Paper Discussion: 3a. The Signpost Platform for City-Scale Sensing

[pcodes]

• @lrshpak Lily Shpak, Comprehensive: Would authors look into passing legislation to clear up privacy policy regarding sensor data collection?
• @lrshpak Lily Shpak, Comprehensive: Will cities need to hire more technically experienced employees to maintain system?
• @albero94 Alvaro Alberto, Critical: Does the given API restrict a developer's control over low-level parts of their application?
• @albero94 Alvaro Alberto, Critical: Is the message queue for the communication options shared between applications?
• @rachellkm Rachell Kim, Critical: Is the energy provided for each module dynamically allocated or equal for each module?
• @nikorev Niko Reveliotis, Comprehensive: How would the authors handle validating modules if they are the central authority?
• @s-hanna15 Sam Hanna, Critical: What are the effects of other types of weather on solar harvesting?
• @s-hanna15 Sam Hanna, Critical: How is onboard storage kept secure?
• @grahamschock Graham Schock, Critical: What mechanical components could be added to keep Signpost installations secure?
• @grahamschock Graham Schock, Critical: How do we authenticate signposts?
• @samfrey99 Sam Frey, Critical: Could communication be improved with the advent of 5G?
• @mjhegarty Michael Hegarty, Comprehensive: Is the inclusion of the Intel Edison worth the additional power, versus handing off the computation to an edge node?
• @mjhegarty Michael Hegarty, Comprehensive: What is the lifetime of a Signpost device?
• @huachuan Huachuan Wang, Comprehensive: is a GPS module on a static device really necessary?
• @hjaensch7 Henry Jaensch, Critical: What are the overheads of providing isolation?
• @rebeccc Becky Shanley, Comprehensive: What can be done to prevent nefarious use of collected data?
• @Others Gregor Peach, Critical: Why can't I read on time?

Shared concerns:

• The cost of the platform seems super high
• There is zero mention of physical security, but it's ok because Berkeley is safe. What are ways to improve the security?
• How are modules kept secure? Does this system allow for untrusted modules to co-exist?

[lrshpak]

Reviewer: Lily Shpak Reviewer Type: Comprehension

Problem Being Solved

The authors of this paper are presenting their idea for Signpost, which is a smart city device. They want to implement these systems in different cities to better tailor the city to the people living in it. They are trying to make it so that the system is all contained in one device, instead of having sensors for each module they want to implement.

Main Contributions

One of the main contributions this paper makes to creating smart cities, is the creation of devices that will contain all sensors for creating small cities. They acknowledge that it is much easier to install these devices in a city if they are contained in one unit.

Three Questions

1. The paper acknowledged that installing these systems would cause significant privacy concerns of citizens of the city they devices are in, would the authors consider trying to get legislation passed to make sure the privacy policy is clearly outlined?
2. The authors say that the system will cost about the same as the general infrastructure in cities already, but will cities have to hire more expensive and technically experienced employees to up keep these systems?
3. How will cities make the transition between their current systems to Signpost without losing any data or power that is necessary to keep the city running?

[albero94]

Reviewer: Alvaro Albero Review Type: Analytical

Problem Being Solved

Deploying sensors at a city scale can provide a great amount of information is be very useful for many different applications. However, the number of sensors that need to be deployed and the places where they are deployed, many times isolated from energy resources and communication networks, make this type of projects very challenging and expensive.

Main Contributions

Signpost is a project that has successfully created a platform to deploy many of these sensors at a city scale. The platform provides the necessary resources for the proper operation of multiple sensors that can be installed in pluggable modules. There is no need of wired networks as there are integrated solar panels deployed where the sensors can harvest energy from. This platform provides an easy way to deploy sensors at a city scale which is be very helpful for developers as they can focus on creating the desired applications instead of solving platform engineering problems.

Questions

• The authors say that the energy for the Linux system and GPS accounts as application energy used and not system overhead, what if none of the current applications use them? Is the Linux system still kept alive? I guess yes in case some other application wants to use it later. How is this energy consumption accounted?
• With the given and limited API for communication and system services will the developers lose some control over low level application development?
• The authors talk about building a community and bringing developers and they have performed different actions to advance in this direction. If a developer wants to develop in Signpost, how hard would it be to migrate an application that is already in production to this platform?

Critiques

• In the multi tenancy section there are no comments about how the memory and processor is shared, just internal communications and energy. Both the Linux OS and the other micro-controllers can be running multiple applications at the same time, there is no detail on how the system handles this situation.
• The security and privacy section is very small and there are no comments about how the system is secured or would react against different attacks. The system is connected to the internet and physically accessible, so implementing security measurements or at least talking about it and mentioning which are the challenges should be a very important point.
• The authors talk about cellular network used to send messages if the message queue is getting to long. However, they do not specify if this is a shared queue or individual by application. In case it is shared, how is that extra energy cost charged? There is no information about cost per message sent or other ways to account this cost.

[rachellkm]

Reviewer: Rachell Kim Review Type: Critical

Problem Being Solved

This paper lists deployment as one of the key reasons for slow progress in applying smart technology to city scale. Because most existing platforms utilize mains power, which requires wiring that significantly limits possible deployment locations, the authors state the need for a wire-free, solar-powered platform that would allow for lightweight and easy deployment for sensors and applications.

Main Contributions

The authors propose a platform called Signpost to enable easier sensor deployment, which in turn would help exploratory technology meant to support the larger goal of enabling smart cities. They have developed a solar-powered platform which comes with a number of useful sensing modules to allow for easy access, maintenance, and updates of smart sensors and the applications that run within these modules.

Questions

1. This paper states that the Signpost platform comes with 8 pluggable slots, but does this mean most of the focus will be on trying to get clients to plug in their own sensors that may also have the option to access the default services on Signpost or to deploy their applications on the platform inside existing sensing modules?
2. Is the amount of energy provided to each module allocated equally or is it dynamic in the sense that it provides more for those that require more power than the rest connected to the backplane? It says energy is never taken to support other applications, but what about modules?

Critique

1. I felt that the claim that their deployments have experienced no vandalism or theft despite placing them near an area with “high property crime” was weak and sort of pointless to the paper because of the initial disclaimer that they have taken zero precautions to prepare for them. I don’t think this proves anything about the physical safety of the systems once it does become public knowledge that they exist, and I just felt that this “Deployment Metrics” section was a bit insufficient overall.
2. Although I did notice the big emphasis on using Signpost for experimentation, I wish they provided more detail on how they envision this would transition into a more permanent solution or what other technical improvements and investments would be needed to benefit this direction of development.

[nikorev]

Reviewer: Niko Reveliotis Review Type: Comprehension

Problem Being Solved

Creating an energy efficient, sustainable solution to collect data in urban areas (weather/air quality/etc). In the intro they describe, "Signpost, a modular, energy-harvesting platform for city-scale sensing." This system mounts onto signposts and allows for 8 modular ports that can be used for multiple uses. The main benefit is it's wireless charging through solar, while allowing for developers to create new install-able modules. Signpost aims to be a self-supplying energy device which can provide valuable sensing data across modules within a city.

Three Questions

1. In section 4.1 they talk about how decreasing energy requirements will allow Signpost to gain greater performance capabilities; specifically how much greater performance would we be considering? The processor would need to be swapped, unless we are strictly speaking of "battery life" in this case. Is this what they were referring to here?

2. Is the cost of implementing Signposts across a city currently feasible? At $2000 a piece they seem great, but do they operate as a service (i.e. charging for technical support/repair)?

3. In the final paragraph they refer to having to validate modules as a central authority. I know this paper is discussing the product they created from a technical standpoint, but is there more information pertaining to them as a company? This is a product that I'd love to see develop.

[s-hanna15]

Reviewer: Sam Hanna Review Type: Critical

Problem Being Solved: This paper talked about urban sensing and the particular technology they were talking about was called Signpost. Signpost is a new method of urban sensing with a focus on making sensing more efficient and applicable for more modules than the current options.

Important Areas: Signpost is an urban sensing technology that’s the main draw is that it uses solar panels as its energy source instead of drawing straight from a city grid, allowing it to be more modular on where it can go. It has many modules allowing different sensors to be added and allows for many different applications to run.

Questions:

1. The paper talks about testing Signpost in different environments including rainy and sunny places. What they never discuss is the effects of other weather events such as wind or snow, so I am curious as to how other inclement weather affects the device?
2. Signpost has multi-tenancy as one of its core design features, and it talks about energy use by each application. One of the examples uses its energy allotment very early then gets turned off. It also talks about how these can be used for key city features. What if one of the key city features application takes up too much of the energy and thus must get turned off?
3. They mention the storage being on the device itself, would this lead to security concerns since the device doesn’t seem to have any physical security?

Critiques:

1. This paper has a section about security, and it mentions that they feel like policy and practice are both necessary, but it doesn’t go into any detail about what needs to be implemented. They seem to believe the only worry is in Personally Identifiable Information (PII), which is always a privacy concern, but if city features will be based on this then it is important that the information be correct and secure.
2. During the Deployment Metrics section, they seem to make a lot of assumptions about the platform being unobtrusive because of a lack of vandalism and theft over a short period of time in one particular area.
3. They talk about some limited applications and the need for this type of product, but I feel like they could talk about a bigger application and how it can actually make an impact on how to make cities more efficient.

[grahamschock]

Reviewer: Graham Schock Review Type: Critical

Problem being solved Due to the effects of urban sprawl more and more people are moving to urban areas. This has led to more technology and devices to support growing urban infrastructure. However, these devices have issues. The root of these problems is because the technology is difficult to deploy. This is due to a wide variety of limiting factors such as re-implementation of resources to sensors being reliant upon wired power sources.

Main Contributions Signpost aims to solve some of these issues by using an energy-harvesting, modular architecture. By having the signpost applications solar powered the applications are no longer dependent on wired mains. This means they can be placed in a wide variety of locations. They also provide a modular architecture which allows for easier development of different components as well as giving the signpost architecture more flexibility about which sensors it will deploy.

Questions

1. Sign stealing and vandalism is a problem. https://en.wikipedia.org/wiki/Street_sign_theft I wonder what mechanical components would need to be in place to deter theft. I know that a lot of the rent-able scooters probably face the same problem and they use alarms as well as GPS tracking to monitor the equipment. Would something like this be feasible?

2. Another logistical problem would be weather and temperature. Weather and temperature can vary between locations even with just in the United States. Would there be different models for a device in Death Valley versus Fairbanks, Alaska?

3. How do we authenticate these signposts? The paper talks about this briefly. These devices are required to be on some type of network and communicate with other devices. Regular authentication is a fairly expensive process, but so vital to a system like this so it does not turn into a IoDDOS.

Critique

1. I do not think the data power consumption and metrics are rigorous enough to say that the signposts can be effective in multiple places with multiple seasons. I think it is funny that they show "Solar harvesting in four cardinal directions and two seasons in Michigan"! When you originally read you think "wow these results must show that it can be used in the snow and in bad weather". However, after further inspection it seems that the two "seasons" are... late march and late july for... one week?

2. In the cost section I wish they would have elaborated more on the details. How does a $2,000 signpost that (probably needs maintenance as well as insurance, which is not covered in the cost break down) on a $250 street sign make economical sense? I think this cost is especially pervasive when one of the top (only) three functionalities it does is "count cars" (who cares?).

[samfrey99]

Reviewer: Sam Frey Review Type: Critical

Problem Being Solved: City-wide deployment of environmental sensors could provide constant access to endless data streams to help solve problems caused by rapid urban expansion, an urban planner's dream. Unfortunately, we still face issues of cost and scalability in these endeavors. When cities already have limited funds to invest in infrastructure, city-wide connectivity must offer a strong value proposition. The Signpost project seeks to address these issues to bring widespread distribution of sensors to urban areas.

Main Contributions: Signpost focuses on efficiency in energy use, cost, and scalability. Baked-in support for multi-tenancy allows the financial burden to be spread across a larger group. Multiple groups can invest in Signposts together and spread them around a city, helping each group scale more quickly. Virtualization of energy as a system resource means that developers writing applications for Signpost don't need to meet specific energy requirements prior to deployment.

Questions:

• When training for cross country in high school, my teammates and I had a (probably bad) habit of hitting signs on our runs. The authors state that they've faced no issues with vandalism or theft, but I still question whether or not widespread use of Signpost beyond a college campus is doable. What efforts could be made to ensure Signpost is durable enough to face not just thieves, but stupid teenagers?
• I find the wide range of energy harvested from sunlight concerning. Enough rainy days could leave a Signpost useless until the sun comes back out. Would it be possible to adapt Signpost to fit existing power line poles and draw from the grid when required, or could additional batteries be fit when low energy harvesting is expected?
• Given current technology, using cellular when the communication queue is full seems like a neccessary evil. Could this be remedied in the future with the expansion of 5G as an edge computing option?

Critique: There are two areas where I found the paper to be lacking in detail. The first to jump out at me is security. Without a concrete security plan, the seemingly open API that the SIgnpost team seeks to provide could quickly become a problem. While they don't intend to use Signpost to gather data about people, a security breach could quickly turn Signpost into public surveillance. Additionally, is there any measure in place to prevent a third party from buying a Signpost and using it that way on their own, ie. are there any security measures in place to prevent misuse of their product?

Secondly, I would've liked more detail into the support for multiple styles of applications. Given the limited power supply, an application running simultaneously across multiple Signposts seems a bit far-fetched. I don't doubt that they support the feature, but I'm left wondering how they've actually implemented this support.

[mjhegarty]

Reviewer: Michael Hegarty Review Type: Comprehension

Problem Being solved

With so called smart cities right around the corner, a infrastructure for citywide smart sensing needs to be set up in order to gather the data needed for this system. Some other solutions to this problem either require an opt-in model using residents smart phones and cars, which can have trouble getting participation, or requires to be wired up to mains(electrical grid of the city), which has time consuming and location limiting factors associated with it.

Main Contribution

The paper showcases a platform called signpost, which aims to solve the city sensing problem using small solar powered modules that can attach to signposts across the city. It features a multi tenant system, with physical separation between modules and a shared data bus for individual modules to communicate with the controller. It also includes multiple radio modules, system storage, a processor specifically for batching computations as well as a policies for power management.

Questions

1. One of the more interesting ideas presented in the paper to me is how they have an Intel Edison module specifically for handling batch computations and importing libraries that would be difficult to include in normal microcontrollers. Do you think that the inclusion of an entire extra module warrants the on board computations as opposed to just handling them off board via edge computing?
2. They talk about how their system is a multi tenant system via hardware separation, but doesn't the shared data bus prevent them from being a true multi tenant system? For example if one malicious module just kept the i2c line high wouldn't that have impacts across the whole system?
3. One other thing I'm curious about is the lifetime of each device. For devices deployed onto signposts with a large number of different components on them how long will they last when exposed to rain, wind, hail, etc? While they seem useful and are self sustaining, the price point of needing to regularly replace a $200 device might get pricey, especially considering how many would need to be deployed in a hypothetical smart city.

[huachuan]

Reviewer: Huachuan Wang Review Type: Comprehension

Overview This paper presents a modular and energy-harvesting flatform for city-scale sensing called Signpost. Signpost harvest energy from an integrated solar panel and can support multiple pluggable sensor modules. They deployed Signpost with several sensor modules and evaluate its ability to support sensing applications.

Contribution This paper introduces a solar energy-harvesting modular platform designed to enable city-scale deployments called Signpost. Signposts can provide energy, communications, storage, processing, time, and location services. Signposts can significantly benefit the city and residences if it can be made widely available. Signpost reduces the difficulty inherent in city-scale deployments enable new experimentation and provide improved insights into urban health.

Questions

1. Each Signpost device has a GPS module, however, because the device cannot move once installed, the location of the device is a constant, the location information could write in the cloud, does that GPS design necessary?

2. The vehicle counting section sounds very interesting, what is the benefit of implementing the frequency bin instead of the sound detector? The application can record the volume and identify the vehicle movement. However, I do not understand how it can count the occurrence of cars if more than one car passes by with small distances.

[hjaensch7]

Reviewer: Henry Jaensch

Review Type: Critical Review

Problem Being Solved

This paper recognizes the advantages of city scale sensing but acknowledges the logistical difficulties of deploying and maintaining city scale sensing. City scale sensing also comes with the question of multi-tenancy and long term use. Existing city wide sensor deployments have been single use and short term, this paper proposes a deployment solution that would also provide an option for multi-tenancy.

Main Contributions

Signpost is a platform that aims to solve the existing problems with city scale sensing. It leverages solar power to avoid having a wired electrical connection. Signpost also has modular sensor options, this allows for future proofing, sensors can be swapped without removing and reinstalling the system. This paper provided thurough analysis of energy availablity at different times of year. This work is useful to defend the viability of a solar powered sensing platform.

Questions

1. Guaranteeing isolation between tenants is difficult and necessary. Does signpost incur unwanted overheads by using virtualization to provide isolation, and can these overheads be avoided?

2. In practice, how are these protected? Signposts are vandalized regularly and the public usually isn't fond of more sensors/cameras. Would this solution incur unwanted costs in policing?

3. This is less for the paper and more for my own curiosity. "pedestrian route planning based on air quality" appears to be a poor allocation of resources. Why not fix the air instead of showing people how to walk around it?

Critiques

• When discussing multi-tenancy the paper proposed a platform energy tax. The idea of programming to manage energy that's potentially not constant sounds like a nightmare. As a broader critique this paper avoided providing implementation details for apps that are supposed to manage their own energy consumption. I'm not sure if this has been done before for battery powered systems but it doesn't seem trivial.

• This paper had sections addressing security but I'm not convinced this hardware wouldn't be physically destroyed. There are also security concerns regarding isolation. The paper provides some details on how energy isolation will be managed but there's a lot of shared hardware including a data bus and i/o ports. This design based on shared hardware opens up multiple attack vectors for malicious tenants, perhaps a component based operating system should be considered. ;-)

[rebeccc]

Reviewer: Becky Shanley Reviewer Type: Comprehension

Problem Being Solved This paper is trying to solve a specific issue of smart cities, the issue of a decentralized smart city. They propose a single system, Signpost, that contains all required sensors

Main Contributions While smart cities existed before this paper, this paper introduces a system that includes all required sensors on one device. This is a major contribution because it simplifies the prior model of a smart city significantly.

Three Questions

1. Although Signpost provides cities with a significantly easier smart system, wouldn’t it be significantly easier to compromise one system as opposed to many small systems?

2. The modularity of Signpost is said to be beneficial in its ability to scale up in complexity. At a point, even if the system is uniform, wouldn’t adding more modules add significant system complexity?

3. The lack of time spent on discussing the safety of Signpost left me with a lot of questions. They acknowledged that there is a risk of the data collected being used nefariously, however, proposed no solution to this problem. How can the current model of Signpost be deployed in cities with this major oversight? (I could be completely misunderstanding the severity of this and it could be fine)

[gkahl]

Reviewer: Greg Kahl

Review Type: Comprehension

Problem Being Solved

This paper discussed the design and implementation of an urban statistic sensing device. It would be implemented on signposts and harvest solar energy in order to power its modules which could perform various statistical and sensing tasks, such as measure noise pollution or analyze traffic.

Contributions

By design, the device would not need access to a power main or connection to WiFi. By harvest solar power, it allows the sensing device to be independent from power mains and not be limited by their availability when deciding on placement location. In addition to this, it will communicate using radio connection to be less dependent on WiFi access. Finally, they propose a modular design with isolation between modules. The system shall partition power to each of the modules and guarantee each a minimum to perform their tasks. This allows for modules which don't trust each other to be implemented on the device, and for sensors and modules to be upgraded in the future.

Questions

1 - They brought up both the ideas of being able to be placed in a wide range of places and the idea that the modules could be easily upgraded in the future. Is it really feasible to be going to Signpost devices distributed across an entire city to replace or add a new module? 2 - The paper discussing not reducing available power to the module below a minimum threshold, what happens in situations where the device is starved of solar power and there just isn't enough power to go around? 3 - They briefly discuss security and authentication, however they do not discuss any sort of physical security. With a device being secured to signposts, would sensitive data being stored locally cause a security concern if physically taken?

[tuhinadasgupta]

Reviewer: Tuhina Dasgupta Type: Comprehension

Problem:
The authors of this paper present a novel idea, Signpost, which is a smart city device. They present a plan to implement this system in different cities to improve the quality of life for the citizens of the city. Rather than having sensors for each desired independent module, they aim to integrate it into one system.

Main Contributions The primary contribution of this paper is to create a singular device that contains all of the sensors needed to create a smart city.

Three Questions The paper concedes that there are privacy concerns but doesn't address possible solutions and how people could be convinced of the security? The maintenance of the system will most likely cost much more than that of a regular system so how will that be addressed? There are critical systems (water,power) that will need to be changed from a regular system to Signpost wo/losing function, how will that be done?

[Others]

Reviewer: Gregor Peach

Type: Critical Review

Problem:

A big idea in "The IoT" is smart cities. A smart city is a city where sensors and instrumentation have been distributed around the city, in order to gather data about what's going on. The idea is that the huge amount of data gathered can be leveraged by "smart systems" that would inform more sophisticated city wide applications. (In the paper they mention about traffic management and weather.)

A huge problem, however, is the deployment and management of such a large number of sensors. That is the problem they attempt to tackle in the paper.

Contributions:

Their contribution is to come up with a sensor platform for "the IoT" city of the future. They argue a smart city could use something like their "signpost" for mass deployment. This is a super cool platform because it can handle multiple arbitrary applications. The platform itself only handles essential stuff like power. They've made the platform itself so easy to deploy that it would be scalable in the "smart city" situation.

Questions:

• Could this be scaled to tops of buildings or antenna?
• What other cool applications could we come up with? (beyond weather and car counting)
• How fit were the students doing the installation? (could this be a scaling problem)

Concerns:

• Why can't I read on time?
• I was unhappy with the length of the security section. I felt they didn't address snooping on the content of their transmissions.
• On a related note, I feel this has disturbing surveillance applications that were not addressed.

[mralexjacobson]

Reviewer: Alex Jacobson Review Type: Critical

Summary of problem: As humans seek to learn more about our urban environments, we may turn to city-scale sensing, which poses some challenges. A city scale deployment requires, physical installation, power management, and communications. The paper proposes a model called Signpost to address those concerns.

Main contributions: Signpost is a platform to reduce the difficulty of city scale deployments, enable experimentation and provide insights into urban health. It does this via 4 components: deployability, to enable large and frequent deployments. Accessibility reduces developer burden, and modularity allows different sensors on the signpost to change as time goes on, so that the signpost won’t be rendered obsolete. It also supports multi-tenancy, so that different sensors/applications can take advantage of the sign post without interfering with each other.

Questions:

If the different modules share a data bus, how do they ensure isolation and that the modules cannot interfere with each other?

The authors state that it may be harder to program for Signpost than other platforms. Why is this? Is it just because the programmers must worry about power consumption and resource management more so than if those resources were abundant?

What kind of modules would be added to a signpost to improve our understanding of the urban environment? The article talked about detecting gunshots, so perhaps there is a crime control advantage, but in what other ways would Signpost improve human life.

Critiques:

In section 4.2.2, the authors discuss several different ways in which Signpost could communicate, including cellular radios. However, they do not indicate which method Signpost uses or which method they discuss is most optimal. Later in the paper, they reveal that Signpost uses a variety of those previously discussed technologies. Why not just address the communication issue in one location, instead of interspersing it throughout the paper?

If the SIgnpost, modular in its design, is widely deployed around a city, it seems as if some security challenges may be posed. What is to stop a malicious actor from opening a signpost and adding their own, potentially harmful, module to it?

The article talks a lot about how the Signpost does not need to be connected to a power grid, as there may not always be a source available for it to connect to. However, that seems slightly dubious. While not an expert, the urban environments that Signpost seems to be targeting for its use probably do have a power grid that it could connect to. What urban areas don’t have power?