May 22, 2018

Announcing IoT Inspector: Studying Smart Home IoT Device Behavior

By Noah Apthorpe, Danny Y. Huang, Gunes Acar, Frank Li, Arvind Narayanan, Nick Feamster

An increasing number of home devices, from thermostats to light bulbs to garage door openers, are now Internet-connected. This “Internet of Things” (IoT) promises reduced energy consumption, more effective health management, and living spaces that react adaptively to users’ lifestyles. Unfortunately, recent IoT device hacks and personal data breaches have made security and privacy a focal point for IoT consumers, developers, and regulators.

Many IoT vulnerabilities sound like the plot of a science fiction dystopia. Internet-connected dolls allow strangers to spy on children remotely. Botnets of millions of security cameras and DVRs take down a global DNS service provider. Surgically implanted pacemakers are susceptible to remote takeover.

These security vulnerabilities, combined with the rapid evolution of IoT products, can leave consumers at risk, and in the dark about the risks they face when using these devices. For example, consumers may be unsure which companies receive personal information from IoT appliances, whether an IoT device has been hacked, or whether devices with always-on microphones listen to private conversations.

To shed light on the behavior of smart home IoT devices that consumers buy and install in their homes, we are announcing the IoT Inspector project.

Announcing IoT Inspector: Studying IoT Security and Privacy in Smart Homes

Today, at the Center for Information Technology Policy at Princeton, we are launching an ongoing initiative to study consumer IoT security and privacy, in an effort to understand the current state of smart home security and privacy in ways that ultimately help inform both technology and policy.

We have begun this effort by analyzing more than 50 home IoT devices ourselves. We are working on methods to help scale this analysis to more devices. If you have a particular device or type of device that you are concerned about, let us know. To learn more, visit the IoT Inspector website.

Our initial analyses have revealed several findings about home IoT security and privacy.

[Read more…]

Is It Time for an Data Sharing Clearinghouse for Internet Researchers?

Today’s Senate hearing with Facebook’s Mark Zuckerberg will start a long discussion on data collection and privacy from Internet companies. Although the spotlight is currently on Facebook, we shouldn’t forget that the picture is  broader: companies from device manufacturers to ISPs collect network traffic and use it for a variety of purposes.

The uses that we will hear about today are largely about the widespread collection of data about Internet users for targeted content delivery and advertising.  Meanwhile, yesterday Facebook announced an initiative to share data with independent researchers to study social media’s impact on elections. At the same time Facebook is being raked over the coals for sharing their data with “researchers” (Cambridge Analytica), they’ve announced a program to share their data with (presumably more “legitimate”) researchers.

Internet researchers depend on data. Sometimes, we can gather the data ourselves, using measurement tools deployed at the edge of the Internet (e.g., in home networks, on phones). In other cases, we need data from the companies that operate parts of the Internet, such as an Internet service provider (ISP), an Internet registrar, or an application provider (e.g., Facebook).

  • If incentives align, data flows to the researcher. Interacting with a company can work very well when goals are aligned. I’ve worked well with companies to develop new spam filtering algorithms, to develop new botnet detection algorithms, and to highlight empirical results that have informed policy debates.
  • If incentives do not align, then the researcher probably won’t get the data.  When research is purely technical, incentives often align. When the technical work crosses over into policy (as it does in areas like net neutrality, and as we are seeing with Facebook), there can be (insurmountable) hurdles to data access.

How an Internet Researcher Gets Data Today

How do Internet researchers get data from companies today? An Internet operator I know aptly characterizes the status quo:

“Show Internet operators you can do something useful, and they’ll give you data.”

Researchers get access to Internet data from companies in two ways: (1) working for the company (as an employee), or (2) working with the company (as an “independent” researcher).

Option #1: Work for a Company.

Working for a company offers privileged access to data, which can be used to mint impressive papers (irreproducibility aside) simply because nobody else has the same data. I have taken this approach myself on a number of occasions, having worked for an ISP (AT&T), a DNS company (Verisign), and an Internet security service provider (Damballa).

How this approach works. In the 2000s, research labs at AT&T and Sprint had privileged access to data, which gave rise to a proliferation of papers on “Everything You Wanted to Know About the Internet Backbone But Were Afraid to Ask”.  Today, the story repeats itself, except that the players are Google and Facebook, and the topic du jour is data-center networks.

Shortcomings of This Approach. Much research—from projects with a longer arc to certain policy-oriented questions—would never come to light if we only relied on company employees to do it. By the nature of their work, however, company employees lack independence. They lack both autonomy of selecting problems and in the ability to take positions or publish results that run counter to the company’s goals or priorities. This shortcoming may not matter if what the researcher wants to work on and what the company want to accomplish are the same. For many technical problems, this is the case (although there is still the tendency for the technical community to develop tunnel vision around areas where there is an abundance of data, while neglecting other areas). But for many problems—ranging from problems with a longer arc to deployment to those that may run counter to priorities—we can’t rely on industry to do the work.

#2: Work with a Company. 

How this approach works. A researcher may instead work with a company, typically gaining privileged access to data for a particular project. Sometimes, we demonstrate the promise of a technique with some data that we can gather or bootstrap without any help and use that initial study to pique the interest of a company who may then share data with us to further develop the idea.

Shortcomings of this approach. Research done in collaboration with a company often has similar shortcomings as the research that is done within a company’s walls. If the results of the research align with the company’s perspectives and viewpoints, then data sharing is copacetic. Even these cooperative settings do pose some risks to researchers, who may create the perception that they are not independent, merely by their association with the company. With purely technical research risks are lower, though still non-zero: for example, because the work depends on privileged data access, the researcher may still face challenges in presenting the research in a way that could help others reproduce it in the future.

With technical work that can inform or speak to policy questions, there are some concerns. First, certain types of research or results may never come to light—if a company doesn’t like the result that may result from the data analysis, then they may simply not share the data, or they may ask for “pre-publication review” for results based on that data (this practice is common for research that is conducted within companies as well). There is also a second, more subtle concern. Even when the work is technically watertight, a researcher can still face questions—fair or unfair—about the soundness of the work due to the perceived motivations or agendas of cooperative parties involved.

Current Data Sharing Approaches are Good, But They are Not Sufficient

The above methods for data sharing can work well for certain types of research. In my career, I have made hay playing by these rules—often working with a company, first by demonstrating the viability of an idea with a smaller dataset that we gather ourselves and “pitching” the idea to a company.

Yet, in my experience these approaches have two shortcomings. The first relates to incentives. The second relates to privacy.

Problem #1: Incentives.

Certain types of work depend on access to Internet data, but the company who holds the data may not have a direct incentive to facilitate the research. Possible studies of Facebook’s effect on elections certainly fall into this category: They simply may not like the results of the research.

But, there are plenty of other lines of research that fall into the category where incentives may not align. Other examples range from measurements of Internet capacity and performance as they relate to broadband regulation (e.g., net neutrality) to evaluation of an online platform’s content moderation algorithms and techniques. Lots of other work relating to consumer protection falls into this category as well. We have to rely on users and researchers measuring things at the edge of the network to figure out what’s going on; from this vantage point, certain activities may naturally slip under the radar more easily.

The current Internet data sharing roadmap doesn’t paint a rosy picture for research where incentives don’t align. Even when incentives do align, there can be perceptions of “capture”—effectively shilling an intellectual or technical finding in exchange for data access.

It is in the interests of everyone—the academics and their industry partners alike—to establish more formal modes of data exchange when either (1) there is determination that the problem is important to study for the health of the Internet, or for the benefit of consumers; (2) there is the potential that the research will be perceived as not objective due to the nature of the data sharing agreement.

Problem #2: Privacy.

Sharing Internet data with researchers can introduce substantial privacy risks, and the need to share data with any researcher who works with a company should be evaluated carefully—ideally by an independent third party.

When helping develop the researcher exception to the FCC’s broadband privacy rules, I submitted a comment that proposed the following criteria for sharing ISP data with researchers:

  1. Purpose of research. The data satisfies research that aims to promote security, stability, and reliability of networks. The research should have clear benefits for Internet innovation, operations, or security.
  2. Research goals do not violate privacy. The goals of the research does not include compromising consumer privacy;
  3. Privacy risks of data sharing are offset by benefits of the research. The risks of the data exchange are offset by the benefits of the research;
  4. Privacy risks of the data sharing are mitigated. Researchers should strive to use de-­identified data wherever possible.
  5. The data adds value to the research. The research is enhanced by access to the data.

Yet, outlining the criteria is one thing. The thornier question (which we did not address!) is: Who gets to decide the answers?

Universities have institutional review boards that can help evaluate the merits of such a data sharing agreement. But, Cambridge Analytica might have the veneer of “research”, and a company may have no internal incentive to independently evaluate the data sharing agreement on its merits. In light of recent events, we may be headed towards the conclusion that such data-sharing agreements should always be vetted by independent third-party review. If the research doesn’t involve a university, however, the natural question is: Who is that third party?

Looking Ahead: Data Clearinghouses for Internet Data?

Certain types of Internet research—particularly those that involve thorny regulatory or policy questions—could benefit from an independent clearing house, where researchers could propose studies and experiments for independent evaluation and have them evaluated and selected by an independent third party, based on their benefits and risks. Facebook is exploring this avenue in the limited setting of election integrity. This is an exciting step.

Moving forward, it will be interesting to see how Facebook’s meta-experiment on data sharing plays out, and whether it—or some variant—can serve as a model for Internet data sharing for other types of work writ large. In purely technical areas, such a clearinghouse could allow a broader range of researchers to explore, evaluate, reproduce and extend the types of work that for now remains largely irreproducible because data is under lock and key. For these questions, there could be significant benefit to the scientific community. In areas where the technical work or data analysis informs policy questions, the benefits to consumers could be even greater.

Artificial Intelligence and the Future of Online Content Moderation

Yesterday in Berlin, I attended a workshop on the use of artificial intelligence in governing communication online, hosted by the Humboldt Institute for Internet and Society.


In the United States and Europe, many platforms that host user content, such as Facebook, YouTube, and Twitter, have enjoyed safe harbor protections for the content they host, under laws such as Section 230 of the Communications Decency Act (CDA), the Digital Millenium Copyright Act (DMCA), and in Europe, Articles 12–15 of the eCommerce Directive. Some of these laws, such as the DMCA, provide immunity to platforms for copyright damages if the platforms remove content based on knowledge that it is unlawful. Section 230 of the CDA provides broad immunity to platforms, with the express goals of promoting economic development and free expression. Daphne Keller has a good summary of the legal landscape on intermediary liability.

Platforms are now facing increasing pressure to detect and remove illegal (and, in some cases, legal-but-objectionable) content. In the United States, for example, bills in the House and Senate would remove safe harbor protection for platforms that do not remove illegal content related to sex trafficking. The European Union has also considering laws that would limit the immunity of platforms who do not remove illegal content, which in the EU includes four categories: child sex abuse, incitement to terrorism, certain types of hate speech, and intellectual property or copyright infringement.

The mounting pressure on platforms to moderate online content coincides with increasing attention to algorithms that can automate the process of content moderation (“AI”) for the detection and ultimate removal of illegal (or unwanted) content.

The focus of yesterday’s workshop was to explore questions surrounding the role of AI in moderating content online, and the possible implications of AI for the moderation of online content and how online content moderation is governed.

Setting the Tone: Challenges for Automated Filtering

Malavika Jayaram from Digital Asia Hub and I delivered the two opening “impulse statements” for the day. Malavika talked about some of the inherent limitations of AI for automated detection (with a reference to the infamous “Not Hot Dog” app) and pointed out some of the tools that platforms are being pressured to use automated content moderation tools.

I spoke about our long line of research on applying machine learning to detect a wide range of unwanted traffic, ranging from spam to botnets to bulletproof scam hosting sites. I then talked about how the dialog has in some ways used the technical community’s past success in spam filtering to suggest that automated filtering of other types of content should be as easy as flipping a switch. Since spam detection was something we knew how to do, then surely the platforms could also ferret out everything from copyright violations to hate speech, right?

In practice Evan Engstrom and I previously wrote about the difficulty of applying automated filtering algorithms to copyrighted content.  In short, even with a database that matches fingerprints of audio and video content against fingerprints of known copyrighted content, the methods are imperfect. When framing the problem in terms of incitement to violence or hate speech, automated detection becomes even more challenging, due to “corner cases” such as parody, fair use, irony, and so forth. A recent article from James Grimmelmann summarizes some of these challenges.

What I Learned

Over the course of the day, I learned many things about automated filtering that I hadn’t previously thought about.

  • Regulators and platforms are under tremendous pressure to act, based on the assumption that the technical problems are easy.  Regulators and platforms alike are facing increasing pressure to act, as I previously mentioned. Part of the pressure comes from a perception that detection of unwanted content is a solved problem. This myth is sometimes perpetuated by the designers of the original content fingerprinting technologies, some of which are now in widespread use. But, there’s a big difference between testing fingerprints of content against a database of known offending content and building detection algorithms that can classify the semantics of content that has never been seen before. An area where technologists can contribute to this dialog is in studying and demonstrating the capabilities and limitations of automated filtering, both in terms of scale and accuracy. Technologists might study existing automated filtering techniques or design new ones entirely.
  • Takedown requests are a frequent instrument for censorship. I learned about the prevalence of “snitching”, whereby one user may request that a platform take down objectionable content by flagging the content or otherwise complaining about it—in such instances, oppressed groups (e.g., Rohingya Muslims) can be disproportionately targeted by large campaigns of takedown requests. (It was not known whether such campaigns to flag content have been automated on a large scale, but my intuition is that they likely are.) In such cases, the platforms err on the side of removing content, and the process for “remedy” (i.e., restoring the content) can be slow and tedious. This process creates a lever for censorship and suppression of speech.The trends are troubling: according to a recent article, a year ago, Facebook removed 50 percent of content that Israel requested be removed; now that figure is 95 percent.  Jillian York runs a site where users can report these types of takedowns, but these reports and statistics are all self-reported. A useful project might be to automate the measurement of takedowns for some portion of the ecosystem or group of users.
  • The larger platforms share content hashes of unwanted content, but the database and process are opaque. About nine months ago, Twitter, Facebook, YouTube, and Microsoft formed the Global Internet Forum to Counter Terrorism. Essentially, the project relies on something called the Shared Industry Hash Database. It’s very challenging to find anything about this database online aside from a few blog posts from the companies, although it does seem in some way associated with Tech Against Terrorism.The secretive nature of the shared hash database and the process itself has a couple of implications. First, the database is difficult to audit—if content is wrongly placed in the database, removing it would appear next to impossible. Second, only the member companies can check content against the database, essentially preventing smaller companies (e.g., startups) from benefitting from the information. Such limits in knowledge could ultimately prove to be a significant disadvantage if the platforms are ultimately held liable for the content that they are hosting. As I discovered throughout the day, the opaque nature of commercial content moderation proves to be a recurring theme, which I’ll return to later.
  • Different countries have very different definitions of unlawful content. The patchwork of laws governing speech on the Internet makes regulation complicated, as different countries have different laws and restrictions on speech. For example, “incitement to violence” or “hate speech” might mean a different thing in Germany (where Nazi propaganda is illegal) than it does in Spain (where it is illegal to insult the king) or France (which recently vowed to ferret out racist content on social media). When applying this observation to automated detection of illegal content, things become complicated. It becomes impossible to train a single classifier that can be applied generally; essentially, each jurisdiction needs its own classifier.
  • Norms and speech evolve over time, often rapidly. Several attendees observed that most of the automated filtering techniques today boil down to flagging content based on keywords. Such a model can be incredibly difficult to maintain, particularly when it comes to detecting certain types of content such as hate speech. For one, norms and language evolve; a word that was innocuous or unremarkable today could take on an entirely new meaning tomorrow. Complicating matters further, sometimes people try to regain control in an online discussion by co-opting a slur; therefore, a model that bases classification on the presence of certain keywords can produce unexpected false positives, especially in the absence of context.


Aside from the information I learned above, I also took away a few themes about the state of online content moderation:

  • There will likely always be a human in the loop. We must figure out what role the human should play. Detection algorithms are only as good as their input data. If the data is biased, if norms and language evolve, or if data is mislabeled (an even more likely occurrence, since a label like “hate speech” could differ by country), then the outputs will be incorrect. Additionally, algorithms can only detect proxies for semantics and meaning (e.g., an ISIS flag, a large swath of bare skin) but have much more difficulty assessing context, fair use, parody, and other nuance. In short, on a technical front, we have our work cut out for us. It was widely held that humans will always need to be in the loop, and that AI should merely be an assistive technology, for triage, scale, and improving human effectiveness and efficiency when making decisions about moderation. Figuring out the appropriate division of labor between machines and humans is a challenging technical, social, and legal problem.
  • Governance and auditing is currently challenging because decision-making is secretive. The online platforms currently control all aspects of content moderation and governance. They have the data; nobody else has it. They know the classification algorithms they use and the features they use as input to those algorithms; nobody else knows them. They also are the only ones who have insight into the ultimate decision-making process. This situation is different from other unwanted traffic detection problems that the computer science research community has worked on, where it was relatively easy to get a trace of email spam or denial of service traffic, either by generating it or by working with an ISP. In this situation, everything is under lock and key.This lack of public access to data and information makes it difficult to audit the process that platforms are currently using, and it also raises important questions about governance:
    • Should the platforms be the ultimate arbiter in takedown and moderation?
    • Is that an acceptable situation, even if we don’t know the rules that they are using to make those decisions?
    • Who trains the algorithms, and with what data?
    • Who gets access to the models and algorithms? How does disclosure work?
    • How does a user learn that his or her content was taken down, as well as why it was taken down?
    • What are the steps to remedy an incorrect, unlawful, or unjust takedown request?
    • How can we trust the platforms to make the right decisions when in some cases it is in their financial interests to suppress speech? History has suggested that trusting the platforms to do the right thing in these situations can lead to restrictions on speech.

Many of the above questions are regulatory. Yet, technologists can play a role for some aspects of these questions. For example, measurement tools might detect and evaluate removal and takedowns of content for some well-scoped forum or topic. A useful starting point for the design of such a measurement system could be a platform such as Politwoops, which monitors tweets that politicians have deleted.


The workshop was enlightening. I came as a technologist wanting to learn more about how computer science might be applied to the social and legal problems concerning content moderation; I came away with a few ideas, fueled by exciting discussion. The attendees were an healthy mix of computer scientists, regulators, practitioners, legal scholars, and human rights activists. I’ve worked on censorship of Internet protocols for many years, but in some sense measuring censorship can feel a little bit like looking for one’s key under the lamppost—my sense is that the real power over speech is now held by the platforms, and as a community we need new mechanisms—technical, legal, economic, and social—to hold them to account.