All caught attacksAttack walkthrough / evidence trace / SWaT 2015

Attack #37P501 + FIT502

RO high-pressure pump and RO permeate flow meter. Attacker's intended impact: Reduced output.

Single Stage, Multi PointWater Treatment7m 53s windowfirst catch 44s4 of 4 engines flagged it
1.0Plant

Where it hit the plant

P1

raw water intake

P2

chemical dosing

P3

ultrafiltration

P4

dechlorination / UV

P5

reverse osmosis

P501FIT502
P6

permeate transfer & backwash

2.0Timeline

Attack window and detections

Labelled window2015-01-02 11:17:02 to 2015-01-02 11:24:50plant local time
Physical Residual
46s
Temporal Sequence
45s
Process Invariants
44s
Fused Ensemble
44s

Shaded band: the labelled attack window from the iTrust ground-truth catalog. Marks: the first second each engine flagged the window. "No flag" means the engine stayed silent for the whole window.

3.0Evidence

Why it was flagged

At 2015-01-02 11:17:02, attack #37 (SSMP) manipulated P501, FIT502 — “Close P-501; Set value of FIT-502 to 1.29 at 11:18:36”. The serving model raised the alert 46 seconds after onset (peak score 5.2621x the operating threshold). The dominant physical evidence: FIT504 — RO recirculation flow meter, Stage 5 (reverse osmosis) ran 29.2 standard deviations above the value predicted by the plant's learned physics, carrying 16% of the anomaly score — 6.3x its share in normal operation. Supporting deviations were concentrated in FIT501, FIT503, P403. Independently, the mined process rule “PIT503 ≈ 0.7834·PIT501 + -7.309, residual ∈ [-0.7047, 0.6132]” was violated for 429 seconds inside the attack window. The fused detection stack corroborated the alert across channels (physics residual + process rules), which is the platform's high-confidence state. In plain terms: the plant's own physics stopped agreeing with what the sensors were reporting, in exactly the part of the process the attacker touched.

From the official explain run 2026-06-12T06:42:56Z
Corroborated across channelsPeak score 5.26x of the operating thresholdFused peak 1.50x of the operating threshold

Sensors carrying the anomaly score

SensorShare of scoreDeviationShare vs normal
FIT504
RO recirculation flow meter, Stage 5 (reverse osmosis)
15.7%29.2σ6.3x
FIT501
RO feed flow meter, Stage 5 (reverse osmosis)
14.0%26.1σ10.0x
FIT503
RO reject flow meter, Stage 5 (reverse osmosis)
10.9%20.3σ3.9x
P403
UV/RO feed pump (standby), Stage 4 (dechlorination / UV)
7.4%13.9σ7.3x
FIT502
RO permeate flow meter, Stage 5 (reverse osmosis)
5.3%9.8σ3.1x
P501
RO high-pressure pump, Stage 5 (reverse osmosis)
5.2%9.7σ45.9x
AIT504
RO permeate conductivity analyzer, Stage 5 (reverse osmosis)
4.9%9.3σ2.9x
PIT501
RO feed pressure transmitter, Stage 5 (reverse osmosis)
4.0%7.0σ1.8x

Process rules violated inside the window

  • PIT503 ≈ 0.7834·PIT501 + -7.309, residual ∈ [-0.7047, 0.6132]
    correlation bandviolated for 7m 9sconfidence 99.97%
  • FIT501 ≈ 0.9242·FIT401 + 0.1401, residual ∈ [-0.007591, 0.007749]
    correlation bandviolated for 7m 8sconfidence 99.95%
4.0Provenance

Every figure has a run.

Validation run

2026-06-11T22:34:18Z, commit 97752838a3f3d5271b7e43335fcf1f839fdf057a

Model image

sha-97752838a3f3

Metrics artefact SHA-256

80c98149b000d88af3fc7b58ab407a24c81f3839426e7d4e53c2011bbf15c52a

Explain run

2026-06-12T06:42:56Z

Explain artefact SHA-256

2616f67199c281fc558b4b05e3cccdeb223910e4c07addf0c80bc50f0ec5b333

Generator

vardar-portal/scripts/build-gallery-data.mjs (DEL-34/DEL-43)

SWaT and WaDi datasets courtesy of iTrust, Centre for Research in Cyber Security, Singapore University of Technology and Design. Detection figures trace to the reproducible model runs shown with each dataset.

The next record is your process.

A technical evaluation on one critical process, with success criteria agreed in advance, produces the only figure that applies to your site.

Evaluate one critical process