|
Did you receive this newsletter as a forward? Subscribe here
|
|
| How can we make TinyML secure? (And why we need to) |
| |
| By Stacey Higginbotham |
Lately, I've been reporting more and more on algorithms that parse incoming data locally in order to make some sort of decision. Last week, for example, I wrote about a company doing voice analysis locally to detect Alzheimer's. I've also covered startups that process machine vibrations locally to detect equipment failures. Each of these examples benefits from machine learning (ML) algorithms that run on microcontrollers, or what we call TinyML.
Running machine learning algorithms locally helps reduce latency. That means a burgeoning machine problem can be detected and the machine turned off quickly if needed. Running machine learning algorithms locally also protects privacy, something especially important in the medical sector. Indeed, I would prefer that neither Google nor Alexa be aware if I develop Alzheimer's.
But as companies push sensitive and necessary algorithms out to the edge, ensuring they perform as intended becomes essential. Which is why I spent time this week learning about the security risks facing our future sensor networks running TinyML. |
|
| — The three branches of AI, based on the amount of computing and power available. Image courtesy of Prof. Dr. Muhammad Shafique, Department of Electrical and Computer Engineering at New York University Abu Dhabi. |
|
|
On Tuesday, during an online presentation hosted by the Tiny ML Foundation, Prof. Dr. Muhammad Shafique, Department of Electrical and Computer Engineering New York University Abu Dhabi, covered ways to design both hardware and algorithms to securely run on constrained silicon at low power. It covered TinyML and also edge ML where machine learning algorithms run on more robust hardware such as smartphones, gateways, or even a car.
Dr. Shafique highlighted two types of security. The first security focus is on ways to prevent adversaries from "tricking" a sensor and its algorithm. One way to do this is by overwhelming the sensor with "noise." For example, it's possible to trick the algorithms governing lane changes in a self-driving car simply by placing black squares on the pavement. The squares introduce "noise" that can confuse the algorithm, leading the car astray.
Hackers have already demonstrated the use of "noise" to confuse other types of sensors and algorithms, such as those in medical devices. Shafique called these "adversarial attacks" and offered different ways to design the ML models that could make them less susceptible to such noise. (This is especially relevant for TinyML because the process of shrinking algorithms to run on less robust chips makes them more vulnerable to any weird outlier data coming in.)
The other type of security is related to the security of the chip running the ML algorithm. This might involve a bad actor getting malware on the chip that could affect it or the software it runs. Or it might simply cause the sensor to produce inaccurate results, which essentially turns the sensor into a node that lies about the state of the thing it's trying to measure. In an industrial system, it might mean the sensor that's monitoring temperature reports an inaccurate value, causing a process to fail, whereas with medical devices it might give false readings that could lead to a delay in diagnoses or even the active infusion of an unnecessary drug.
It is, in other words, super scary stuff. And what makes it even scarier is that we're gradually going to move more and more machine learning to the edge because that's where such processing belongs. As I've noted in previous posts, it simply take too much time, money, and energy to move data to the cloud. Because most of us recognize that edge processing also means data stays local to a computer network and thus, more private — or at least more controlled — we also tend to think of machine learning at the edge or on a sensor as more secure.
But as Dr. Shafique's presentation made clear, that's an erroneous assumption. This is why I recommend that anyone planning to build sensors and systems that rely on edge-based machine learning start thinking now about the security needs of both the hardware and the algorithms that will soon be making more of the decisions in our smarter and more connected world. |
|
|
|
|
| Connecting the Physical and Digital Worlds SPONSORED |
| |
|
As IoT continues to dominate our world, there’s an ever-increasing business need to connect the physical and digital ecosystems with a frictionless user experience.
Rightpoint believes the business opportunity lies in pairing a human-centric approach with cutting-edge connected device solution development, offering clients end-to-end IoT services, from concept to consumer.
Learn more about how leading organizations like General Motors, Chamberlain, Sanofi and more are leveraging IoT to create more meaningful experiences.
|
|
|
|
|
| Konnected embraces crowdfunding for an equity raise |
| |
DIY security company Konnected is in the midst of raising up to $1.07 million on the StartEngine crowdfunding platform to fund its next phase of growth. Unlike the startup's previous crowdfunding campaigns, which were designed to raise money for building new products, this is an equity raise, with supporters getting stock in the six-year-old company.
Konnected founder and CEO Nate Clark said he decided to raise money to help grow the company; he chose crowdfunded equity as a way to avoid having to go to venture capitalists. "I saw firsthand what happens when you have venture funding and lose your mission because of a push to grow the business," he said. "It would be a different vibe and different business model if [Konnected] were led by a VC." |
|
| — A Konnected interface board that ties an old alarm system to the internet. Image courtesy of Konnected. |
|
|
That doesn't mean he'll never seek venture capital, just that based on his active community of users and previous experience with crowdfunding he's excited to sell shares to a broader audience of existing customers and those who search for investments on StartEngine. So far, the funding has netted a bit more than $556,000, with 393 investors participating.
Konnected got its start in 2016 with a crowdfunded campaign to build a board that consumers wire into an existing alarm panel in their home. Both the original and the second-generation Konnected panels let buyers repurpose the existing sensors from an earlier or existing security installation and control it from their smartphone. Konnected subsequently added integrations that tied the older components to new platforms such as Amazon Alexa and Samsung's SmartThings. The third-generation Konnected product added support for Ethernet and Power over Ethernet.
Installing any of the Konnected panels requires a bit technical skill because users have to pull out the existing board inside an alarm panel and rewire the sensors through the Konnected panel. It's on a level of difficulty on par with replacing a thermostat, so non-experts can actually do it. I did.
When I used it, I thought it was an excellent product, both because it used what was already in an average home and because Clark was so clearly trying to meet the needs of the user community. Konnected responds to user feedback with product updates and features, and seems focused on bringing security to a modern era.
And now, like HomeAssistant, Samsung SmartThings, and other old-school DIY smart home platforms, Konnected is trying to boost its offerings to appeal to more mainstream users. With this in mind, the company is building a smart garage door opener and a cloud platform. But to fund these growth plans, Konnected needs more money.
I'm intrigued by Clark's decision to use crowdfunding, for two reasons. The first is that Konnected is one of a few remaining IoT hardware startups that built something necessary and useful for early adopters and is now trying to parlay that into a product for more mainstream customers. I'd love to see it succeed, and the market for connected security products is only expanding.
The second reason I'm intrigued is because I think the wave of VC money that flowed into the smart home space caused problems. The wash of funds resulted in poorly thought-out hardware businesses that strove to grow hardware sales without thinking about the ongoing cost of connected hardware. This led to the creation of subscription business models, which can be a really hard to sell for consumers who just spent money to buy the connected device itself. Pressure from venture investors also drove many startups to sell to larger companies, shrinking the overall number of smart home options consumers might have otherwise had.
Nor is Konnected big enough for a venture capital-style investment. The company reported $1.1 million in revenue for 2020 and a $187,000 profit. While it is raising money, it considers the fundraising as an accelerator to growth based on the fact that it is currently profitable and hopes to continue being profitable. Indeed, Konnected's financials are almost inverse to those of traditional venture-backed startups, which emphasize growth over profitability and spend accordingly.
At least one of Konnected's investors is fine with this. Jason Passwaters, who is also a Konnected customer, invested more than $50,000 in the startup after getting an email about the fundraising campaign. He said that when he looked at Konnected it reminded him of how smart devices such as Ring and the thermostat from Ecobee changed the market. After all, as he noted, "Ring was bought by Amazon for, like, a billion dollars" back in 2018. Last year, Generac purchased Ecobee for $770 million.
However, Passwaters said he isn't "some big investor," and that he invested the money in part because he believed in the tech, liked the overall opportunity, and was impressed with Clark's energy. While he did read the financial documents affiliated with the offering, he seemed happy to participate in the fundraising and figures that if he simply gets his money back that would be "cool."
I really hope it works out for both Clark and Passwaters. And not just because it will benefit Konnected, but because I think this is a model that will benefit some of the smaller, grittier DIY smart home projects out there as well, such as Hubitat or HomeAssistant. |
|
|
|
|
| Cellular IoT SIM cards SPONSORED |
| |
|
Own your IoT deployment with Hologram’s SIM card and device management dashboard. Their SIM grants you access to a global cell network with easy activation, developer-friendly tools, and new connectivity partnerships.
Order a Hologram IoT SIM card today! |
|
|
|
|
| Episode 357: Too many companies are paying ransomware |
| |
| This week we kick off the show with an explainer on Sigfox's receivership before delving into grim news from Claroty, a cybersecurity firm. Claroty surveyed 1,100 IT and OT (operational technology) pros and discovered that three out of five of them worked at companies that paid up after a ransomware attack. I can't believe it, but the research offers compelling reasons why. We then move over to some Matter news, starting with an update from Assa Abloy on backward compatibility for its Yale locks, and concluding with an update from Nanoleaf, which is delaying some of its products thanks to Matter being delayed. We then cover Walmart teaming up with Angi to install some smart home products for consumers as part of the rush to invest in smart devices that we discussed last week with Angi's CEO. In quick news, we talk about a new IoT satellite service, an acquisition by connected health company Withings, and Peloton's new $90 smart heart rate monitor. |
|
| — Withings purchased a subscription workout app to add to its wellness hardware. Image courtesy of Withings. |
|
|
| Our guest this week is Simon van der Jagt, CEO of Nowi. Nowi makes an energy-harvesting chip that is easier to use in a variety of sensors. We talk about why before moving to a discussion of plug and forget devices and what those will entail. The idea is that companies can build energy-harvesting sensors that could last for the life of a road or container or other product, and thus we can forget about them. We also talk about trends in designing chips for the IoT such as sales challenges and the growing demand for highly specific chips that can consume the least amount of power while getting the work done. Enjoy the show. |
|
|
This week on the IoT Podcast Hotline, we hear a good idea from one of our listeners who set up a sort of call system using Wyze outlets and smart bulbs.
The IoT Podcast hotline is brought to you this week by SpinDance. SpinDance can help you design, build and support your IoT product. Learn more at SpinDance.com. |
|
|
|
|
|
| Correction |
| |
| Last week's newsletter erroneously said that Assa Abloy would make both its Yale and August locks backward compatible with the Matter standard. Assa Abloy has only committed to building a module to ensure its Yale Assure locks can be retrofitted for backward compatibility. I regret the error. |
|
|
|
|
|
| News of the Week |
| |
From hot mugs to cold boxes: Ember, the company behind the Ember self-heating coffee mug (the best gift I've ever given!) has released its first B2B product, a refrigerated cube designed for shipping medical items that need to stay cold. Last year, Ember said it would broaden its portfolio of products, and shocked many when it said it would pivot from warm to cold. Sure, you might think that making temperature-controlled enclosures is basically the same, but the mechanisms for keeping things hot and keeping things cold are very different. And with this product, Ember is proving that it can do it, in the process opening up a much broader business opportunity and probably reassuring investors who didn't want to get stuck with a one-hit-wonder product. Ember has signed a partnership with Cardinal Health, which has experience shipping medicine, to prepare for the release of the product later this year. (TechCrunch)
Public opprobrium might work...sometimes: Life360 has said it will stop widely selling location data after a report by The Markup pointed out it was one of the top sellers of location data to data brokers. The news horrified many, especially since Life360 is popular with families trying to keep track of their kids. In response, Life360, which earned $16 million from selling location data in 2020, has scaled its location data sales to only two companies, Allstate Airity and Placer.ai. The Allstate sales are tied to insurance policies that use data to set premiums based on risk, and will only be allowed to use aggregated data going forward. Go journalism! Also, this is why we need far more transparent data policies, even if not everyone cares to read them. (The Verge)
Australia's Telstra Purple acquires two companies to build out IoT capabilities: Telecom carrier Telstra Purple has agreed to purchase Alliance Automation and Aqura Technologies. The terms of the Alliance Automation deal are undisclosed but the seller of Aqura says that deal was valued at $30 million in cash. Alliance Automation makes industrial control systems software and will help boost Telstra's sales efforts with firms requiring digital transformation while Aqura provides consulting and services and will help set up private networks. It's a good time to be in the services business since IoT deployments are still a pretty heavy lift. (ARN)
Keep an eye on this company (and trend): This isn't exactly an IoT story, but I couldn't help but notice this story on Metronome, which has raised $30 million to make usage-based pricing easier to implement. Today it is aimed at helping SaaS companies that typically charge a monthly fee for access to a software service adapt their pricing to usage-based fees, which involves a surprising amount of work. Usage-based pricing already software exists (think telco OSS billing software), but building software that can charge per use and change pricing plans quickly is tough. And yet, it's clear that for both software and the physical world, connectivity and sensors make usage-based pricing viable and likely to happen. (TechCrunch)
Insurers are finally adopting smart home tech to reduce claims: It's taken far longer than I thought it would, but the insurance world is finally starting to ship out connected sensors to policyholders in order to detect leaks and prevent claims. Chubb has apparently embraced both leak and temperature sensors and is offering discounts for customers that deploy them. According to a Chubb executive, residential customers can save about 3% on their premiums with sensors and 8% if they install devices that shut off the water. In the meantime, Chubb has acquired StreamLabs, an Australian sensor company, to control its own destiny when it comes to sensors, a strategy I expect to see from more companies that want to ensure the quality of their data streams. Unfortunately for consumers, this means if you switch insurers you may need a new round of sensors. (Insurance Journal)
Learn about LoRaWAN: In case my overviews of low-power wider area networks are confusing, Kevin has written a LoRaWAN 101 post that provides some basic terminology and information to help decipher what can be a confusing sector. (StaceyonIoT)
Maybe smart home folks need to talk to carmakers about energy resiliency: I've been very excited about how connected appliances and smarter breaker boxes can combine with software to add resiliency when it comes to electricity in homes, and now there's a new player in the form of Ford's new F-150 Lightning truck. Ford is working with Sunrun, a solar panel installer, to install and sell a specially designed charger that can handle bidirectional energy flow and charge the truck. The F-150 extended-range battery system can store 131 kilowatt-hours of electricity which is enough to meet the energy demands of the average home for three days. This is pretty awesome andsomething I'd like to see more of as more people embrace electric vehicles in the home. We already use our Tesla as a giant battery to recharge devices during power outages, but I can't take advantage of its battery to power my coffee maker or a space heater at the moment. I'd love to have that for future vehicles. (Electric Cars Report)
Samsung SDS is out of the commercial lock business: Zigbang, a property tech company in South Korea, will acquire Samsung SDS' home IoT business for an undisclosed amount. Samsung SDS is a managed security service that also provides hardware such as locks and wall switches. The smart home business inside the security business was kind of a weird fit, and Samsung had tried to sell its IoT hardware business to Allegion (the parent company of Schlage locks) back in 2016. So, it has taken a while, but now Zigbang has purchased the business and will likely end up providing services and hardware to multifamily and other commercial housing providers going forward. (Business Standard)
The Thread Group preps to take Thread commercial: The Thread radio standard was developed for connecting IoT devices by companies more commonly associated with the smart home, but the Thread Group has long had commercial ambitions. And it looks like the Thread Group plans to make 2022 the year it shifts to serious commercial consideration. The Thread protocol allows for a distributed, low-power mesh network that can connect directly to the internet without requiring an intermediary as Zigbee and Z-Wave do. This is probably even more useful in commercial buildings where sensors networks are much larger and need to be more robust. The Thread Group is working to help get several existing smart building protocols, such as BACnet and DALI, ready to transition to IP networks that can also directly communicate to the internet. So keep an eye on Thread for commercial deployments to see if it will make it in the wider commercial world. (The Thread Group)
Hi! Want to support this newsletter and my podcast? We have some upcoming advertising opportunities. Request a media kit for more details. Thanks!
|
|
|
|
|
|
|
|
|
|
