Summary
On 23 December 2015, attackers remotely opened breakers at three Ukrainian electricity distribution companies and cut power to approximately 225,000 customers.1 A joint analysis by the SANS ICS team and the E-ISAC describes these as the first publicly acknowledged cyber incidents to result in power outages.2
The outage itself was not caused by an exotic exploit. The attackers logged in with legitimate credentials, crossed from the business network into the control network over existing remote access, and operated the control systems the way an operator would.1 For anyone who designs or runs firewalls, that makes this case unusually useful: nearly every step ran through a control that network engineers own.
What happened
According to the US ICS-CERT alert, which is based on interviews with staff at six Ukrainian organizations, remote intrusions at three regional distribution companies (oblenergos) caused unscheduled outages. The attacks on the three companies took place within 30 minutes of each other and affected multiple central and regional facilities.1
The SANS and E-ISAC report gives the view from one of them. Kyivoblenergo reported that, starting at about 3:35 p.m. local time, seven 110 kV and twenty-three 35 kV substations were disconnected for three hours. The outage was first thought to affect about 80,000 customers. It later emerged that three distribution companies had been attacked, bringing the total to roughly 225,000.2
Power was restored within several hours, with operators switching to manual mode to do it. Recovery of the control systems took far longer: ICS-CERT noted that the affected companies were still running their distribution systems in an operationally constrained mode when it reported two months later.1 2
The authors of the SANS and E-ISAC report are careful about scale. Measured against the whole Ukrainian power system, they rate the impact as low: a small share of consumers, for a limited time. For the companies involved, it was critical.2
How the attack worked
The public reporting describes two stages: a long, quiet intrusion into the business networks, then a short, coordinated attack on operations.
Stage one: getting in and staying in
- Spear phishing. Staff on the business networks received Microsoft Office documents carrying BlackEnergy 3. Opening the document prompted the user to enable macros, and doing so installed the malware. No software vulnerability was exploited.2
- Credential theft. With a foothold on the business network, the attackers harvested credentials and information. ICS-CERT concluded that the attackers had acquired legitimate credentials before the attack in order to use the companies' remote access.1
- Reconnaissance. The SANS and E-ISAC report judges that the attackers had the opportunity to persist in the environment for six months or more, learning how the networks and the distribution systems worked. It calls that patience, rather than any single tool, their strongest capability.2
Stage two: the afternoon of 23 December
- Into the control network over VPN. The attackers used virtual private network connections to enter the ICS network.2
- Opening the breakers. Multiple people, working remotely, operated the breakers using either remote administration tools already present at operating-system level or remote ICS client software.1
- Slowing the recovery. Several further actions were aimed at making restoration harder:1 2
- Firmware on serial-to-Ethernet devices at substations was corrupted, leaving the devices inoperable.
- KillDisk malware erased files and corrupted the master boot record on targeted systems, including, in at least one case, Windows-based operator interfaces embedded in remote terminal units.
- Uninterruptible power supplies for servers were scheduled to disconnect through their remote management interfaces.
- At one company, a flood of thousands of telephone calls kept customers from reaching the call center to report the outage.
What the malware did not do
Early coverage suggested that BlackEnergy or KillDisk had switched the power off. The SANS and E-ISAC analysis rejects that reading directly:
… neither BlackEnergy 3 nor KillDisk contained the required components to cause the outage. The outages were caused by the use of the control systems and their software through direct interaction by the adversary.SANS ICS and E-ISAC, Defense Use Case, March 2016
BlackEnergy enabled the initial access. KillDisk delayed the recovery. The outage came from people using the operators' own tools with the operators' own access. ICS-CERT, for its part, stated that the role of BlackEnergy was still being evaluated and that it had not confirmed a causal link, although the malware was reported on each company's network.1
That distinction matters for defenders. An organization that had detected and blocked every malware sample would still have been exposed if the stolen credentials and the remote access path remained valid.
The network security view
Read as a network engineer, the public reports describe a small number of access paths that should not have been open, or should have been watched. The table pairs each reported observation with the control that addresses it. The first column comes from the sources. The other two are our own analysis.
| What was reported | Why it mattered | Control that addresses it |
|---|---|---|
| VPNs from the business network into the ICS network appear to have lacked two-factor authentication 2 | A stolen password was a complete key to the control network. | Multi-factor authentication on every path into OT, using factors that cannot be stolen together. |
| The firewall allowed remote administration out of the environment through a remote access capability native to the systems 2 | Built-in remote tools became the attackers' operating console. | Default-deny rules in both directions at the IT/OT boundary, with each permitted flow listed by source, destination and service. |
| Legitimate credentials were used to operate the systems 1 | Activity looked like normal operator work. | Separate accounts and directories for OT, so that a compromise of business credentials does not carry across. |
| No resident capability to continuously monitor the ICS network was apparent, based on media reporting 2 | Months of reconnaissance went unnoticed. | Network security monitoring inside OT. Control networks change little, which makes new sessions and new talkers stand out. |
| Firmware on serial-to-Ethernet devices was overwritten 1 | Field devices could be reached and rewritten from the control network. | A separate, restricted management path for field devices, and firmware updates allowed only from named maintenance hosts. |
| UPS units were shut down through their remote management interfaces 1 | Supporting infrastructure was reachable like any other host. | Management interfaces for power, cooling and out-of-band access placed in their own segment. |
| KillDisk wiped workstations and servers 1 | Operators lost their tools at the moment they needed them. | Offline backups, rehearsed restoration, and a tested plan for operating manually. |
A checklist for engineers
These points draw on the mitigations ICS-CERT published with its alert1 3 4, restated as questions you can put to your own network.
- Can you list every path from IT into OT? Not the intended ones: the ones the rule base actually permits. If the list is long or uncertain, start there.
- Does any remote session reach the control network directly? ICS-CERT recommends isolating ICS networks from untrusted networks and, where two-way communication is necessary, using a single open port over a restricted network path. A VPN that lands a user inside the control network is not that.
- Is remote access switched on only when someone needs it? ICS-CERT's wording is that remote access should be operator controlled, time limited, and procedurally similar to "lock out, tag out". Persistent vendor connections into the control network should not be allowed.
- Would stolen business credentials work in OT? If the same account, or the same directory, opens both, the boundary is weaker than the firewall suggests.
- Could "read only" be bypassed? ICS-CERT advises enforcing monitoring-only access with a one-way device rather than relying on software settings or permissions.
- What can execute on an HMI? Operator stations and servers change rarely, which makes them good candidates for application allow-listing.
- Who would notice a new remote session at 3:35 p.m.? Monitoring that nobody reads does not count. Decide who looks, and what they do next.
- Can you run without the control system? ICS-CERT recommends contingency plans that assume the ICS is actively working against safe operation. The Ukrainian operators restored power by hand. That option has to be kept alive on purpose.
Limits of the public record
Anyone using this case should know how much of it rests on reporting rather than evidence.
- The ICS-CERT account comes from interviews. The team stated that it was not able to independently review technical evidence of the attack, although the accounts it heard were consistent across the organizations it visited.1
- The SANS and E-ISAC report consolidates open-source information and the authors' own analysis, and includes an appendix rating the credibility of what was publicly available. Some of its observations, such as the absence of network monitoring, are explicitly based on media reporting.2
- On responsibility: the SANS and E-ISAC report does not address attribution. The ICS-CERT alert was revised in July 2021 to attribute the activity to Russian nation-state cyber actors.1
- Three other organizations, some from other critical infrastructure sectors, were also intruded upon but saw no operational impact.1 Little is public about why.
Sources
- ICS-CERT (now CISA). Cyber-Attack Against Ukrainian Critical Infrastructure, alert IR-ALERT-H-16-056-01. Released 25 February 2016, last revised 20 July 2021. cisa.gov/news-events/ics-alerts/ir-alert-h-16-056-01
- R. M. Lee, M. J. Assante and T. Conway. Analysis of the Cyber Attack on the Ukrainian Power Grid: Defense Use Case. SANS ICS and the Electricity Information Sharing and Analysis Center (E-ISAC), TLP: White, 18 March 2016.
- NCCIC / ICS-CERT. Seven Steps to Effectively Defend Industrial Control Systems. cisa.gov (PDF)
- NCCIC / ICS-CERT. Improving Industrial Control Systems Cybersecurity with Defense-in-Depth Strategies, 2016. cisa.gov (PDF)
About this briefing. It is compiled entirely from the public sources listed above and contains no non-public information. No testing was carried out against any real system. The analysis and the checklist are i4secure's own and are offered for defensive and educational use. Corrections are welcome.