A financial services firm is adopting SageMaker and is concerned about preventing data exfiltration. Which mechanisms can an ML engineer use to control data egress from SageMaker? (Select three.)

Connecting to SageMaker via a VPC interface endpoint (AWS PrivateLink) ensures all traffic between your VPC and SageMaker remains within the AWS network, preventing data from traversing the public internet. Enabling network isolation for training jobs and deployed models restricts outbound internet access for these resources, allowing them to communicate only with specified VPC resources, thus preventing unauthorized data exfiltration. Restricting SageMaker notebook presigned URLs to specific IP addresses ensures that even if a URL is compromised, it can only be accessed from approved corporate networks, limiting potential data egress. Service Control Policies (SCPs) control permissions at the organizational level but don't directly prevent data egress from SageMaker itself. Disabling root access on notebook instances is a security best practice for preventing unauthorized system changes but doesn't directly control data egress. Encrypting data protects it if exfiltrated but doesn't prevent the exfiltration event itself.