Build1 distinct publisher2 min readUpdated
A public database, a 0.0.0.0/0 rule and a wildcard IAM role all get filed as risk. Only one of them shows up on this month's invoice, and that changes who can champion the fix.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
Trace the upsizing path, because it is the one that never reaches the security queue at all. A database with a public endpoint and a permissive group absorbs continuous scan and connection-attempt traffic, which burns connections and CPU, and eventually produces the "we need a bigger instance" conversation that a private subnet would have made unnecessary [2]. What lands in the record is a larger instance class. Nothing in the billing data says why.
Now count the shared list against the bill. The dev.to write-up proposes four audit items: public databases, unrestricted security groups, wildcard IAM, and a tripwire test on cost alerting [12]. Only the first carries standing monthly charges, through egress rates on served bytes and through the instance size that scan load provoked [13]. Open security groups and wildcard roles price at zero until someone monetizes them, and the fourth item is a detection check, not a spend item [13].
That asymmetry is where the joint-label pitch gets thin. Wildcard IAM costs nothing while nothing goes wrong; it sets the blast radius, and the author's own term for it is unpriced tail risk [6]. The difference described is between an incident report saying an attacker enumerated one bucket and one saying the attacker launched 400 instances in three regions the company does not use [7]. Finance has no current line to point at, so this item stays a security ask with a cost label stapled on, and it will lose to the RDS finding every time.
The organisational logic still holds for the items that do have a number. Cost carries a monthly forcing function executives already read; security carries veto power; a finding wearing both labels clears prioritization queues that pure-risk findings sit in for quarters, according to the same piece [8][9].
The audit itself is cheap enough that the argument is mostly about routing. Public exposure is one `describe-db-instances` query filtered on `PubliclyAccessible`, cross-checked against each instance's security group; open groups are a `describe-security-groups` filter on `0.0.0.0/0` with the intentional web listeners removed [11]. The disposition rules matter more than the commands: every remaining open rule gets closed or documented today rather than sprint-after-next, and every wildcard role either gets scoped or gets a named owner and an expiry date [15]. A role with an owner and a date is the only version of that finding that survives contact with a prioritization meeting, because someone has to renew it.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Security findings and cost findings are usually filed by different teams into different tools and treated as different kinds of problem, but a specific set of configurations sits in both piles: a database open to the internet, a security group accepting traffic from anywhere, and an IAM role with a wildcard where a scope should be.
Every byte an exposed endpoint serves outward is data transfer billed at internet egress rates.
Roles with AdministratorAccess or "Principal": "*" do not cost anything while nothing goes wrong; they define the blast radius. Wildcard IAM is unpriced tail risk, and the mitigation (scoping roles, requiring MFA, alerting on use of broad roles) is nearly free.
One leaked key with admin scope is the difference between an incident report that says "attacker enumerated one bucket" and one that says "attacker launched 400 instances in three regions we don't use."
Teams that fix public exposure with a bastion host per VPC buy a small standing cost: an instance, its EIP, its patching. Session Manager and equivalents removed the need for most of these years ago, so retiring bastions removes cost and attack surface in the same change.
The public-exposure audit is one query, aws rds describe-db-instances filtered on PubliclyAccessible, cross-checked against each instance's security group; the open-group audit is aws ec2 describe-security-groups filtered on ip-permission.cidr 0.0.0.0/0, then filtering out the intentional web-facing rules.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Verifiable mechanics, unsourced magnitudes
The cluster is a single practitioner post. Its checkable content - the two CLI audits, the existence of Session Manager as a bastion replacement, the fact that wildcard IAM carries no direct charge, egress billing on served bytes - is stated precisely enough to verify. Its load-bearing empirical claims - cryptomining as the dominant monetization, discovery on the bill rather than the SIEM, the $40,000 GPU weekend, scan load forcing instance upsizes, dual-labelled findings clearing queues - carry no dataset, incident reference, or citation of any kind, and no second publisher corroborates them.
No adoption signal in cluster
The supplied material contains no release, deployment, benchmark, disclosed usage, pricing change, or incident record - only a recommended practice. Nothing indicates how many teams run such a shared audit list, so adoption cannot be measured without inventing facts.
Mildly overstated on magnitudes, honest on scope
Slightly positive rather than strongly so. The piece inflates on two axes - a vivid unsourced $40,000 anomaly figure and an unsupported claim that dual-labelled findings clear queues where pure-risk findings sit for quarters - while pricing two of its four items at zero until a compromise occurs, which the framing 'the egress bill' partly obscures. Against that, the author explicitly disclaims a security program, names identity, detection and response as out of scope, and concedes that closing a port only makes you not-negligent, which is unusually restrained and pulls the gap back toward alignment.
Practitioner post, no vendor stake disclosed or visible
The single source promotes no product, tool, or paid service: every remediation it names is a native AWS capability or a configuration change, and no pricing, funding, or commercial relationship appears. The visible incentive is audience-building on a developer platform - the appended FAQ-style question-and-answer blocks and keyword-shaped title read as reach optimization - which is a mild slant toward confident, quotable framing rather than a financial conflict.
Low-moderate: one self-limiting source, no corroboration
Confidence is capped by cluster shape: one publisher, one item, zero adoption evidence, and no independent check on the frequency and cost claims. It is not lower because the operationally useful core - the audit commands, the pricing split across items, and the explicit scope boundary - is internally consistent and checkable by any reader with CLI access.
build
Two Actions, One Loose Policy: The Bedrock Wildcards That Widen A Least-Privilege Grant1 distinct publisher
security
Leaked AWS keys keep working: 526 root, 242 admin, and a rotation rate of 13.7 percent2 distinct publishers
build
Bedrock's evaluation modes grade what they can see, and the dataset outlives both1 distinct publisher
build
Two mechanisms, one vCPU floor: why db.t3.micro cannot meet a 1-second RPO on RDS1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
dev.to
1 article · August 23, 2026