Then,

*TPRate*(sensitivity) = \_*TP TP* <sup>+</sup> *FN* (7)

$$\begin{array}{rcl} \text{TP}\_{Rate\text{(sensitivity)}} & = & \frac{\text{TP}}{\text{TP} + \text{FN}} & & \text{(7)} \\\\ \text{TN}\_{Rate\text{(specificity)}} & = & \frac{\text{TN}}{\text{TN} + \text{FP}} & & \text{(8)} \\\\ \text{FN}\_{Rate\text{(1-sensivity)}} & = & \frac{\text{FN}}{\text{FN} + \text{TP}} & & \end{array} \tag{9}$$

$$
\begin{array}{lcl}
F \text{N}\_{\text{Rate}\{1-\text{s}\,\text{sensitivity}\}} & = & \frac{F \text{N}}{F \text{N} + TP} & & \\ \\ \\ \text{} & & & \\ \\ \text{} & & & \\ \text{} & & & \\ \text{} & & & \\ \text{} & & & \\ \text{} & & & \\ \text{} & & & \\ \text{} & & & \\ \text{} & & & \\ \end{array}
\tag{9}
$$

$$\left(FP\_{Rate\{1\text{-}species\}}\right) = \frac{FP}{FP \text{+ TN}}\tag{10}$$

In addition, some extra metrics are utilised to evaluate system performance such as packet delivery rate (PDR), throughput, and end-to-end delay.

• Packet Delivery Ratio (PDR): the ratio between the number of packets sent from the origin and the proportion of packets received at the destination.

$$\text{PDR} = \sum N\_r / \Sigma N\_r \tag{11}$$

where *Nr* = number of packets received and *Ns* = number of packets sent.

• Throughput: the total number of packets that are transferred in the system. Throughput of a system can be presented as shown in the following equation.

$$\text{Rate of throughput(kbps)} = N\_r \ast \text{S/ST} \,\text{(12)}\tag{12}$$

where *Nr* = number of packets received and *S* = packet size and *ST* = simulation time.

• Average end-to-end delay: the average time for packets reaching from the origin to the destination. Average end-to-end delay is explained in the following equation:

$$\text{Rate end - to - end delay (ms) } = \left(\frac{\sum end\_{time} - \text{start}\_{time}}{\sum N}\right) \tag{13}$$

**105**

rate.

**16. Future work**

*Embedded Devices Security Based on ICMetric Technology*

device is unique in its internal environment.

generation of a device identification.

under certain conditions.

throughput, end-to-end delay.

sensor where every sensor will have unique readings.

1.ICMetric technology can be used to apply identification and improve security

2.ICMetric technology relies on the special internal features of device where each

3.MEMS gyroscope, magnetometer, and accelerometer sensors embedded in intelligent wheelchair are utilised. Three axes readings achieved from every

4.Readings generated from MEMS sensors are analysed statistically to generate ICMetric number utilised for device identification. A statistical study of the readings generated from sensor shows practical use of MEMS sensors for the

5.Support vector machine exploited in this chapter to evaluate and test system

6.In order to evaluate system performance, we need to calculate the performance

metrics, which are accuracy rate, error rate, and four types of alarms.

7.Additional performance metrics can be evaluated for system such as, PDR,

During the training and testing processes, the SVM with dataset is extracted from the trace file generated by ns-2; it was found that using ICMetric technology for embedded devices identification provides better rate of accuracy and low error

1.This chapter has established the design of an ICMetric by using different

It can be integrated into bitcoins to provide secrecy of transactions.

features which can strengthen the device ICMetric.

features of device. Therefore, the aim of the future research is to discover more

2.The ICMetric technology has not been researched in bitcoins and block chains.

*DOI: http://dx.doi.org/10.5772/intechopen.89240*

of embedded devices.

where *N* represents a number of connections.

#### **15. Conclusions**

The main contribution of the present chapter is to achieve an identification model for intelligent wheelchair application with high rate of accuracy and with low error rate. This was done through the design of an identification process by using ICMetric technology. From the given results, the following substantial remarks were obtained:

*Service Robotics*

Then,

equation.

equation:

**15. Conclusions**

time.

FN = attack connection record classified as normal

*TPRate*(sensitivity) = \_*TP*

*TNRate*(specificity) = \_*TN*

*FNRate*(1−*s*ensitivity) = \_*FN*

*FPRate*(1−specificity) = \_*FP*

In addition, some extra metrics are utilised to evaluate system performance such

• Packet Delivery Ratio (PDR): the ratio between the number of packets sent from

system. Throughput of a system can be presented as shown in the following

where *Nr* = number of packets received and *S* = packet size and *ST* = simulation

• Average end-to-end delay: the average time for packets reaching from the origin to the destination. Average end-to-end delay is explained in the following

The main contribution of the present chapter is to achieve an identification model for intelligent wheelchair application with high rate of accuracy and with low error rate. This was done through the design of an identification process by using ICMetric technology. From the given results, the following substantial remarks were obtained:

Rate end − to − end delay (ms) = (

where *N* represents a number of connections.

Rate of throughput(kbps) = *Nr* ∗ *S*/*ST* (12) (12)

the origin and the proportion of packets received at the destination.

• Throughput: the total number of packets that are transferred in the

as packet delivery rate (PDR), throughput, and end-to-end delay.

where *Nr* = number of packets received and *Ns*

*TP* <sup>+</sup> *FN* (7)

*TN* <sup>+</sup> *FP* (8)

*FN* <sup>+</sup> *TP* (9)

*FP* <sup>+</sup> *TN* (10)

PDR = ∑ *Nr*/∑ *Ns* (11)

= number of packets sent.

<sup>∑</sup>*endtime* <sup>−</sup>*starttime* \_\_\_\_\_\_\_\_\_\_\_\_\_\_ ∑*<sup>N</sup>* ) (13)

**104**


During the training and testing processes, the SVM with dataset is extracted from the trace file generated by ns-2; it was found that using ICMetric technology for embedded devices identification provides better rate of accuracy and low error rate.
