UNIT V
Security, Ethics, and Case Studies in Cloud AI
,Security and privacy in cloud-based AI
• It have become critical as organizations migrate sensitive data to the cloud,
introducing complex, interconnected risks that traditional security measures
struggle to manage.
• The integration of AI into cloud computing offers immense processing power but
requires robust, proactive security frameworks to combat threats like data
poisoning, model stealing, and unauthorized (phishing) access.
,Key Challenges and Security Risks
• Data Breaches and Unauthorized Access: Misconfigurations (the cause of 99% of cloud
breaches) and weak credentials often lead to data breaches, threatening the
confidentiality of sensitive information.
• Model Theft and Poisoning: Malicious actors may steal proprietary AI models (model
stealing) by querying them, or inject corrupted data (data poisoning) into training sets to
corrupt AI reliability.
• Adversarial Attacks: Input manipulation (e.g., modified images) can trick AI into creating
incorrect outputs, which is a critical concern for autonomous vehicles and similar systems.
, Continue…
• Shared Responsibility Misunderstandings: A significant challenge is defining the
boundaries of responsibility between cloud providers (security of the cloud) and
customers (security in the cloud).
• Lack of Visibility: The distributed, multi-tenant nature of cloud environments
makes it difficult for companies to monitor where their data is stored, processed,
and who has access.
Security, Ethics, and Case Studies in Cloud AI
,Security and privacy in cloud-based AI
• It have become critical as organizations migrate sensitive data to the cloud,
introducing complex, interconnected risks that traditional security measures
struggle to manage.
• The integration of AI into cloud computing offers immense processing power but
requires robust, proactive security frameworks to combat threats like data
poisoning, model stealing, and unauthorized (phishing) access.
,Key Challenges and Security Risks
• Data Breaches and Unauthorized Access: Misconfigurations (the cause of 99% of cloud
breaches) and weak credentials often lead to data breaches, threatening the
confidentiality of sensitive information.
• Model Theft and Poisoning: Malicious actors may steal proprietary AI models (model
stealing) by querying them, or inject corrupted data (data poisoning) into training sets to
corrupt AI reliability.
• Adversarial Attacks: Input manipulation (e.g., modified images) can trick AI into creating
incorrect outputs, which is a critical concern for autonomous vehicles and similar systems.
, Continue…
• Shared Responsibility Misunderstandings: A significant challenge is defining the
boundaries of responsibility between cloud providers (security of the cloud) and
customers (security in the cloud).
• Lack of Visibility: The distributed, multi-tenant nature of cloud environments
makes it difficult for companies to monitor where their data is stored, processed,
and who has access.
