|
Work item:
|
X.sreai-ec
|
|
Subject/title:
|
Security requirements of delivering edge AI on edge computing
|
|
Status:
|
Under study
|
|
Approval process:
|
TAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
-
|
|
Equivalent number:
|
-
|
|
Timing:
|
2027-09 (Medium priority)
|
|
Liaison:
|
SG13
|
|
Supporting members:
|
China Telecom, China Unicom, Ministry of Industry and Information Technology (MIIT)(China), DSIT - United Kingdom
|
|
Summary:
|
Artificial intelligence (AI) is essential for enhancing natural language processing, automating complex tasks, and providing intelligent insights across various applications. Edge AI involves deploying and operating AI capabilities on edge computing devices, bringing AI processing closer to data sources like smartphones and IoT devices. This architecture enables real-time data analysis and decision-making.
However, the rapid growth of edge AI introduces new security threats and risks, particularly targeting the management and services of AI capabilities on edge computing devices, which often have limited computing power. These threats attack data integrity, model reliability, AI service availability, and secure communication channels. Compromised data integrity can lead to manipulated outputs, while attacks on model reliability can distort decision-making. Threats to AI service availability can cause significant disruptions, rendering AI capabilities inaccessible when needed most. Vulnerabilities in communication channels can expose sensitive information to interception and tampering. These limitations make it essential to implement lightweight, yet robust security controls that can protect AI assets, sensitive data, and communications. These security challenges can severely impact the functionality and stability of AI on edge devices, leading to service disruptions and increased system vulnerabilities.
Hence, establishing a standard specifically for edge AI security within edge computing environments is crucial. This standard would offer clear, actionable measures tailored to the unique risks and requirements of edge AI, ensuring consistent and robust security practices for all stakeholders. This would enhance the overall security of edge AI, fostering greater trust and adoption across various industries. By addressing these specific security challenges, the standard would play a vital role in protecting sensitive data, ensuring secure model management, and maintaining the reliability of edge AI services.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2024-09-12 17:34:17
|
|
Last update:
2024-09-16 14:25:17
|
|
