Surveilled

Can you survive a day with extreme digital surveillance?

Surveilled was developed using Twine for my Surviving Social Media class’ final project. Its goal is to delve deeper into the world of digital surveillance and specifically look at how it is currently being applied to protests. The game takes place in a made-up country where the government strictly opposes any mention of climate change. Your goal as the player is to attend a climate change protest and make it back home safely.

This post contains spoilers for the game as I will be talking about how I designed the storyline and explaining the impacts of each decision in more detail. Please try playing the game first and see how far you can get!

Play Now!

Choices

Like with any chose your own adventure game, there are many choices that you can make. This section will break down each one, explaining the reasoning behind each one and how they impact the ending you get. Given that the game is trying to show the worst-case scenario, making a single mistake will always give you a bad ending. The game was mostly modeled after the How Hong Kong Protesters Evade Surveillance With Tech video by the Wall Street Journal, but uses other articles to dive deeper into specific choices.

Will You Go?

Selecting Yes progresses the player to the next decision. However, selecting No leads the player to the False Arrest ending. This is meant to motivate the player to explore the rest of the game while also commenting on the current accuracy of facial recognition technology.

What Will You Wear?

Selecting Everyday Clothes flags the player for arrest since they will be easily recognizable. Their face is visible and they are wearing clothes that are tied to their every day appearance.

Selecting Plain Clothes with Hat, Sunglasses, and a Mask is the safe option in this choice. There is nothing distinctive about the clothing and the combination of a plain black mask, sunglasses, and a hat hide key points of the face that facial recognition algorithms use to identify people. According to this Digital Trends article about the CV Dazzle makeup style:

“Normal black masks and sunglasses, as we have them now, offer the best widespread protection, as we don’t know for sure what the police are using,” White told Digital Trends.

Selecting Elaborate Makeup to Confuse Cameras also flags the player for arrest. While seemingly a good option, these will only target specific facial recognition algorithms. This choice is also meant to represent other methods of defeating the cameras such as QR codes, lenses, masks, glasses, and even wearable prints. There was a study covered by Vice about using more natural-looking makeup to fool the algorithms. This method proved to be very effective while also using conventional and neutral makeup, making sure you don’t stand out.

The experiment saw 100 percent success in the digital experiments on both the FaceNet model and the LResNet model, according to the paper. In the physical experiments, the participants were detected in 47.6 percent of the frames if they weren’t wearing any makeup and 33.7 percent of the frames if they wore randomly applied makeup. Using the researchers’ method of applying makeup to the highly identifiable parts of the attacker’s face, they were only recognized in 1.2 percent of the frames.

Examples of CV Dazzle styles:

What Will You Bring?

Phone

Choosing the Personal Phone flags the player for arrest. It is possible to set up your personal phone properly to be untraceable during a protest. However, since I wanted this to model a worst-case scenario, we assume that the player is not quite tech savvy enough to execute the set up perfectly, allowing themself to be tracked.

This Wired article details how to avoid getting caught by various law enforcement tools such as wireless interception of text messages and stingray devices (fake cell towers that collect data on all phones that connect to them). They recommend putting your phone into a Faraday bag, which will block any wireless transmissions.

Another concern that comes with taking your personal phone is what happens with your data after you are arrested. For example, in the Hong Kong protests, the best method of locking your phone was with the standard PIN system. Since there is a law that protects people from being required to reveal knowledge, police are not able to get into their phones. However, facial scans and finger prints do not fall under its protection.

Choosing the Burner Phone is the safe option for this one. A burner phone is a cheap and disposable pre-paid phone, perfect for maintaining your anonymity. However, purchasing these also comes with the risk of your identifying information being collected upon checking out (for example, your credit card information). This HowToGeek article excerpt goes though multiple weak points in the purchasing process:

In the process of this: If you took your normal phone with you, your cellular carrier will know that you were at the store at the time the phone was purchased. License plate cameras on the route may have captured your license plate and recorded your movements. A camera in the store may have recorded you buying the phone. Your credit card company will have a record of you buying the phone. When you turn the phone on at home, the cellular carrier your phone uses will have a pretty good idea of where your home address is.

Choosing Neither seems like it will be a safe option, but in this case it leads to the Out of the Loop ending.

Money

Choosing to bring the Cards (ID, Credit, Train) flags the player for arrest. However, it also triggers a special event when they arrive at the protest. The player will be approached by another protester who offers to wrap the player’s cards in tinfoil like they showed in the Hong Kong protest video. This is to protect the cards from getting scanned by the police.

Choosing to bring Cash is the safe option. Cash purchases are untraceable, so buying a train ticket back home won’t leave evidence like a train pass would.

Choosing to bring Neither flags the player for the Curfew ending. Since the player has moved far from the original staring point, they must buy a train ticket to get back home on time. However, if the player has no money, they are forced to try and walk back.

Additional Items

This choice has no impact on the ending of the game. It only affects what the player does when they get to the protest and explains how the chosen item would potentially be used. These are also inspired by the Hong Kong protesters.

Choosing the Spray Paint prompts an event where the player joins a group of protesters who are covering security camera lenses with spray paint. This helps guard everyone’s identity from potential facial recognition software.

Choosing the Laser Pointer prompts an event where the player pulls out their laser pointer and joins the larger group in confusing the higher-up cameras with a light show. The bright light also serves as a distraction to the police. Laser pointers have been declared illegal, with officials claiming that they can blind people or cause permanent eye damage.

Choosing Neither leads to an uneventful protest interaction.

How Are You Communicating?

Choosing the popular, short-form social media platform Squaqr leads to the Tracked ending.

Choosing the anonymous, encrypted messaging app Telephone is one of the safe options. This is meant to represent apps like Telegram and Signal. The appeal of these apps is their security and the anonymity that they provide. They also keep everyone up-to-date with real-time messaging.

Choosing the anonymous file sharing service PortalDrop is the other safe option. This is meant to be an off-brand version of AirDrop. The appeal of a service like AirDrop is the anonymity of the sender. People can share things like links and images amongst each other in real-time.

Endings

This section lists all possible endings in the game. It explains the research behind each one and how to get them.

False Arrest

This ending details the player getting arrested in their home after deciding to not attend the protest. Facial recognition software incorrectly identified them as a problematic protestor.

This Center of Strategic and International Studies (CSIS) article talks about a series of studies done by the National Institute of Standards and Technology (NIST). They vetted several facial recognition algorithms using a standardized test called the Facial Recognition Vendor Test (FRBT). In perfect conditions, the algorithms performed with 99.97% accuracy. However, when these conditions shift, this accuracy drops dramatically (36 - 87% depending on camera placement). Real life conditions, will almost always be far from perfect, so relying on recognition algorithms like these would be disastrous for making arrests. However, if this method were combined with other methods of identification, it may be a more viable approach.

To get this ending:

Out of the Loop

This ending details the player getting swept up into a crowd and ending up cornered by the police. Having no form of communication leads to a lack of awareness of the groups plans. Often, police will set up areas to corral the protestors into and arrest them. If you don’t know where these would be (due a lack of real-time updates), you could easily get caught up in one.

To get this ending:

Tracked

This ending is meant to resemble an Intercept article about how posts from Twitter can be scraped and used maliciously. The article looks into Dataminr, an artificial intelligence startup. Twitter provides them with a content stream known as the firehose, which then gets analyzed in real time and provides clients with geographic information about crisis events. This technology is very helpful, but also depends on the client and how they will use it. For example, it was discovered that police were using using this information to monitor and track the George Floyd protests:

Dataminr’s Black Lives Matter protest surveillance included persistent monitoring of social media to tip off police to the locations and activities of protests, developments within specific rallies, as well as instances of alleged “looting” and other property damage.

“It’s true Dataminr doesn’t specifically track protesters and activists individually, but at the request of the police they are tracking protests, and therefore protesters,” this source explained.

To get this ending:

Curfew

This ending is unlocked by choosing to bring no form of money. The protest has moved far away from the original gathering point while dodging police and the only way back home before the country’s curfew is by train. However, since the player has chosen to bring no money, they get caught by the police after trying to stealthily make their way back home.

To get this ending:

Home Arrest

In this ending, the player successfully makes it home unscathed. However, police were able to digitally track their presence at the protest.

To get this ending (any of these options):

Safely Home

This is the only good ending in the game. The player returns home safely after the protest.

The Story Outline

![[../assets/surveilled.png]]

Future of Digital Surveillance

A Computer Science project done at Stanford on the ethics of surveillance gives a summary of MIT professor Gary Marx’s framework for evaluating surveillance methods. Some questions that I thought were especially important were:

Golden rule: would those responsible for the surveillance (both the decision to apply it and its actual application) agree to be its subjects under the conditions in which they apply it to others?

Human review: is there human review of machine generated results?

Right to challenge and express a grievance: are there procedures for challenging the results, or for entering alternative data or interpretations into the record?

Information used for original vs. other unrelated purposes: is the personal information used for the reasons offered for its collection and for which consent may have been given and does the data stay with the original collector, or does it migrate elsewhere?

Currently the trends of digital surveillance in the United States are worrying. This Just Security article talks about how aerial and social media surveillance, as well as facial recognition technology, were extensively used to monitor the recent Black Lives Matter protests. Facial recognition in particular has been a very popular tool for police. For example:

The New York Police Department (NYPD) has made nearly 3,000 arrests based on facial recognition searches in the first five and a half years of using the technology. Florida law enforcement offices run an average of 8,000 searches per month using Pinellas County’s facial recognition system.

However, this article also cites the more recent NIST study about the accuracy of facial recognition algorithms. Spoiler alert, it’s not very good if you’re not a while man:

certain algorithms were more likely to misidentify African American or Asian individuals than White males “by factors of 10 to beyond 100 times.”

Current algorithms are biased. This is partly because the data sets are skewed towards whoever we have more images of, which in the United States, tends to be white men. Additionally, the bias will also come from the creators of the algorithm. Often the teams that write these programs are not very diverse, making it very difficult to be aware other peoples’ points of view and how they would be affected by the technology.

The solutions that I can see for the future of digital surveillance are:

  • Hire more diverse teams that can rigorously test the algorithms
  • Stop using biased ones as the sole basis for important decisions
  • Set up a robust set of laws concerning data collection and use