TY - GEN
T1 - Challenges in Implementing Artificial Intelligence on the Raspberry Pi 4, 5 and 5 with AI HAT
AU - Steadman, Phil
AU - Jenkins, Paul
AU - Rathore, Rajkumar Singh
AU - Hewage, Chaminda
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.
PY - 2024/12/20
Y1 - 2024/12/20
N2 - As technology interacts with all areas of daily life, similarly the risks from cybersecurity attacks have increased proportionally. Furthermore, as the demand for computer-controlled devices and infrastructure has increased, and whilst devices have miniaturized, cybersecurity has not kept pace. Therefore, the importance of malware detection for smaller devices is a necessity. Artificial Intelligence (AI) is improving malware detection, however, training AI models traditionally demands powerful computational resources, far more powerful than the capabilities of lightweight devices such as the Raspberry Pi. Researchers have explored alternative methodologies to adapt AI training to these constraints, given the limited performance while maintaining reliability. This paper examines the design and construction of a lightweight secure network infrastructure tailored to the Raspberry Pi's capabilities. Key considerations include network segmentation, firewall implementation, and device configuration management using automation. Networking setups prioritise wired connectivity for low-latency, high-security applications, and implementing security measures such as browser proxy containers and VPN forwarding to mitigate potential threats, particularly in environments prone to malware infiltration. Training challenges on small-board computers reveal the limitations of Raspberry Pi's computational power for AI training. Comparative data identifies the delta between Raspberry Pi’s and conventional computing devices, concluding the need for more powerful platforms for efficient AI training.
AB - As technology interacts with all areas of daily life, similarly the risks from cybersecurity attacks have increased proportionally. Furthermore, as the demand for computer-controlled devices and infrastructure has increased, and whilst devices have miniaturized, cybersecurity has not kept pace. Therefore, the importance of malware detection for smaller devices is a necessity. Artificial Intelligence (AI) is improving malware detection, however, training AI models traditionally demands powerful computational resources, far more powerful than the capabilities of lightweight devices such as the Raspberry Pi. Researchers have explored alternative methodologies to adapt AI training to these constraints, given the limited performance while maintaining reliability. This paper examines the design and construction of a lightweight secure network infrastructure tailored to the Raspberry Pi's capabilities. Key considerations include network segmentation, firewall implementation, and device configuration management using automation. Networking setups prioritise wired connectivity for low-latency, high-security applications, and implementing security measures such as browser proxy containers and VPN forwarding to mitigate potential threats, particularly in environments prone to malware infiltration. Training challenges on small-board computers reveal the limitations of Raspberry Pi's computational power for AI training. Comparative data identifies the delta between Raspberry Pi’s and conventional computing devices, concluding the need for more powerful platforms for efficient AI training.
KW - Artificial Intelligence
KW - Convolutional neural networks. Image processing
KW - Malware
KW - Raspberry Pi
UR - http://www.scopus.com/inward/record.url?scp=85214194606&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-74443-3_8
DO - 10.1007/978-3-031-74443-3_8
M3 - Conference contribution
AN - SCOPUS:85214194606
SN - 9783031744426
T3 - Lecture Notes in Networks and Systems
SP - 147
EP - 157
BT - Contributions Presented at The International Conference on Computing, Communication, Cybersecurity and AI - The C3AI 2024
A2 - Naik, Nitin
A2 - Grace, Paul
A2 - Jenkins, Paul
A2 - Prajapat, Shaligram
PB - Springer Science and Business Media Deutschland GmbH
T2 - International Conference on Computing, Communication, Cybersecurity and AI, C3AI 2024
Y2 - 3 July 2024 through 4 July 2024
ER -