A team from the University of British Columbia (UBC, Canada) has found a vulnerability that allows attackers to disrupt water treatment plants and power grids without penetrating a facility's internal systems or breaking encryption.
The attack targets facilities that exchange data with cloud-based monitoring and control systems over encrypted channels. First, the attacker needs to gain control of the router through which traffic passes between the facility and the cloud.
The attacker then analyzes metadata - information about message timing and size, not their contents. These patterns can predict when critical commands are likely to be transmitted, even if the messages themselves remain unreadable. Based on this analysis, the attack selectively blocks a small share of packets.
Industrial control systems rely on computers that control physical equipment - pumps, valves, generators. Unlike attacks that require access to equipment inside a facility or network flooding, the new method uses patterns in encrypted data exchange to anticipate critical moments and intervene at those points. According to the authors, this makes it possible to disrupt operations without entering the facility's own systems and to evade the attack detectors they tested.
"What makes this attack alarming is that the attacker does not need to read the messages," explained Dr. Aastha Mehta, a professor of computer science at UBC and co-investigator on the project. "Even when the content is encrypted, patterns in communication can reveal enough information to identify and attack critical messages."
The work "ICS-Sniper: A Targeted Blackhole Attack on Encrypted ICS Traffic" has been accepted for presentation at the Network and Distributed System Security Symposium 2027 in Seoul, South Korea. The researchers call their attack ICS-Sniper, by analogy with a sniper who selects targets precisely and leaves minimal traces.
In laboratory tests, the team demonstrated it on a cloud version of a water treatment testbed at the Singapore University of Technology and Design. Two models of the testbed were used: one with software controllers and one with standard industrial equipment; both reproduced physical processes. In one test, the attack delayed the start of clean water pumping by about 15 minutes. In another, a simulated valve remained open longer than it should have - a failure that at a real facility could lead to the loss of treated water.
"An attacker does not necessarily need to break encryption to cause a failure," noted Dr. Karthik Pattabiraman, a professor at UBC and co-investigator on the project. "Simply by observing traffic patterns, they can determine when critical messages are likely to be sent and selectively block them."
Water treatment was chosen because publicly available schemes made it possible to reproduce the facility's control systems. The traffic analysis method was also tested on several industrial communication protocols, including a power grid dataset, but the disruptions were demonstrated specifically on the water treatment models.
The team is already working on defenses. One approach is to make network communications less predictable, for example by changing message sizes or timing so that an attacker has more difficulty distinguishing critical commands from ordinary traffic. Such changes must preserve industrial equipment's latency requirements. Another approach is to make control more resilient to network failures, including through alternative routes for critical messages. The researchers have also published code, testbed reproduction instructions and traffic data so that other specialists can verify the results and develop defenses.