Researchers use sound waves to crash drones by confusing sensors. APM vs MultiWii. You'll never guess which one goes down ;-)

A very interesting paper entitled Rocking Drones with Intentional Sound Noise of Gyroscopic Sensors.  A MultiWii and APM drone were tested. One went down (see below for answer as to which):


Sensing and actuation systems contain sensors to observe the environment and actuators to influence it. However, these sensors can be tricked by maliciously fabricated physical properties. In this paper, we investigated whether an adversary could incapacitate drones equipped with Micro-Electro-Mechanical Systems (MEMS) gyroscopes using intentional sound noise. While MEMS gyroscopes are known to have resonant frequencies that degrade their accuracy, it is not known whether this property can be exploited maliciously to disrupt the operation of drones. We first tested 15 kinds of MEMS gyroscopes against sound noise and discovered the resonant frequencies of seven MEMS gyroscopes by scanning the frequencies under 30 kHz using a consumer-grade speaker. The standard deviation of the resonant output from those gyroscopes was dozens of times larger than that of the normal output. After analyzing a target drone’s flight control system, we performed real-world experiments and a software simulation to verify the effect of the crafted gyroscope output. Our real-world experiments showed that in all 20 trials, one of two target drones equipped with vulnerable gyroscopes lost control and crashed shortly after we started our attack. A few interesting applications and countermeasures are discussed at the conclusion of this paper.


Views: 3190

Comment by Mike T on August 16, 2015 at 10:22pm

hahahahaaha love the click bait!

Comment by Hein du Plessis on August 16, 2015 at 10:30pm
Could this be because of the filter in the apm software?

Comment by RC on August 16, 2015 at 10:31pm

8200 Hz at 97 dB? You will at least hear where it is coming from.

Here is a picture of a MDTU (Mobile Drone Takedown Unit) :)

Comment by Rob_Lefebvre on August 17, 2015 at 4:31am

Very interesting.  I was going to assume the double-IMU with EKF did it's job, then realized that they tested an APM, so none of that applies.  I wonder then, why the difference?

Comment by Roberto Navoni on August 17, 2015 at 5:48am

This is the reason because on VR Brain 5 i deprecated the use of  

  • ST Micro LSM333D 14 bit Gyro + accelerometer / magnetometer

In the old VR Brain 4.0 i used it but after a lot of flight test i decided to don't use inside my design. I don't understand why in PX4 and Pixhawk decide to use it . 

But all the ST series had this kind of problem ... infact at the begin of the Pixhawk project need to add a MPU6000 for have better INS performances . 

@Robert i agree with you the EKF cannot work fine with sensor with this kind of problem that mean that on px4 or pixhawk a strange vibration or sound noise can change the attitude estimation ...

Comment by Ben on August 17, 2015 at 6:05am

Interesting article, but the sound level required is quite high (97 dB), stronger than a lawn mower at 1 meter :

If the targeted drone is at 128m, a sound source of 97 + (7*6) dB = 139dB will be required, which is more than a jet taking off at 100 m :

So we could say the sound power level required is superior to a jet taking off, and this is only for a distance of 100 meter. If the target is farther, higher power will be required. Of course all people around the sound source will have to wear good ear protections. Well, I think the operator will become crazy or deaf or both in a matter of seconds.

In addition it isn't very difficult to add a layer of acoustic isolation around the gyro sensor (or the whole board). I wonder if there will be any real use of this concept.

What's interesting is that an original weak point is identified.

Comment by James Betker on August 17, 2015 at 10:41am

The article mentions that there are several COTS solutions available which would be able to convey the attack to up to 100m. These are directional horns so the environmental noise would not be as bad as you are making it out to be.

Also, the article mentions that the best attack vector for most of the tested accelerometers are in the ultrasound range, thus humans would not be able to hear it.

Comment by Quadzimodo on August 17, 2015 at 1:41pm

RC - The MDTU you pictured above is no good for the task. While that thing is probably good for about 135db at 1-metre, the peak would be somewhere down near the tuning frequency of those enormous enclosures (~50hz). All that cone area is useless at >8kHz.

Thinking out loud...

A better approach might be to purchase one of these (or similar) and load it with a straight bit of pipe instead of a horn. For example - SPL of upwards of 160dB is possible ~8.2kHz @ 1-metre from these things.  Loosing approximately 6dB each doubling of distance between the source and the target would see the SPL drop to something like 112dB @ 256-metres.

For a more serious effort 8, 16, 32 or more of these compression drivers could be configured as a focused line array type arrangement.  Ideally, the array would be dynamic so allow the focus point of the drivers to be altered based on the distance to the target... however sound is very difficult to focus accurately so a dish like arrangement (at least 8 drivers across) statically focused around the ideal interception range would probably still be effective.

A line array has major advantages in projecting sound over long distances and focusing the line array further maximises efficiency. A simple line array will loose just 3dB each doubling of the distance from the source to the target (up to certain limits)... A focused line array would be even better. For example - 160dB at the source theoretically becomes something like 133dB @ 512-metres.  This can be extrapolated further... with 121dB @ 8.2km or 97dB @ 21km... but these numbers would not be achievable in the real world because the effectiveness of a line array drops rapidly with distance and it tends to start behaving more like a point source.  But, of course, a larger array would be far harder to focus compared to a relatively small and simple point source... so the disadvantages of a line array almost certainly outweigh the advantages when compared to a simple point source for use as an anti-drone measure.

Comment by Ben on August 17, 2015 at 2:48pm

The article says that the device they could use for 100 meter range is this one :

Have a look a the photo, it's *really* huge.

It looks non portable, and this is a directionnal source. If the drone is moving, which it likely is, such a large device is unpractical, and requires some kind of aiming (even if done electronically by chaning the phase of the sound sources)

With their experimental setup (loudspeaker), they mention sub meter distance at maximum volume.

I wonder if something can be done by software : it looks like the attack is clearly visible as the "noise" on the gyros is visible, could the APM/Pixhawk enter a special mode where it would not use the gyros axis that are jammed ? (because it seems that not all axis are resonnant at the same frequency)

Also we can assume that the attacker will have to try different frequencies if he doesn't know the exact model of gyro used.

And let's say only use the accelerometer, I know it's very difficult (or impossible) for a quad to be stabilized only with an accelerometer, but for a plane it should be possible ?

I remember the Dragon OSD / autopilot that had no gyroscopes and only accelerometers, it was still OK for planes : it was programmed for very slow movements, so that the acceleration during turns was small in comparison to gravity.

Comment by Gary McCray on August 17, 2015 at 3:14pm

The French are noted for and I believe the US military has experimented with phased array focused infrasonic sound projectors.

They do take a lot of energy to operate and they are large, but you could probably come up with a directable one that you could use to protect specific facilities and they would probably induce sufficient vibration to seriously confuse most drones.

I don't think even optical flow would be immune because the image would be bouncing all over the place.

That said the high intensity sound may be a lot worse danger than the "drone".




You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2020   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service