Researchers create 'self-aware' algorithm to ward off hacking attempts
Published at : October 15, 2021
Researchers create 'self-aware' algorithm to ward off hacking attempts
Original article: https://techxplore.com/news/2021-10-self-aware-algorithm-ward-hacking.html
It sounds like a scene from a spy thriller. An attacker gets through the IT defenses of a nuclear power plant and feeds it fake, realistic data, tricking its computer systems and personnel into thinking operations are normal. The attacker then disrupts the function of key plant machinery, causing it to misperform or break down. By the time system operators realize they've been duped, it's too late, with catastrophic results.
The scenario isn't fictional; it happened in 2010, when the Stuxnet virus was used to damage nuclear centrifuges in Iran. And as ransomware and other cyberattacks around the world increase, system operators worry more about these sophisticated "false data injection" strikes. In the wrong hands, the computer models and data analytics—based on artificial intelligence—that ensure smooth operation of today's electric grids, manufacturing facilities, and power plants could be turned against themselves.
Purdue University's Hany Abdel-Khalik has come up with a powerful response: To make the computer models that run these cyberphysical systems both self-aware and self-healing. Using the background noise within these systems' data streams, Abdel-Khalik and his students embed invisible, ever-changing, one-time-use signals that turn passive components into active watchers. Even if an attacker is armed with a perfect duplicate of a system's model, any attempt to introduce falsified data will be immediately detected and rejected by the system itself, requiring no human response.
"We call it covert cognizance," said Abdel-Khalik, an associate professor of nuclear engineering and researcher with Purdue's Center for Education and Research in Information Assurance and Security (CERIAS). "Imagine having a bunch of bees hovering around you. Once you move a little bit, the whole network of bees responds, so it has that butterfly effect. Here, if someone sticks their finger in the data, the whole system will know that there was an intrusion, and it will be able to correct the modified data."
"Traditionally, your defense is as good as your knowledge of the model. If they know your model pretty well, then your defense can be breached," said Yeni Li, a recent graduate from the group, whose Ph.D. research focused on the detection of such attacks using model-based methods.
Abdel-Khalik said, "Any type of system right now that is based on the control looking at information and making a decision is vulnerable to these types of attacks. If you have access to the data, and then you change the information, then whoever's making the decision is going to be basing their decision on fake data."
To thwart this strategy, Abdel-Khalik and Arvind Sundaram, a third-year graduate student in nuclear engineering, found a way to hide signals in the unobservable "noise space" of the system. Control models juggle thousands of different data variables, but only a fraction of them are actually used in the core calculations that affect the model's outputs and predictions. By slightly altering these nonessential variables, their algorithm produces a signal so that individual components of a system can verify the authenticity of the data coming in and react accordingly.
"When you have components that are loosely coupled with each other, the system really isn't aware of the other components or even of itself," Sundaram said. "It just responds to its inputs. When you're making it self-aware, you build an anomaly detection model within itself. If something is wrong, it needs to not just detect that, but also operate in a way that doesn't respect the malicious input that's come in."
For added security, these signals are generated by the random noise of the system hardware, for example, fluctuations in temperature or power consumption. An attacker holding a digital twin of a facility's model could not anticipate or re-create these perpetually shifting data signatures, and even someone with internal access would not be able to crack the code.
"Anytime you develop a security solution, you can trust it, but you still have to give somebody the keys," Abdel-Khalik said. "If that person turns on you, then all bets are off. Here, we're saying that the added perturbations are based on the noise of the system itself. So there's no way I would know what the noise of the system is, even as an insider. It's being recorded automatically and added to the signal."
Original article: https://techxplore.com/news/2021-10-self-aware-algorithm-ward-hacking.html
It sounds like a scene from a spy thriller. An attacker gets through the IT defenses of a nuclear power plant and feeds it fake, realistic data, tricking its computer systems and personnel into thinking operations are normal. The attacker then disrupts the function of key plant machinery, causing it to misperform or break down. By the time system operators realize they've been duped, it's too late, with catastrophic results.
The scenario isn't fictional; it happened in 2010, when the Stuxnet virus was used to damage nuclear centrifuges in Iran. And as ransomware and other cyberattacks around the world increase, system operators worry more about these sophisticated "false data injection" strikes. In the wrong hands, the computer models and data analytics—based on artificial intelligence—that ensure smooth operation of today's electric grids, manufacturing facilities, and power plants could be turned against themselves.
Purdue University's Hany Abdel-Khalik has come up with a powerful response: To make the computer models that run these cyberphysical systems both self-aware and self-healing. Using the background noise within these systems' data streams, Abdel-Khalik and his students embed invisible, ever-changing, one-time-use signals that turn passive components into active watchers. Even if an attacker is armed with a perfect duplicate of a system's model, any attempt to introduce falsified data will be immediately detected and rejected by the system itself, requiring no human response.
"We call it covert cognizance," said Abdel-Khalik, an associate professor of nuclear engineering and researcher with Purdue's Center for Education and Research in Information Assurance and Security (CERIAS). "Imagine having a bunch of bees hovering around you. Once you move a little bit, the whole network of bees responds, so it has that butterfly effect. Here, if someone sticks their finger in the data, the whole system will know that there was an intrusion, and it will be able to correct the modified data."
"Traditionally, your defense is as good as your knowledge of the model. If they know your model pretty well, then your defense can be breached," said Yeni Li, a recent graduate from the group, whose Ph.D. research focused on the detection of such attacks using model-based methods.
Abdel-Khalik said, "Any type of system right now that is based on the control looking at information and making a decision is vulnerable to these types of attacks. If you have access to the data, and then you change the information, then whoever's making the decision is going to be basing their decision on fake data."
To thwart this strategy, Abdel-Khalik and Arvind Sundaram, a third-year graduate student in nuclear engineering, found a way to hide signals in the unobservable "noise space" of the system. Control models juggle thousands of different data variables, but only a fraction of them are actually used in the core calculations that affect the model's outputs and predictions. By slightly altering these nonessential variables, their algorithm produces a signal so that individual components of a system can verify the authenticity of the data coming in and react accordingly.
"When you have components that are loosely coupled with each other, the system really isn't aware of the other components or even of itself," Sundaram said. "It just responds to its inputs. When you're making it self-aware, you build an anomaly detection model within itself. If something is wrong, it needs to not just detect that, but also operate in a way that doesn't respect the malicious input that's come in."
For added security, these signals are generated by the random noise of the system hardware, for example, fluctuations in temperature or power consumption. An attacker holding a digital twin of a facility's model could not anticipate or re-create these perpetually shifting data signatures, and even someone with internal access would not be able to crack the code.
"Anytime you develop a security solution, you can trust it, but you still have to give somebody the keys," Abdel-Khalik said. "If that person turns on you, then all bets are off. Here, we're saying that the added perturbations are based on the noise of the system itself. So there's no way I would know what the noise of the system is, even as an insider. It's being recorded automatically and added to the signal."
cybersecuritycyber securityransomware