Protected Park: A Cornerstone of Trust in the Smart City

The Internet of Things (IoT) enables the integration of physical devices into the digital realm by granting them Internet connectivity and analytical capabilities. When IoT is combined with other domains can make an efficient system that can help humans. The Internet of Things (IoT) is revolutionizing [1] daily life by connecting smart devices to the internet. While promising, IoT still faces critical challenges. This paper reviews its key issues, architecture, and applications, offering insights into its real-world potential. In the past two decades, IoT devices have become integral across industries, yet their security remains a major challenge due to vulnerabilities [2] that cyberattacks exploit, causing widespread damage. This work examines the IoT security landscape, identifying key challenges, security goals, common threats, and necessary countermeasures for securing IoT systems. One such system is proposed in this article that creates a fearless and secure park as part of a smart city. A smart city uses technology and innovation to improve efficiency, sustainability, and inclusivity, tackling issues like pollution, population growth, and healthcare. IoT-based sensor networks, enhanced by machine learning, offer promising solutions in healthcare. Ghazal et al. [3] explores their application and lays the groundwork for future research on IoT’s impact in smart cities. Smart cities have advanced rapidly, driven by IoT innovations that enable new services and applications. Bellini et al. [4] in their work reviews the research on IoT-enabled smart cities, highlighting key trends and challenges in building sustainable, efficient urban environments. It also surveys IoT technologies and classifies smart city approaches into eight domains, expanding on the traditional six-domain model.

Incorporating hardware components and deploying requisite software constitutes integral facets of the system design workflow. The ensuing section delineates the system design prerequisites essential for the envisaged project.

3.1 Hardware setup

The project mandates the installation of the official Raspberry Pi Operating System onto a microSD card. Subsequently, an apt Raspberry Pi web camera must be chosen, followed by the establishment of a connection between the Raspberry Pi 3 B+ and the selected web camera through a micro USB cable, interfacing with a designated USB port [18].

1.png

Figure 1. Overall system view

The project is delineated by a sequence of five steps, as depicted in Figure 1.

The steps include:

1.  Acquisition of data through a webcam.
2. Generation of a digital signature utilizing the ECDSA algorithm.
3. Activation of an alarm system in the event of sharp object detection.
4. Storage of data in either a local drive or a cloud environment, contingent upon the client's specifications.
5. Retrieval of data and subsequent verification.

The initial two steps are executed on a Raspberry Pi platform, while the third step involves data storage. The fourth step is carried out on end devices such as mobile devices, laptops, or desktops. This paper proposes utilizing the Raspberry Pi Model 3B V1.3 along with a Raspberry Pi-compatible USB webcam for image capture. The USB webcam offers greater reliability compared to the Pi Camera, which uses a serial connection and is more fragile. A 64GB microSD card was employed to ensure faster OS performance. For end-user interaction, devices such as laptops, desktops, tablets, or smartphones can be used. The detailed configuration specifications for each device can be found in Table 2.

Table 2. Hardware components and configurations

3.2 Web camera configuration and connection

For the execution of this project, the utilization of a USB-connected web camera is recommended, as it exhibits superior performance in capturing both video streams and still images. The inherent advantage of USB-connected cameras lies in their seamless plug-and-play functionality, ensuring straightforward integration into the system. The Raspberry Pi device is appropriately configured to accommodate the requirements of the project, leveraging the fswebcam software package. This software facilitates the capture of both high-quality photos and video frames. The operational command employed for capturing images through the fswebcam software is exemplified below:

fswebcam - r 1280x720--no-banner/smart_park

/image1.jpg

The acquired image of the smart park sample, obtained via a webcam, is illustrated in Figure 2.

2.png

Figure 2. Smart park

In the project workflow, when timestamping is required for images, ImageMagick is utilized to embed temporal metadata, including both time and date, onto the images. This preprocessing step precedes the archival of the images in the cloud storage system, ensuring temporal context is preserved for future retrieval and reference [19]. In this endeavor, the protection of image data captured from the webcam is executed by leveraging the MIRACL (Multi-precision Integer Rational Arithmetic Crypto Library) cryptographic library package. This entails employing advanced cryptographic algorithms and protocols provided by MIRACL to ensure the confidentiality, integrity, and authenticity of the acquired image data. The library facilitates multi- precision arithmetic operations, enabling robust cryptographic computations necessary for securing sensitive visual information. The utilization of MIRACL enhances the overall security posture of the system, fortifying against potential threats and unauthorized access to the webcam-derived image data [20].

3.3 ECDSA algorithm

Digital signatures serve as a means of data authentication, with the utilization of elliptic curve cryptography representing an advanced technique in this domain. Elliptic Curve Cryptography (ECC) stands out as one of the most efficient cryptographic methodologies, demanding lower processing overhead. Within ECC, the Elliptic Curve Digital Signature Algorithm (ECDSA) [21, 22] is a prominent mathematical algorithm employed to validate the integrity and authenticity of digital information. Public-key cryptography, exemplified by ECDSA, operates on key pairs, consisting of public keys distributed openly and private keys maintained in secrecy, enhancing the overall security of cryptographic processes.

Point arithmetic procedures like point doubling and point addition are used to implement ECC protocols (ECDSA). The ECC provided by is a curve equation $y^2=x^3+A x+b \bmod p$. This equation uses variables like [15A, B]. It is provided by NIST P192 elliptical curves with prime field. Below is information on the variables used in the equation (NIST P192) as described in Figure 3.

3.png

Figure 3. NIST P192 curve parameters

The elliptical curve digital signature algorithm consists of three phases.

• Phase 1: Creation of keys
• Phase 2: File signatures (R, S)
• Phase 3: Validation of the signed document

Phase 1: Creation of keys:

Random seed: A random seed serves as an initial input for the initialization of a pseudorandom number generator (PRNG). In the context of computational algorithms, particularly those involving stochastic processes, a random seed is a numeric value that kickstarts the deterministic sequence of seemingly random numbers generated by the PRNG. The selection of an appropriate random seed is crucial for ensuring reproducibility in computational experiments and simulations, as the same seed will consistently produce the same sequence of pseudorandom numbers, enabling researchers to replicate and verify their results. During the random seed generation phase, temporal information, notably the precise time and date of the image, is incorporated.

ECDSA generates the key using the basis point G and a significant prime number (d_sp) as described in Eq. (1).

$Q_{s p}=d_{s p} \times G$    (1)

• “d_sp” - private key which is derived from the random seed.
• “Q_sp” - public key that is shared in addition to the signatures.
• “G” is NIST P192’s definition of the base point.

This random seed is utilized as a crucial input parameter in the key generation module, which resides within most of the cryptographic library. This module is subject to modifications to enable the generation of both public and private keys. Following this cryptographic process, the resultant keys are stored as public key and private key value, representing the public and private aspects of the cryptographic key pair, respectively.

Phase 2: File signatures (R, S):

A cryptographic process generates a 40-byte signature, which is subsequently divided into two 20-byte components, constituting the pair (R, S). This binary pair is then stored in a file and is contemporaneously disseminated alongside the corresponding image. Simultaneously, the indispensable public key file is shared in tandem with the image. The public key file assumes a crucial role in the image verification procedure.

• Finding the value of R

Utilizing a pseudo-random number generator (PRNG) or an initial seed, generate a random 20-byte value denoted as 'k'. Subsequently, employ this generated 'k' value to perform scalar multiplication with the elliptic curve's base point, thereby determining a new point on the curve.

$(x, y)=k \times G$       (2)

'R' will represent that point p's 'x' coordinate value.

• Finding the value of S:

The equation used to find the part of signature value S is described in Eq. (3).

$S=K^{-1}\left(Z+d_{s p} \times R\right) \bmod P$   (3)

where, P represents the finite field in the Ed25519 curve, which is given by 2²⁵⁵-19. K^-1 is the multiplicative inverse of K. The x-coordinate value (R) of point P is determined by employing Eq. (2). The respective R and S values corresponding to Smart Park, as depicted in Figure 2, are illustrated in Figure 4.

4.png

Figure 4. R and S values generated for smart park

• Z Computation:

The SHA-1 algorithm generates a 128-bit hash independent of the size of the input. The SHA-1 function inflates to a 128-bit hash when it receives an input of 100 bits. A 128-bit hash will be produced from the output even if an input of one gigabyte is given.

This basic SHA1 hashing can be represented by the following Eq. (4):

Hash1 = SHA1 (ln)    (4)

In the above equation, more than one input that leads to Hash1 may exist in the domain space. To avoid the hash value collision, this article proposes the use of a 4-byte SALT extra information that is added to SHA1. After that, the equation would be as follows (Eq. (5)):

$H a s h_1=S H A 1(l n)+S A L T$  (5)

Once after generating a SHA1 hash of 128 bits for more collision resistance, this will be given to SHA2 resulting in a 256-bit hash value. The Eq. (6) that would result from passing the 128-bit Hash1 value to SHA-2 is as follows:

$Z=S H A 2\left(H a s h_1\right)$    (6)

There is less chance of a collision in this scenario because Hash1 is smaller than Z.

The SALT is extra information to better protect an image. The function of SALT in the equation is to protect Z's uniqueness from collision assaults. “Z” is a combination of two hash functions of different output sizes. To guarantee uniqueness, this method first compresses the result of the SHA1 procedure using a reduced size before appending the SALT. Finally, the whole equation used to compute Z is as described in Eq. (7).

Hash1 = SHA1 (ln)  （7）

Z stores 256 bits of output.

3.4 Alarm in case of sharp object detection

The alarm system utilizes sophisticated image analysis algorithms to expeditiously alert authorized personnel upon identifying anomalous objects with high acutance within an image dataset. This functionality enhances forensic investigation capabilities in cases of theft, allowing for swift and targeted responses. The park, functioning as a pivotal recreational enclave for children, serves a dual purpose. It not only acts as a deterrent against potential abduction scenarios through the implementation of the advanced alarm system but also plays a vital role in fostering the overall societal well-being.

This study analyzes biometric protocols for securing wireless sensor networks (WSNs), focusing on their characteristics and limitations. It emphasizes the importance of unique biometric authentication and reviews attack defense schemes [23]. The proposed WSN-MAS enhances WSN security and is applicable to smart environments such as healthcare and intelligent transportation.

This paper presents a new automated model for anomaly detection and localization in crowd videos using optimized Self-Organizing Maps (SOM) [24]. The model generates three video frame patterns and employs feature extraction techniques such as HOG, LGP, and PCA for dimensionality reduction.

The dataset for the detection of weapons is chosen from the study [25]. The dataset consists of 4794 images, with 4409 images assigned to the training class and 385 images assigned to the test class. Machine learning models that input or output data sequences are known as sequence models. In this paper, we will use a sequential model. The Sequential model is a simple and convenient way to build a linear stack of layers for a neural network. which can create a Sequential model and add layers to it one by one, specifying the input shape and the number of neurons in each layer.

The Dense layer represents a fully connected layer, where each neuron is connected to every neuron in the previous and next layers. The input shape parameter is specified only for the first layer, as the subsequent layers can automatically infer their input shapes. MaxPooling is a downsampling operation that reduces the spatial dimensions of the input data. Here, the activation functions are 'relu' and 'sigmoid'. Finally, train the model on a dataset using the fit method.

5.png

Figure 5. Model to predict weapons

If the model predicts the 'Gun' or 'Knife' then it signals the alarm in the park and notifies concerned security officers (see Figure 5).

The performance metrics accuracy, precision, and recall, were evaluated across multiple deep learning models. A detailed comparison is provided in Table 3.

Table 3. Performance evaluation of various models using a real dataset

3.5 Storage of image data in the drive

The acquired images, accompanied by their corresponding date and time stamps, are systematically archived in cloud storage or a designated storage drive, facilitating their integration into image detection applications. Typically, these images are encoded in the JPEG (Joint Photographic Experts Group) file format to optimize data compression, achieving a judicious compromise between maintaining satisfactory image quality and minimizing overall file size. Users are afforded the flexibility to tailor parameters such as compression levels during the image storage process.

3.6 Verifying the sign

The preliminary communication phase encompasses the transmission of the primary message, concomitant with the public key (Q_sp) and the signature (R, S). During the verification procedure, the message is employed to derive the hash value, and subsequently, the value p is computed using Eq. (8) in accordance with Eq. (3). The acceptance criterion for the authenticity of the message is satisfied only if both the value R and the computed value are in concordance.

$P=S^{-1} Z \times G+S^{-1} \times R \times Q_{s p}$    （8）

The value of Z is calculated again for the received image at the receiver side and then the value of P is found out using the public key. If a signature is valid then the data is authenticated else the data is modified by some external malicious activity.

6.png

Table 4. The R and S values of the image's public key

Pictures	Public Key	R Value	S Value
4-1.jpg	3318638065693882606455708635923689058643664680677954560261215	118466103783852272334 702306108854626694975 3134157049206621	5492873726759280974 0132350409498796606 46875108678525692463
4-2.jpg	3867505862548321950 4038569147896272473 1383032544319407405 2	233116634303835366947 537254679334578488537 3921764329518959	1954677835897677742 0644016292921013470 8988916225755992428 9
4-3.jpg	5922009533165673920 7847343750542426183 2192057813459147755 8	198358275997014613662 426175066463631849314 7765595072761196	3706882835897790607 0407438501563891236 1237047524540061599 6
4-4.jpg	4941152402187364020 3242568816934471220 8586696520430181884 0	557404485900940816442 006983941534292681910 7218543741978939	1563411350322933790 8401639634476268547 7812162069833661300 9
4-5.jpg	3242838646291286492 5044340178232125733 5565768932777383013 8	458092092585878401048 005995236401701042862 0391780936267915	2073232607295099081 1379851935096572930 8962684113865658956 9