KASPR Datahaus · measurement record
Iran, June 2025: eight days, measured from outside
During the June 2025 war with Israel, Iran experienced the most complete national internet shutdown KASPR has recorded there. The share of measured addresses responding fell from a baseline near 11% to 0.14% at the deepest point — and because the measurement runs from outside the country, on a fixed schedule, the record continued uninterrupted through the blackout. The Economist drew on this measurement in its reporting at the time; this page sets out the full record.
8
days below 90% of baseline
0.14%
of measured addresses responding at the lowest point
175 of 254
districts with episodes of their own
81%
of normal, after partial restoration
The sequence
Strikes, cyber-attacks, then the switch
The reported sequence, drawn from The Economist and public reporting: Israeli strikes on Iranian nuclear facilities began on 13 June 2025. On 17 June, hackers linked to Israel attacked one of Iran’s largest banks and took roughly $90m from an Iranian cryptocurrency exchange. Officials presented the restrictions that followed as protection against further cyber-attacks; incoming international calls were also cut. United States strikes followed on 21 June, and a ceasefire was agreed on 24 June.
In the measurement record, the national series leaves its normalcy band on 2025-06-18. Two days later the connected share reaches its floor. Monitors at IODA, a research group at the Georgia Institute of Technology, logged the same near-total outage over those days — two independent instruments, one picture.
The national record
The floor was not zero — quite
The chart below is the daily share of measured addresses responding, against the baseline the anomaly model froze when the episode began. At the deepest point, 0.14% of measured addresses still answered. Analysis by Simon Angus of the Monash IP Observatory, reported in The Economist at the time, put the surviving fraction at about 1% of monitored addresses — and their locations suggested state-linked facilities: universities, military sites, airfields.
That remainder is informative. A shutdown that leaves official infrastructure connected while the public network goes dark has a different signature from a cable cut or a power failure, where the loss is indiscriminate. The measurement cannot attribute intent; it can characterise the shape of what happened, and this shape is selective.
Daily share of measured addresses responding (solid) against the frozen baseline (dashed). The shaded span is the episode window. The partial recovery from 21 June settles below the pre-war baseline.
A daily connectivity series that collapses nearly to zero for two days in mid-June 2025, recovers partially, and settles below its earlier baseline.The subnational record
Province by province, district by district
175 of 254 measured districts recorded negative episodes of their own inside the window. The provinces below fell furthest. The pattern is the same one the national series shows — near-simultaneous, near-total, and lifted in step — which is what a centrally coordinated restriction looks like in measurement, whatever its stated purpose.
| Province | From | Days | Depth |
|---|---|---|---|
| Kermanshah | 2025-06-18 | 3 | −100% |
| Lorestan | 2025-06-18 | 3 | −100% |
| Qazvin | 2025-06-20 | 7 | −100% |
| Hamadan | 2025-06-17 | 9 | −100% |
| Semnan | 2025-06-13 | 13 | −100% |
| Kordestan | 2025-06-18 | 3 | −100% |
| Ilam | 2025-06-18 | 3 | −100% |
| North Khorasan | 2025-06-18 | 3 | −100% |
| Ardebil | 2025-06-13 | 8 | −100% |
| Bushehr | 2025-06-18 | 7 | −100% |
Depth is the deepest daily shortfall against the frozen baseline during the overlapping episode, in per cent.
The pattern
Not the first, and not the last
Iran restricts internet access often. Access Now, an advocacy group, counted 34 shutdowns there in 2023 — more than any country except India and Myanmar, as The Economist noted in its coverage. Searches for VPNs and satellite-internet services surged during the June blackout, echoing the pattern of the 2022 protests. The Economist also reported a loosening before the war — WhatsApp formally unblocked in late 2024 — and, after the blackout, restoration to more than 80% of addresses on IODA’s measure, close to what the KASPR series shows.
The record since supports the caution in that reporting. Seven months later the same signature returned at greater length: twenty days below baseline in January 2026, documented in the district brief alongside this story.
Sources & method
The record behind this page
Connectivity figures, episode windows, depths and the province table are computed live from KASPR measurement — ICMP probing of measured IPv4 addresses from outside Iran, aggregated daily, with episodes defined against a 99.9% prediction band. The reported sequence of strikes, cyber-attacks and the ceasefire, the Access Now count, the IODA observations and the analysis of surviving addresses are drawn from the article cited below. Windows are windows, not attributions of cause.
“Iran has come back online — for now.” The Economist, Graphic detail, 26 June 2025. The article’s map of online and offline addresses during the blackout is Monash IP Observatory / KASPR Datahaus measurement. The Economist article
The same record updates for every measured country as events open. Open Alerts