Announcements from AWS re:Inforce 2024 Keynote

#aws #security #ai #news

AWS re:Inforce is the Amazon Web Services annual event focused on security.

The event was led by Chris Betz, CISO of AWS.

During the keynote, Chris shared some of the insights that AWS embed security as part of their company's culture.

He talked about the Security Guardians program, a mechanism for distributing security ownership, and a culture of escalation – a process of making sure that the right people know about the problem at the right time.

AWS infrastructure is secured by design at all layers – from hardware, virtualization, compute, networking, storage, and finally at the apps and data layer.

AWS Graviton4 was designed with built-in security features such as:

Pointer authentication
No simultaneous multi-threading (SMT)
Full encryption of all high-speed physical interfaces
Branch target identification

AWS Nitro System supports full isolation of customer's AI data from AWS operators:

Encrypt sensitive AI data using keys that customers own and control
Store data in a location of the customer's choice
Securely transfer the encrypted data to the enclave for inferencing
Encryption for ML accelerator, to Nitro, to the network, and back Reference:
A secure approach to generative AI with AWS

AWS is using Automated Reasoning for multiple purposes, such as:

Verify the correctness of cryptographic protocols, authorization logic, and consistency of storage systems (such as Amazon S3 ShardStore)
Verify security mechanisms such as firewalls, detection, and coding practices

Zero Trust challenges:

A strong identity and access management
Hybrid environments
Complex network segmentation
Expanding application landscape and workforce mobility

Announcement - AWS Private CA Connector for SCEP (Currently in Preview)

Simple Certificate Enrollment Protocol (SCEP), lets you use a managed and secure cloud certificate authority (CA) to enroll mobile devices securely and at scale.

References:

Announcement - Passkeys as 2nd Factor Authenticators in AWS IAM

AWS now allows customers the options for strong authentication by launching support for FIDO2 passkeys as a method for multi-factor authentication (MFA) as we expand our MFA capabilities. Passkeys deliver a highly secure, user-friendly option to enable MFA for many of our customers.

References:

Announcement - IAM Access Analyzer unused access findings recommendation (Currently in Preview)

AWS IAM Access Analyzer provides tools to set, verify, and refine permissions. With the new announcement, IAM Access Analyzer offers actionable recommendations to guide you to remediate unused access.

References:

Announcement - Malware Protection for S3 Amazon GuardDuty

Amazon GuardDuty is a threat detection service that continuously monitors, analyzes, and processes specific AWS data sources and logs in the AWS environment. This expansion of GuardDuty Malware Protection allows scanning newly uploaded objects to Amazon S3 buckets for potential malware, viruses, and other suspicious uploads and taking action to isolate them before they are ingested into downstream processes.

References:

AWS Generative AI stack and built-in security controls:

Amazon Q - Tools and services to write secure and robust code (Amazon Q Developer, Amazon Q Business)
Amazon Bedrock - Helps keep data secure and private. All data is encrypted in transit and at rest. Data used for customization is securely transferred through the customer's VPC
AWS Nitro System - Allows customers to secure AI infrastructure includes zero trust access to sensitive AI data

Announcement - Generative AI-powered query generation AWS CloudTrail Lake (Currently in Preview)

AWS CloudTrail Lake lets customers run SQL-based queries on your events. This new feature empowers users who are not experts in writing SQL queries or who do not have a deep understanding of CloudTrail events.

References:

Steve Schmidt, the Chief Security Officer of Amazon, shared some of the experiences Amazon has had using generative AI.

Generative AI security scoping matrix:

Consumer App - Using "public" generative AI services
Enterprise App - Using an app or SaaS with generative AI features
Pre-trained models - Building an app on a versioned model
Fine-tuned models - Fine-tuning a model based on customer's data
Self-trained models - Training a model from scratch, based on the customer's data References:
An Introduction to the Generative AI Security Scoping Matrix
Securing generative AI: Applying relevant security controls

Handling service data properly:

Know what you have, where is it, how it is stored, who has access for what purposes, and how that data is used over time
Trust boundaries for retrieval-augmented generation (RAG)
Continued testing
Security guardrails (such as GuardRails for Amazon Bedrock)

The full keynote is available at: https://www.youtube.com/watch?v=skH3Q90llss

About the Author

Eyal Estrin is a cloud and information security architect, and the author of the books Cloud Security Handbook, and Security for Cloud Native Applications, with more than 20 years in the IT industry.

You can connect with him on Twitter.

Opinions are his own and not the views of his employer.

👇Help to support my authoring👇

☕Buy me a coffee☕