Why spray campaigns still clear weak auth edges in 2026
Password spray remains one of the cheapest ways into remote access, webmail, and partner SSO. Attackers try a short list of common passwords across many accounts, stay under classic lockout thresholds, and harvest valid sessions before your SOC treats the noise as a helpdesk incident. Extortion crews such as Luna Moth have returned to law firms with the same playbook: valid credentials, quiet mailbox access, then pressure. Parallel reporting on 0ktapus-style campaigns against more than a hundred firms shows how quickly stolen or guessed logins become tenant-wide compromise once MFA prompts or recovery flows are weak.
Exchange and VPN fronts stay in the crosshairs. Turla-linked tooling against messaging platforms and weekly ISC Stormcast briefings keep reminding operators that failed auth is rarely random when the same source prefixes rotate through RDP, OWA, and IdP endpoints in the same window. Cursor’s confirmation that a supply-chain breach fed a malicious Claude Code extension is a different initial-access path, yet it lands on the same lesson for IT admins: treat authentication surfaces as high-value control planes, and assume credential probing will coincide with other noise such as PHP gateway scouts or cloud-provider employee data leaks that refresh password lists.
Practical detection starts with pattern shape, not a single bad password. Sequential brute force against one account looks like a lockout storm. Spray looks like many usernames, few passwords, low per-account failure counts, and shared ASN or residential-proxy exits. Your job is to promote that shape to a same-hour response before the helpdesk resets everyone and erases the attacker’s trail.
Signals that separate spray from ordinary lockouts
Wire these observations into SIEM or IdP analytics with a rolling one-to-four-hour window:
- Username cardinality: dozens to hundreds of distinct users failing from one /24, ASN, or proxy family, each with only a few attempts.
- Password reuse fingerprint: identical failure codes and attempt timing across unrelated accounts (for example the same three passwords every 90 seconds).
- Protocol fan-out: the same source hits VPN, RDP Gateway, and IdP password endpoints within minutes.
- Success after sparse failures: one soft success amid a spray is higher severity than a noisy lockout with no success.
- Off-hours admin and service accounts: spray against break-glass, scanner, or shared mailbox identities.
- Geo and ASN churn: consumer VPN and set-top exits that recycle after each soft block.
Concrete scenario: a mid-size firm sees 40 attorney accounts each fail twice against OWA between 02:10 and 02:40 UTC from two hosting ASNs. Nobody is locked out. At 02:41 one partner account succeeds, then a new device registers for MFA. That is spray plus account takeover staging, and it should page the on-call before anyone opens a bulk password-reset ticket.
Controls that slow attackers without melting the helpdesk
Pair detection with controls that raise cost without blanket account freezes:
- Per-source and per-ASN auth rate caps on IdP, VPN, and RDP, separate from per-user lockout.
- Adaptive MFA step-up on sparse multi-user failure clusters, including number matching and phishing-resistant factors for admins.
- Temporary quarantine of source prefixes that cross spray thresholds, with allow exceptions for known corporate egress.
- Password spray canaries: monitored non-person accounts that never receive legitimate logons; any attempt is high fidelity.
- Breach-dump window watches: tighten caps when large credential dumps hit public channels or when partner cloud breaches (for example employee data exposed via a provider) refresh corporate email passwords.
Operator checklist for the next spray window
Run this list during the first hour of a suspected campaign:
- Confirm the failure shape: many users, few attempts each, shared sources; export the username and source list.
- Check for any successes, new MFA devices, OAuth grants, or mailbox rule changes in the same window.
- Cap or quarantine the offending ASNs and proxy exits at the edge and IdP; keep per-user lockouts as a secondary control.
- Force step-up or session revoke only for accounts that saw success or admin-role proximity; avoid mass resets that flood the helpdesk.
- Hunt lateral paths: VPN profile downloads, RDP jumps, Exchange PowerShell, and SSO apps touched by any successful spray victim.
- Push the source indicators to peer sharing and edge decoys so the next wave is slower across sister tenants.
- Brief helpdesk with a script: verify caller identity out-of-band before reset; spray often overlaps social engineering of the reset queue.
- After containment, retune thresholds using the observed attempt rate so the next shift’s alerts fire before lockout counters do.
Implementation pitfalls that keep spray invisible
Per-user lockout alone is a poor spray detector. Attackers deliberately stay below three-to-five failures per account. If your only alert is “account locked,” you will see helpdesk volume while the campaign completes elsewhere.
Logging gaps break timelines. IdP, VPN concentrator, and reverse-proxy clocks that drift by minutes make “many users, one source” look like unrelated tickets. Sync NTP, join on normalized username and source IP, and keep at least 30 days of failed-auth detail for retrospective hunts when a new blocklist entry appears.
Over-broad geo denies create false confidence. Carrier NAT and anycast make “impossible travel” and country labels noisy. Prefer spray-shaped rate policy and ASN velocity over permanent country blocks on auth planes.
Autodrop blocklists without retrospective search miss the soft successes that already happened. When a prefix is quarantined, immediately query allow and success logs for the prior 24–48 hours against the same usernames.
MFA fatigue and consent lures ride the same week as spray. If your response only resets passwords, OAuth grants and push-bombed sessions remain. Include grant review and session kill in the same playbook.
Metric mismatch stalls funding. Scrubber and edge contracts measured only in gigabits will not protect auth queues. Report spray in attempts per minute per ASN, accounts touched, and time-to-quarantine so leadership sees auth as a control plane worth rate policy.
Tabletop the failure mode: spray at 02:00, one success, helpdesk flooded at 08:00. Your win condition is quarantine and session revoke before the first reset call, with a clear packet of source ASNs, touched accounts, and mailbox or SSO changes ready for escalation.