I came across an example of leveraging AI for cyberattacks in a research paper(https://lnkd.in/gBhwgAsF). While this is focused on grids, this use case is a scary example of how multiple AI methodologies can be combined to create cyberattack approaches that go beyond sophisticated.

This is why I keep saying that the country that will be able to build superintelligence first will have the capability to cripple its competitors at will. The type of approach suggested in this paper seems sophisticated, but is still rigid. One of the capabilities of a superintelligent algorithm will be to devise much more powerful cyberattacks in nanoseconds.

Anyways, coming back to this paper, the authors propose a novel attack framework called CDB-TAS (Cyberattack Design Based on CNN-DQN-Blockchain Technology for Targeted Adaptive Strategy. There are three main components of the suggested solution:

1. Three-stage adaptive attack strategy

Preliminary Reconnaissance Phase: A convolutional neural network (CNN) analyses grid/telemetry data in real time to identify the most vulnerable buses (nodes) in the network; small load fluctuations are applied to gather data stealthily.

Escalation Phase: A Double Deep Q-Network (Double DQN) reinforcement-learning agent refines the attack strategy based on feedback (grid responses, demand profiles, detection risk) produced during the reconnaissance phase.

Sustained Attack Phase: The attack is expanded and executed at higher intensity, across more buses/substations, while continuously adapting via feedback loops to maintain stealth and maximize disruption.

2. Adversarial use of blockchain

A private blockchain network is used by the attacker (rather than purely for defense) for obfuscation, anonymized coordination among attack nodes, decentralized logging of attack metadata, and smart contract automation of attack triggers.

This novel combination (CNN + RL + blockchain) from the adversary’s perspective is claimed to be the first of its kind in the power systems / hybrid energy systems domain.

3. Large-scale simulation/case study

The authors simulate their framework on a synthetic 2000-bus hybrid hydrogen-power network modelled after the Electric Reliability Council of Texas (ERCOT) grid. The system comprises traditional generation + renewables (solar, wind) + hydrogen electrolysis/storage/fueling.

A primitive form of the more sophisticated cyberattack methodologies that are coming our way.