TIA · HYDRA
Cyber Signal
A daily cut from public sources. You don't subscribe to it. It just gets published.
No account. No email address. No tracking.
A daily cut from public sources. You don't subscribe to it. It just gets published.
No account. No email address. No tracking.
no item today reports exploitation in practice
This page went out at 10:00 with 7 items. At 10:26 the radar producer reported a scoring defect: the critical-infra/OT rule fired on the bare word „water", so the phrase „Water resistance: IP68" carried a consumer phone review over the threshold. Three further general AI articles cleared by exactly zero margin (source base 2 + broad tag 12 = 14, against a threshold of 14). After the rubric was fixed and regression tests added, the producer withdrew 4 items as noise. Three remain. The link stays valid and the original receipt stays sealed. Items were not removed because they suited us — they were removed by a corrected rule that now applies to every following day.
The same revision added the weekly edition W3 (31 Aug – 6 Sep) to this page. It was due with the Sunday cut and did not go out. Nothing was added to the cut itself; the weekly is its own section and carries its own limits.
records parsed from the source feeds — before the time window and de-duplication
The digest hands us counts, not the discarded items — so this shows how many and why, not which.
The cut is thin for two reasons and neither of them is a quiet day in security. The window covers a Sunday, when security sources publish little while general technology desks publish daily. And the producer withdrew 4 of the original 7 items as noise after fixing its scoring. Funnel: 12 → 7 → 3.
This is not a quiet day in security. It is a trace of when our own instrument ran. We are not taking a day off and we are not padding the gap — the date stays, because an empty day is a measurement too.
Today’s lowest-scoring item cleared the threshold by exactly zero points (14 out of 14): “Attackers conceal phishing lures using invisible Unicode characters”. We are leaving it in — it is more honest than anything we could replace it with.
July brought 4,084 new advisories for malicious npm packages. Other months run around 731 — July is a spike, not a trend.
All 8,189 of them, across seven months, are rated critical. Not one carries a computed CVSS score — there is nothing to score in a malicious package. It is not a flaw in the code, it is intent. Triage them by severity and every one is a tie.
But reach is distributed extremely unevenly. Of the 354 packages we could resolve, 67% have no dependents at all — their reach is zero. Among the rest the median multiplier is 1.95×, and 7% multiply thirtyfold or more. Highest measured case: engine.io has 118 direct dependents and reaches 15,865 through the tree. An average severity cannot see that spread at all.
Dependencies: a deps.dev v3alpha dependentCount measurement, SINGLE provider, no cross-check, window 2026-07. The package version is picked by a rule frozen BEFORE the run and BLIND to the measured value — the earlier method took the maximum across six versions, i.e. selected on the quantity it was meant to measure, and overstated the tail twofold. Amplification is undefined for packages with no dependents; those are reported separately as zero reach, not as a missing value. And advisories capture a fraction of malicious packages — this is the advisory denominator, not the malware denominator.
Seven collection days, no gaps, 83 published items. The previous edition put the target in the machine room. This one moves the agent from the path of an attack to a party in it — and closes on the question of who keeps the record.
The previous edition ended with the machine room under attack — Ray, MLflow, LiteLLM, the plumbing rather than the model. This period moves one step further: the agent stops being the path of an attack and becomes a party to it. It is used to port an exploit between PLC models. It is used inside a ransomware operation. It coordinates with other agents, for months, on a system belonging to nobody involved.
The two ends of the week matter more than either half. It opens with routers taken over and the security logs blinded. It closes with an agent fleet whose activity was classified by the only body positioned to classify it — the operator that owned it, which called it misalignment rather than a breach. Different technologies, same problem: the record is held by the party with an interest in it.
In these seven days 177 items crossed the threshold and 83 were published. The remaining 94 are in the record and never appeared on any daily page — and five of the seven days hit the publication cap of twelve, so the count is a ceiling, not a volume. One scoring configuration held across the whole period, so the arc is internally comparable. But on 7 September, the day after this period closed, that rubric was corrected: one rule fired on the bare word “water”, and general-technology AI articles cleared by exactly zero margin. Six items published in this period cleared at exactly the threshold, from the same source and the same tag. We are not restating the period under the new rubric and we are not removing them — the correction is dated and it applies forward. The arc above therefore carries no scores at all.
Seven days, seven pages, nothing is rewritten.
Attention classes are how much attention we gave an item. They are not severity verdicts.
N-able has released an emergency hotfix for a maximum-severity remote code execution (RCE) flaw affecting its N-central…
Threat actors have adopted the ASCII smuggling technique in phishing campaigns, using invisible Unicode characters to evade…
Elastic Security Labs has documented four previously unreported programs associated with REVSTEALER, an emerging Windows…