Skip to content

Topic

Detection Engineering

Security discipline of writing, tuning, and maintaining detection rules and alerts that turn system telemetry into signals of malicious activity.

Current stories

buildOne report1 publisher

Wazuh 4.14.7 passes its config check while discarding rules written above their parent

Wazuh 4.14.7 ignores a custom rule whose if_sid parent appears later in the same file, yet wazuh-analysisd -t still exits 0, a dev.to lab test found. The only sign is two warnings in the output, so a deploy gate that reads only the exit code ships a dead rule.

Publishers:dev.to

Reality

Evidence55
Adoption
Not scored
Hype gap+5
Incentives70
Confidence50
leadershipOne report1 publisher

Triage and tuning decided which SOC caught CISA's red team

CISA's red team breached two critical-sector organisations, and only the water-sector one contained it, isolating machines in up to 20 minutes. The gap traces to alert tuning and triage, so the first fix most SOCs need is analyst time.

Publishers:itpro.com

Reality

Evidence45
Adoption
Not scored
Hype gap+20
Incentives55
Confidence45
buildOne report1 publisher

An autonomous AI agent narrated its own post-exploitation moves into DIVD's logs

DIVD says the autonomous AI agent behind its first security incident in seven years wrote its own reasoning into the system logs. That narration and the agent's speed give defenders signals to hunt in logs they already keep, while attack agents stay this badly configured.

Publishers:dev.to

Reality

Evidence35
Adoption
Not scored
Hype gap+25
Incentives
Not scored
Confidence35
securityOne report1 publisher

Conifers finds 47% of the average organization's detections need attention despite showing as deployed

Conifers assessed 14,652 customer detections and found 47% at the average organization need attention while still showing as deployed. Against the ATT&CK techniques relevant to each customer, average protection stood at 64%, leaving one in three without a reliable detection.

Reality

Evidence38
Adoption
Not scored
Hype gap+30
Incentives68
Confidence40
buildOne report1 publisher

AegisGate measured its 100% true-positive rate on the patterns it already had

AegisGate's founder replayed 24 attack prompts gathered from advisories and disclosed incidents against his own gateway, which scores 99.8 out of 100 on evasion resistance. It blocked 13, and seven new regexes followed.

Publishers:dev.to

Reality

Evidence35
Adoption
Not scored
Hype gap+45
Incentives65
Confidence60
securityOne report1 publisher

Huntress rebuilt a 175-endpoint INC ransomware case from scheduled tasks and a driver fragment

Huntress reconstructed an INC intrusion from the logs that had survived. The tradecraft it found was ordinary operator work, the kind a name-based indicator list misses, and how the attackers first got in is still an open question.

Publishers:huntress.com

Reality

Evidence62
Adoption28
Hype gap−5
Incentives65
Confidence55
buildOne report1 publisher

A three-week Elastic baseline stood in for the SIEM proposal at a 70-developer product shop

An anonymized consulting write-up says a product team of 70 to 100 developers got failed-auth, RBAC and container-exec visibility out of the Elastic cluster it already ran for troubleshooting. The post describes the sequence and generalizes the technical details.

Publishers:dev.to

Reality

Evidence22
Adoption12
Hype gap+18
Incentives78
Confidence45
securityOne report1 publisher

Enterprise ATT&CK splits Defense Evasion into Stealth and Defense Impairment

The Enterprise matrix now runs 15 tactics, and the two that replaced Defense Evasion leave different forensic evidence, so detection content mapped to the old tactic has to be reassigned technique by technique.

Reality

Evidence48
Adoption20
Hype gap+18
Incentives82
Confidence55
leadershipOne report1 publisher

Verizon's 2026 breach report puts the median AI-assisted actor at about 15 known ATT&CK techniques

Jon Baker of AttackIQ treats the subscription-gated report as evidence that AI hands any attacker fast access to a technique catalog that has not grown, and builds a spending case for chokepoint detection on it.

Publishers:forbes.com

Reality

Evidence32
Adoption30
Hype gap+32
Incentives78
Confidence52
buildOne report1 publisher

Extended Passport parsing puts SAP's CVSS 10.0 overflow ahead of every logon check

Onapsis says the OVERPASS overflow fires while the SAP kernel processes an Extended Passport at session initiation, which happens before authentication or logon policy is consulted. That leaves reachability and patch state as the only variables.

Publishers:dev.to

Reality

Evidence47
Adoption
Not scored
Hype gap+9
Incentives62
Confidence43
securityOne report1 publisher

Fourteen percent of attack actions raised an alert across Picus's 338 million simulations

Picus's Blue Report blames performance issues and log-collection gaps for the misses, both of them configuration work, which is why an AI SOC pointed at that pipeline would only get faster at the one action in seven that already alerts.

Reality

Evidence32
Adoption28
Hype gap+45
Incentives88
Confidence52
buildOne report1 publisher

A backdoor built as an HAProxy filter suppresses the log lines that would show it

Rapid7's account puts the ted backdoor inside HAProxy 2.8.12 as an internal filter, matching requests by IP, URL and User-Agent and reducing the same counters an operator would use to triage it. Root was already held.

Publishers:dev.to

Reality

Evidence46
Adoption20
Hype gap−5
Incentives55
Confidence50
buildOne report1 publisher

Enrollment approval and role scoping decide what a stolen RMM credential can reach

A guidance post on dev.to argues you cannot catch remote support abuse by identifying the binary. The eight signals it recommends instead all lean on baselines most teams have never written down, and that is where the cost sits.

Publishers:dev.to

Reality

Evidence28
Adoption
Not scored
Hype gap+28
Incentives78
Confidence58
leadershipOne report1 publisher

Attackers repurposed prompt injection's invisible Unicode to break email keyword matching

Microsoft's hunt for email-borne prompt injection instead surfaced a three-month phishing campaign hiding tag characters inside words like funding, which puts both problems on one detection team's desk.

Publishers:microsoft.com

Reality

Evidence58
Adoption62
Hype gap+12
Incentives68
Confidence55
securityConfirmed4 publishers

Phishing crews adopt the AI red team's invisible Unicode trick to break keyword filters

A hunting signature Microsoft built to catch prompt injection in email started firing on ordinary phishing instead, where invisible tag characters were splitting lure words like "funding" so filters never parsed them.

Perspective Coverage

4 publishers
Builder
Builder 31%
Operator
Operator 60%
Investor
Investor 9%

Reality

Evidence64
Adoption72
Hype gap+16
Incentives74
Confidence66
buildOne report1 publisher

Origin validation ships off by default in four of five dominant WebSocket frameworks

The upgrade request is an HTTP GET that carries session cookies cross-origin with no preflight, so CORS never sees it, and the frameworks that could check the Origin header for you mostly hand that job to your route code.

Publishers:dev.to

Reality

Evidence55
Adoption
Not scored
Hype gap+18
Incentives30
Confidence52
buildOne report1 publisher

Elementor Pro's upload validator returns early when the first array element is empty

The bypass needs three separate conditions to line up before it reaches command execution. Wordfence's blocked-attempt telemetry tells you nothing about whether they lined up on your site.

Publishers:dev.to

Reality

Evidence58
Adoption35
Hype gap+15
Incentives55
Confidence55
securityOne report1 publisher

Outsider's affiliates stood up 700 new domains within a month of Operation Ghost Hook

Group-IB says the phishing kit Google sued over in June kept producing pages after the FBI seized its admin servers and wallets, which puts the durable detection signal in the kit's file names rather than its hosts.

Reality

Evidence44
Adoption66
Hype gap+6
Incentives58
Confidence46
securityOne report1 publisher

Sysdig clocks an AI-assisted intrusion reaching admin in eight minutes

The vendor's threat-research figures arrive without CVE identifiers, dates or sample sizes, worth flagging because the operational consequence is a patch window attackers shut before any patch gets applied.

Reality

Evidence34
Adoption45
Hype gap+42
Incentives88
Confidence38
buildOne report1 publisher

Superior ships its wallet drainer as a routine auto-update to extensions users already trusted

An extension allowlist pins an ID, and the ID survives a sale. The operators behind Superior bought or seeded trust that users had already granted, then spent it in an automatic update that stripped page CSP on the way through.

Publishers:dev.to

Reality

Evidence55
Adoption40
Hype gap+12
Incentives55
Confidence50

Earlier coverage

  1. Agent Tesla v4 hides in emoji and never hits disk: an email-rule problem, not a new-malware one

    Security · August 21, 2026 · Confirmed2 publishers

  2. The AI security line item to fund first is log coverage, not another agent

    Security · August 18, 2026 · Confirmed2 publishers