Feed
Grid
What You Could Build If IAM Let You: New Policies From Undocumented Condition Keys
In the previous post, we mapped 36 condition keys that the IAM engine evaluates but has never documented. The decomposition model, the service-specific resource identifiers, the organizational metadata - all of it sitting in the request context, invisible unless you probe for it. That post was about discovery. This one is about what you can…
Show Me the Context: Building the Full Request Context for AWS IAM
This post was written in early August 2026, ahead of our fwd:cloudsec Europe talk. On August 25, AWS launched the Access Troubleshooter (currently in public preview), a first-party feature that surfaces the request context for denied requests. We've updated this post to account for it. When the IAM engine evaluates your request, it matches your…
Building the Future: Introducing the Upwind AI Security Lab
I have always been fascinated by what comes next. Growing up, I watched technology reinvent itself again and again, from early gaming and the dot-com era to SaaS, cloud, and modern software development. I remember wondering when I would get the chance to help build what came next. At the time, I was mostly watching…
5 Back to School Security Predictions: The Attacker Class Average Just Moved
Back to school season is here, which makes this a good moment to look at what changed in the threat landscape over the summer. AI-assisted attackers haven't climbed toward the top of the field so much as filled in the middle of it and the middle is the population almost no security program was designed…
The Org Chart is The Wrong Security Boundary
The department model of dividing companies into groups earned its place. Finance gets the finance applications, engineering gets the repositories, HR gets the HRIS, and the boundaries hold because software was bought per function and used per function. Identity and access management platforms, such as Active Directory and Okta, were built around exactly that. Groups…
What’s Behind the Curtain? AWS Bedrock AgentCore Runtime Tear Down – Part III
Recap In Part I and Part II, we: Networking and VPC Mode Network Isolation Testing The microVM is assigned an IPv6 address matching the value in the JWT. Across multiple runs, all addresses shared the same 2600:1f18::/32 prefix, but cross-microVM communication always failed. We attempted to reach the host EC2 IMDS by manipulating the route…
What’s Behind the Curtain? AWS Bedrock AgentCore Runtime Tear Down – Part II
Recap In Part I, we explored the AgentCore Runtime microVM from the inside and discovered we weren't alone - four platform binaries were running alongside our code, and one of them was quietly shipping logs to an AWS-internal S3 bucket. We left off with a question: what can we learn from these internal components, and…
What’s Behind the Curtain? AWS Bedrock AgentCore Runtime Tear Down – Part I
Introduction When you deploy an AI agent to AWS Bedrock AgentCore Runtime, your code runs inside a Firecracker microVM - but it doesn't run alone. In this three-part series, we tear down the platform internals, document what we found, and assess how well the isolation holds up. Setting the Stage AWS Bedrock AgentCore Runtime is…
Upwind Code Expands Enterprise Coverage to Azure DevOps and Bitbucket Cloud
Modern development organizations rarely keep all their code in one place. Teams may use different version control platforms because of acquisitions, business-unit preferences, regional requirements, or existing development workflows. But when code is spread across multiple providers, application security coverage can become fragmented too. Today, Upwind Code adds support for Azure DevOps and Bitbucket Cloud.…
Upwind Blue Agent – Increasing the Scope and tooling to a new level of incident response
Cloud attacks do not stay within the boundaries of a single security tool. An intrusion can begin with an API request, execute a process inside a Kubernetes workload, modify a file, contact an external host, use a cloud identity, and change cluster state, all as part of the same incident. But the evidence needed to…
Upwind for Microsoft Sentinel – Available on Marketplace and Security Store
Security teams should not have to switch between tools to understand what is happening across their cloud environments. That’s why we’re excited to announce that the Upwind solution for Microsoft Sentinel is now available through the Microsoft Marketplace and the Microsoft Security Store. The integration brings Upwind security data directly into Microsoft Sentinel, helping security…
You Can’t Crowdsource Your Way to a Live Adversary
Bug bounty programs were built on a single assumption: that finding a vulnerability was the hard, scarce, expensive part worth paying for. That assumption held for about a decade, then AI erased it. When anyone can point a model at your code and receive a plausible-looking finding back in seconds, a crowd of finders stops…
What You Could Build If IAM Let You: New Policies From Undocumented Condition Keys
In the previous post, we mapped 36 condition keys that the IAM engine evaluates but has never documented. The decomposition model, the service-specific resource identifiers, the organizational metadata - all of it sitting in the request context, invisible unless you probe for it. That post was about discovery. This one is about what you can…
Latest
Show Me the Context: Building the Full Request Context for AWS IAM
This post was written in early August 2026, ahead of our fwd:cloudsec Europe talk. On August 25, AWS launched the Access Troubleshooter (currently in public preview), a first-party feature that surfaces the request context for denied requests. We've updated this post to account for it. When the IAM engine evaluates your request, it matches your…
Building the Future: Introducing the Upwind AI Security Lab
I have always been fascinated by what comes next. Growing up, I watched technology reinvent itself again and again, from early gaming and the dot-com era to SaaS, cloud, and modern software development. I remember wondering when I would get the chance to help build what came next. At the time, I was mostly watching…
5 Back to School Security Predictions: The Attacker Class Average Just Moved
Back to school season is here, which makes this a good moment to look at what changed in the threat landscape over the summer. AI-assisted attackers haven't climbed toward the top of the field so much as filled in the middle of it and the middle is the population almost no security program was designed…
The Org Chart is The Wrong Security Boundary
The department model of dividing companies into groups earned its place. Finance gets the finance applications, engineering gets the repositories, HR gets the HRIS, and the boundaries hold because software was bought per function and used per function. Identity and access management platforms, such as Active Directory and Okta, were built around exactly that. Groups…
What’s Behind the Curtain? AWS Bedrock AgentCore Runtime Tear Down – Part III
Recap In Part I and Part II, we: Networking and VPC Mode Network Isolation Testing The microVM is assigned an IPv6 address matching the value in the JWT. Across multiple runs, all addresses shared the same 2600:1f18::/32 prefix, but cross-microVM communication always failed. We attempted to reach the host EC2 IMDS by manipulating the route…
What’s Behind the Curtain? AWS Bedrock AgentCore Runtime Tear Down – Part II
Recap In Part I, we explored the AgentCore Runtime microVM from the inside and discovered we weren't alone - four platform binaries were running alongside our code, and one of them was quietly shipping logs to an AWS-internal S3 bucket. We left off with a question: what can we learn from these internal components, and…
What’s Behind the Curtain? AWS Bedrock AgentCore Runtime Tear Down – Part I
Introduction When you deploy an AI agent to AWS Bedrock AgentCore Runtime, your code runs inside a Firecracker microVM - but it doesn't run alone. In this three-part series, we tear down the platform internals, document what we found, and assess how well the isolation holds up. Setting the Stage AWS Bedrock AgentCore Runtime is…
Upwind Code Expands Enterprise Coverage to Azure DevOps and Bitbucket Cloud
Modern development organizations rarely keep all their code in one place. Teams may use different version control platforms because of acquisitions, business-unit preferences, regional requirements, or existing development workflows. But when code is spread across multiple providers, application security coverage can become fragmented too. Today, Upwind Code adds support for Azure DevOps and Bitbucket Cloud.…
Upwind Blue Agent – Increasing the Scope and tooling to a new level of incident response
Cloud attacks do not stay within the boundaries of a single security tool. An intrusion can begin with an API request, execute a process inside a Kubernetes workload, modify a file, contact an external host, use a cloud identity, and change cluster state, all as part of the same incident. But the evidence needed to…
Upwind for Microsoft Sentinel – Available on Marketplace and Security Store
Security teams should not have to switch between tools to understand what is happening across their cloud environments. That’s why we’re excited to announce that the Upwind solution for Microsoft Sentinel is now available through the Microsoft Marketplace and the Microsoft Security Store. The integration brings Upwind security data directly into Microsoft Sentinel, helping security…
You Can’t Crowdsource Your Way to a Live Adversary
Bug bounty programs were built on a single assumption: that finding a vulnerability was the hard, scarce, expensive part worth paying for. That assumption held for about a decade, then AI erased it. When anyone can point a model at your code and receive a plausible-looking finding back in seconds, a crowd of finders stops…
arrayref Supply Chain Attack: A One-Line Build Dependency Ran a Backdoor During cargo build
Key Takeaways Executive Summary arrayref 0.3.10 is a hijacked release of a widely used Rust utility crate that added one dependency, proc-macro1, whose build script downloaded and executed a remote binary at compile time. The release was live on crates.io for 86 minutes on August 20, 2026, alongside [email protected] and [email protected] published from the same…