Breach Notices Count From Confirmation. Attackers Count From Login
Archi3 min read
Contents
The clock in the headline is the wrong clock#
A vendor ships an emergency patch on a Tuesday. Its advisory says exploitation started in January. By Thursday, a dozen outlets have compressed that into one number, a six month window, and readers carry away the idea that everyone was exposed for six months. The window describes how long a flaw existed in code. It says nothing about when any particular network was entered, how long the intruder stayed, or when someone noticed.
Those three numbers decide the cost. They almost never appear in the story.
Appliances do not run your endpoint agent#
Most edge devices, VPN concentrators, firewalls, hypervisor management planes, out-of-band cards, cannot host a third-party agent. Vendors say so in the compatibility matrix, quietly, on page nine. So the box that terminates remote access for a thousand employees reports to the SIEM as a line of syslog, if the team forwarded it, and a NetFlow record if they did not.
Take a firm with 40 such devices across two data centers. If six of them stream logs and the rest send nothing beyond SNMP traps, the detection surface is six boxes. Attackers who read the same advisories you do already assume that ratio.
The patch rate is not the exposure rate#
An inventory sheet showing 94 percent of appliances on current firmware looks like a win. It hides the two devices facing the internet with a management interface on a public address, the four running a branch the vendor stopped supporting, and the shared local admin account that six contractors use.
The exploited path in most appliance incidents needs no credential at all. It reaches a service that is listening before authentication is requested. That is why an MFA rollout announced in the same week as the breach often does nothing to the vector that was actually used.
Notifications start at confirmation, and confirmation comes late#
Legal deadlines in the United States vary by state, and the European 72 hour countdown starts when a controller becomes aware. None of those clocks start when the first malicious packet arrived. They start when someone with authority concludes that a personal data breach occurred.
The gap between those two moments is where the real work happens, and it is the part the headline skips. A common sequence: an anomaly shows up in a firewall log in March, a junior analyst closes the alert, a second alert in May gets escalated, the forensic retainer is signed in June, and letters go out in July. The letter says "recently discovered." The logs say otherwise.
Regulators in several jurisdictions now ask for the log retention windows that support those dates. Keep 30 days of logs and the honest answer becomes "we cannot reconstruct the timeline," which is a second problem stacked on the first.
The recovery bill is not the ransom number#
Headlines quote a ransom demand. The costs that hit the income statement are the rebuild of every appliance from known-good firmware, the reissue of certificates and keys that lived in those boxes, the weeks of reduced capacity while the edge runs on a temporary design, and a forensic firm's hourly rate applied to a fleet with partial log coverage.
Crews that hit an appliance often encrypt the configuration backup first, because it usually sits on a share the appliance can reach. A clean copy stored offline, with a restore tested this quarter, changes the outcome more than any new detection rule.
Four questions your own logs can answer#
, Which internet-facing devices sent a log entry in the last 24 hours, by name rather than by IP?, Can we list every firmware version in the fleet without opening a spreadsheet someone saved in March?, Do we hold authentication records for local accounts on those devices, or only for the directory?, When did we last restore a device configuration and confirm it booted clean?
None of those require a purchase order. Three of them require someone to write down the answer once and recheck it on the first of each month.
The advisory headline will keep reporting the vulnerability's age, because it is the one number a vendor can publish. The notification letter will keep reporting the discovery date, because it is the one date a lawyer can defend. Between those two facts sits the actual incident, in a log file with a short retention window and a hostname nobody on the call recognizes.
More articles