Home

Embeddable algorithm detects cyber attacks at sensor level

Feb. 25, 2019

Mitsubishi Electric develops algorithm to detect measurement inconsistencies

If there’s one thing I’ve learned since I came on board at ControlGlobal.com, it’s been from Joe Weiss. After almost a year of reading his Unfettered blog, the one thing I’m sure of in process automation is that cybersecurity should occur at all levels, especially at the sensor level.

Taking initiative on the matter and seeing the potential of the internal calculations of sensor fusion algorithms, Mitsubishi Electric Corp. recently announced that it’s developed an algorithm which can detect malicious attacks against equipment sensors. According to a press release, the algorithm is embedded in sensor fusion algorithms and detects inconsistencies in sensor measurements that are consistent with malicious cyber attacks.

The new algorithm detects malicious attacks based on more than 42% inconsistencies in measurement data, the press release reports. The algorithm can be implemented as additional software in existing sensor signal processing circuits. It doesn’t require the addition or modification of hardware, and does not compromise sensor accuracy, it adds.

Because the algorithm exploits intermediate values from the sensor fusion algorithm, it doesn’t compromise computation speed, it says.

Source: Mitsubishi Electric Corp.

Taking it further, the company developed an advanced evaluation environment to apply abnormal signals to each individual sensor and to multiple sensors simultaneously. This proved to the company that there are significant differences between fluctuations caused by natural phenomena and those caused by malicious attacks, it says.

Mitsubishi plans to continue development on the project and aims to commercialize the technology starting in 2020, according to the press release.

Let’s hope this is just the beginning to solving this problem, which Weiss has so passionately written and spoken about.