- Date
- January 25, 2003
- Location
- United States
- Industry
- Energy
- Access vector
- External Remote Services
- OT targeting
- Indirect: IT dependency
- Primary peril
- None
- Impact duration
- Under 1 day
Davis-Besse nuclear plant SQL Slammer infection, 2003

The incident path, stage by stage
Each stage of the attack, from intent to financial loss, with this incident’s path highlighted. Solid cells are confirmed, dashed cells are reported by a source but not confirmed, dotted cells are implied by the facts, and Unknown marks what is not publicly known. Click a stage title or a highlighted cell to read that part of the analysis.
Scroll sideways to see every stage.
On this page
Highlights
What happened. On January 25, 2003, the SQL Slammer worm reached the Davis-Besse nuclear power station in Ohio through a consultant’s T1 line that bypassed the plant’s firewall, and overloaded the site network[1][2]. The Safety Parameter Display System was unavailable for 4 hours 50 minutes and the plant process computer for 6 hours 9 minutes; the Nuclear Regulatory Commission (NRC) says plant control and protection functions were not affected[1].
Why it matters. An indiscriminate worm with no industrial logic reached safety monitoring systems at a nuclear plant. Little followed because the reactor was already offline and defueled for unrelated reasons, not because the defenses stopped it[2].
About this analysis. This analysis sets out what is known, stage by stage. The incident became public only in August 2003, through press reporting, and the operational facts are documented far better than the financial ones. It reflects the best available information as of October 22, 2026; new information could confirm, change or discredit parts of it.
Highlights
- The plant’s enterprise firewall was configured to block UDP port 1434, the port Slammer used, but the consultant’s T1 had been implemented behind it[2].
- Both affected systems had redundant analog backups that were unaffected by the worm[2][3].
- Davis-Besse was offline and safely defueled at the time because a large cavity had been found in the reactor pressure vessel head, so there was no generation to lose[2].
- No one has attributed a cost to the event. INL’s estimates of $1 billion in recovery and $1.2 billion in lost productivity cover all Slammer victims worldwide[2].
Attack flow
- Updated: 2026-10-22
- Intent: Malicious [confirmed | worm author’s intent; plant was collateral]
- Threat actors: Not attributed [unknown | No author named]
- Access vector: External Remote Services [confirmed | consultant’s T1 behind the firewall; T1199 inferred]
- IT impact: Malware [confirmed | SQL Slammer, T1210]; Denial of Service [inferred | T1498]
- OT targeting: Indirect: IT Dependency [confirmed]
- OT sophistication: Not applicable [not in path | No OT-aware logic]
- OT impact: Loss / Denial of View [confirmed | T0829]
- Perils: No Perils [confirmed | No physical event; reactor defueled]
- Financial losses: Cyber Incident Response [confirmed]
What happened
Slammer was released on January 25, 2003. Idaho National Laboratory (INL) says it infected about 300,000 unpatched hosts, more than 90% of vulnerable hosts within 10 to 15 minutes. It exploited a buffer overflow in Microsoft SQL Server over UDP port 1434 and carried no payload beyond spreading[2].
It first compromised a server on a consultant’s company network. The consultant provided application software that ran on a server, and a T1 line it used bypassed the corporate firewall[1]. INL says the T1 had been implemented behind Davis-Besse’s external-facing enterprise firewall, whose ruleset blocked UDP 1434[2]. The worm infected corporate-network servers with port 1434 open, residing only in memory, and INL says it spread to an unpatched SQL Server 2000 host in the operations network[1][2].
Its traffic left the site’s computers unable to communicate with each other. The Safety Parameter Display System became unavailable by 4:50 p.m. and stayed down for 4 hours 50 minutes; the plant process computer was lost by 5:13 p.m. and stayed down for 6 hours 9 minutes[1]. No source gives a clock time for either recovery. According to SecurityFocus, the plant had been offline since February 2002[3].
The incident became public on August 20, 2003, in a SecurityFocus report republished by The Register[3]. The NRC issued an information notice on August 29[1], and in October Representative Edward J. Markey wrote to the NRC chairman calling the agency’s earlier answers vague, incomplete or contradictory, and asking about firewalls, patching and the timing of the notice[4].
Intent: malicious, but not aimed at the plant
Slammer was deliberately written and released, so the path starts at Malicious. That records the author’s intent, not any intent toward Davis-Besse. The worm exploited a SQL Server flaw wherever it found one, and the plant was collateral: it was reachable, not chosen[2]. No source describes any demand, message or follow-on activity aimed at the plant.
Threat actors: not attributed
No source in the record names Slammer’s author. The worm spread indiscriminately across the internet, infecting about 300,000 hosts within minutes, and nothing in the NRC or INL accounts suggests any actor chose Davis-Besse[2][1]. No threat actor is highlighted on the map.
Access vector: a contractor’s line around the firewall
The route is the finding. The NRC says the consultant’s T1 line bypassed the corporate firewall[1]. INL says it was implemented behind the external-facing enterprise firewall, whose rules blocked UDP 1434, and calls it an unmonitored backdoor[2]. The rule was right; the traffic never met it.
SecurityFocus quotes an April 2003 e-mail from FirstEnergy to the NRC calling the T1 a backdoor from the internet to the corporate internal network that was not monitored, and saying some Network Services staff knew of it and some did not. The same report says it was one of multiple entry points that bypassed the firewall[3]. The e-mail itself has not been located, so those quotations are unverified at first hand. The map shows the external connection as confirmed; reading it as MITRE’s Trusted Relationship technique is inferred.
IT impact: a worm that only spread
The worm infected corporate-network servers that had port 1434 open, and it resided only in memory; the fix was the Microsoft security patch[1]. It had no payload beyond spreading, and its effect came from that: the NRC says computers on the site network could not communicate with each other while it did[1]. That network denial of service is inferred from the NRC’s statement. Malware is confirmed.
OT targeting: indirect, through a shared network path
Slammer had no interest in anything industrial. The plant systems it disabled were reachable because a corporate-side network path led to them, which makes this an indirect effect through an IT dependency[2]. It did cross into the plant environment: the Safety Parameter Display System and the plant process computer are plant monitoring systems, and both were down because they were infected, not as a precaution[1]. What it reached was monitoring, not control or safety actuation.
OT sophistication: not applicable
Slammer contained no OT-aware logic of any kind: no protocol knowledge, no process awareness and no payload beyond propagation[2]. It disabled a safety parameter display system by consuming resources, not by understanding it. With no OT technique to rate, the stage is marked not applicable.
Nothing in this event required an adversary who knew what a nuclear plant was.
OT impact: loss of view
Operators lost two displays. The Safety Parameter Display System was unavailable for 4 hours 50 minutes and the plant process computer for 6 hours 9 minutes, which in MITRE ATT&CK for ICS terms is Loss of View, confirmed[1][2]. The NRC says plant control and protection functions were not affected[1].
The indications were never wholly lost. Both systems had redundant analog backups that the worm did not touch[2][3]. No source describes any control or protective function as affected.
Perils: none
There was no physical event. The worm caused a loss of view on two monitoring systems, plant control and protection functions were not affected, and the plant was offline and defueled at the time[1][2]. No source reports physical damage, a release or an injury, and the facts imply none.
Financial losses: what it cost
No utility, regulator or researcher has attributed a cost to this event. INL’s Slammer figures, about $1 billion in recovery costs and $1.2 billion in lost productivity, are global estimates across all victims, and DeNexus does not apportion them to one site[2]. With the plant defueled and the analog backups intact, the direct loss is plausibly small, but no number supports that.
Loss evidence
- Confirmed: Network event response and recovery (two plant systems restored the same evening); Voluntary improvements (documenting external connections, patching, a firewall between the plant and corporate networks, monitoring and filtering of plant network data, a patch review process)
- Not supported: Direct business interruption (the plant was already offline and defueled); Third-party physical and societal impact (no customer lost supply)
The corrective actions come from the NRC notice, which is informational, not an order, so the work is counted as the licensee’s own[1]. The absence of business interruption reflects the plant’s state at the time, and cannot be read as evidence that the consequence of such an infection is small.
What this incident shows
Davis-Besse’s firewall blocked the port Slammer used, and Slammer got in anyway, because a contractor’s line had been connected behind it[2]. Perimeter security is about the inventory of connections, not only the quality of the firewall: an organization with a well-configured firewall and an unlisted contractor link has no perimeter at the point that matters. The NRC had issued a security order on external connections that bypass network boundary protections in February 2002, eleven months before the infection[1].
The record also shows a working fallback. Redundant analog backups absorbed a digital failure on the plant’s monitoring systems[2][3]. But the outcome was set by the plant’s state, offline and defueled, more than by its defenses, and the evidence says nothing about how the same event would have played out on a running unit.
References
- U.S. Nuclear Regulatory Commission. “Potential Vulnerability of Plant Computer Network to Worm Infection.” NRC Information Notice 2003-14, 29 Aug. 2003. govinfo.gov/content/pkg/GOVPUB-Y3_N88-PURL-LPS126392/pdf/GOVPUB-Y3_N88-PURL-LPS126392.pdf
- Idaho National Laboratory, CyOTE Program. “SQL Slammer Worm Infection of Davis-Besse Nuclear Power Plant.” Case study, 31 Mar. 2023. cyote.inl.gov/content/uploads/24/2025/12/CyOTE-Case-Study_Davis-Besse-SQL-Slammer.pdf
- “Slammer worm crashed Ohio nuke plant net.” SecurityFocus, republished by The Register, 20 Aug. 2003. theregister.com/2003/08/20/slammer_worm_crashed_ohio_nuke/
- Markey, Edward J. Letter to Nils J. Diaz, Chairman, U.S. Nuclear Regulatory Commission, 20 Oct. 2003. NRC ADAMS ML032970134. nrc.gov/docs/ML0329/ML032970134.pdf


