Bluetooth has spent much of its life ignobly associated with crummy headphones, byzantine connection procedures, and car stereo systems that never quite seem to work right. Now this wireless technology concocted in the ’90s to help PCs and mobile phones communicate is being asked to step up and save the planet from a global pandemic. According to its two co-inventors, there could be some issues.
Named for the 10th century king Harald “Bluetooth” Gormsson, famous in Scandinavia for uniting (and Christianizing) the Danes, the humble, oft-derided wireless technology included in some form in nearly every portable device from the past decade and beyond is central to coronavirus contact tracing apps pushed by Apple, Google, and governments across the world. Banking on the standard’s ubiquity, and considerably improved reliability since the ’90s, these entities hope to turn billions of Bluetooth-enabled devices into an army of public health automatons that can map anyone who came into contact with someone who tests positive for Covid-19.
Although the exact plans for using Bluetooth vary between governments, the gist is simple: In order for your iPhone to connect to your friend’s Bluetooth speaker, it has to essentially shout its existence into the electromagnetic spectrum, sending repeated radio messages that announce that the device is turned on and willing to pair with another. It’s exactly these short, repeating radio wave bursts that tech companies and public health authorities hope can be used for contact tracing, by collecting an anonymized record of every Bluetooth announcement within a certain range. If one of these “HELLO, I AM BLUETOOTH!” messages ends up coming from an individual who later tests positive for Covid-19, the hope is that anyone else whose phone was able to detect that message could then be alerted and tested (or treated) accordingly. It’s a fine, if somewhat ambitious, plan on paper, and the Apple/Google partnership is already drawing generally positive commentary from the privacy and civil liberties-minded set for its efforts to mask the identity of a Bluetooth device’s owner.
But like many great paper plans, the fundamental nature of the universe presents some complications. Underneath its blue runic branding, Bluetooth is just radio waves, the same chunk of the electromagnetic spectrum that allows for FM radio, Wi-Fi, and cell phones. And if you’ve ever used of these things, you know that getting a reliable signal can be a challenge even under seemingly ideal conditions. That’s because radio waves don’t just teleport from transmitter to receiver, but get bonked around and absorbed by objects in the way: Trees, houses, dogs and cats, cars, brick walls, and other human beings all absorb or reflect a Bluetooth signal, affecting the strength of that signal when it reaches its destination or stopping it from getting there entirely.
Given that Google and Apple’s Bluetooth contact tracing relies on using the strength of a received signal (Received Signal Strength Indication, or RSSI) to determine whether you were within coughing distance of a Covid-19 patient that time you walked to the grocery store, this could be a serious problem. Swarun Kumar, a professor of electrical and computer engineering at Carnegie Mellon University, recently estimated (see video below) that environmental factors could make a Bluetooth device that’s 2 meters away appear to another device as if it’s 20 meters away, or vice versa.
I asked Jaap Haartsen and Sven Mattisson, who invented Bluetooth together while working at the Swedish telecom Ericsson, whether their tech was up to the task. While both expressed hope that Bluetooth could be used to save lives, they also noted that problems of accuracy are very real. Bluetooth Low Energy, or BLE, a more modern iteration of the tech that’s a great deal more efficient and accurate, and which is used in the Apple/Google contact tracing system, is still subject to finicky physics.
“One issue that comes to mind at the physical layer is the uncertainty in the detection range,” explained Mattisson, “i.e. how well can you assess the distance to another BLE device. … The radio signal’s path loss will vary significantly depending on conditions (i.e. free space or obscured). Whether knowing the distance only within a factor of, say, ten is a problem or not, depends on the application. E.g. do we get too many or too few contacts…? Temporal aspects can be used to reduce this uncertainty but it is easy to forget about the coarse range resolution of a single radio link.”
“It is easy to forget about the coarse range resolution of a single radio link.”
In other words, Bluetooth’s inaccuracy may not be a problem for typical applications (connecting to that friend’s speaker, for example), but certainly could be for purposes of determining exposure to a pathogen. Mattisson, who has read through the Bluetooth-specific portion of the Apple/Google plan, also pointed out that the “way many BLE devices are built, the RSSI value can be rather crude and not well calibrated.”
Mattisson echoed Kumar’s concerns with Bluetooth signals colliding with walls and being absorbed by your pants: “In addition to this RSSI uncertainty, you will have the path loss variation. If there is an obstacle, like a human body, between your BLE device and the detected one, there may be some 70dB of attenuation of the signal at 1m (3ft; e.g. in a crowd) but in free space this may well correspond to 10m (30ft). Exact numbers vary with objects, signal reflections etc., so it is hard to give anything but rough numbers.” Mattisson added that without tapping other location pinpointing technologies — which privacy advocates say could provide governments and private companies with overly broad maps of our daily movements — false positives and false negatives will be unavoidable.
“My main concern was that many (non-radio specialists) are unaware of the large variability in signal strength vs. contact distance. Since the contacts reported include RSSI, app developers can do something about it. For example, one can track a contact over (a limited) time and get many RSSI readings, giving a better estimate of the range… Further, using the knowledge that a contact may be close, or at some distance, can also be factored in. All these aspects must be considered by the app developers. If you want accurate ranging, you need triangularization or timing measurements (like a radar). The first requires multiple units cooperating (like what is done in cellular networks to locate mobiles dialing e.g. 911) and the latter dedicated hardware. Neither is this is available with vanilla point-to-point Bluetooth links. So, yes, there may be false negatives and false positives and those have to be accounted for.”
“Many are unaware of the large variability in signal strength vs. contact distance.”
Haartsen, Mattisson’s co-inventor, shared his concerns over radio wave wonkiness, but also added that the whole issue is moot unless people end up using the app in great numbers:
“All around the world, even in my country, groups are working on smartphone apps for helping us to get back to normal life after the corona lockdown. Bluetooth is only part of the big puzzle. Yes, Bluetooth could keep track which other phones have been in range, however the accuracy is not great (yet). But at least you do not need to use location data like GPS which would be sensitive to privacy issues. Privacy and data security are very high on the agenda and need to be solved before people (at least in my country) will use it. And of course the sheer number of people that will activate the app. If it is only scarcely used, it will be useless.”
During a joint press briefing, an Apple and Google representative told The Intercept that a recent update to their contact tracing specification called for embedding a tiny fragment of data inside each Bluetooth broadcast that states the strength with which the signal was transmitted, allowing the receiving device to better determine how much of that signal was lost in transit, and better estimate its distance. The companies believe this process will be refined as it is used.
Bluetooth’s reliability problems have never been more than an annoyance because the stakes have been about as low as can be imagined. Not so with contact tracing, where a false positive could mean a healthy person unduly treated as infected or exposed, with all the anxiety and restrictions that might bring, while a false negative means a person who might need live-saving viral screening could go about their routine as usual, instilled with a false sense of Bluetooth-enabled confidence.
Jason Bay, the Singaporean government official who helped oversee that country’s early app-based contact tracing efforts, has echoed public health authorities around the world who say an army of human contact tracers will be more efficacious than any technology. He wrote on Medium, “If you ask me whether any Bluetooth contact tracing system deployed or under development, anywhere in the world, is ready to replace manual contact tracing, I will say without qualification that the answer is, no.”
Daniel Kahn Gillmor, a senior technologist at the American Civil Liberties Union, told The Intercept he agrees that erratic Bluetooth signals could cause more problems than good: “If a [contact tracing] scheme has a high false-positive rate, especially for a high-prevalence disease, you might as well just alert everyone. No need to build a novel in good public messaging to achieve the same effect. And the stress of getting lots of apparently individualized alerts that turn out to be false can contribute to warning fatigue and general dismissal of the messages.”
Perhaps Bluetooth-centric contact tracing will only turn out to be merely useless as opposed to actively harmful. Maybe the Apple/Google tag-team approach will be universally adopted and responsibly implemented by every locale in the country. “If Bluetooth can play a role in stopping something like the present coronavirus pandemic, that is of course extremely gratifying,” Mattisson told The Intercept. “But, it is smart use of a good technology, not just the technology itself that may see this happen.” The trouble with looking to the technology sector to save the world is that their solution will, of course, be more technology. Even if that technology has trouble getting through bushes and trees.