Advanced Search
Content Type: Examples
A 2015 study carried out in Rwanda and published in the journal Science used mobile phone records to study the distribution of wealth and poverty in an attempt to fill in the gap left by the difficulty of collecting accurate statistics. A rought idea of geographic location was derived from anonymised data on billions of interactions such as when calls were made and received, when text messages were sent and which cell towers they were routed through. Combined with responses from about 850…
Content Type: Examples
Automated systems such as the personality test developed by Massachusetts-based workforce management company Kronos are increasingly used by large companies to screen job applicants. To avoid falling foul of regulations prohibiting discrimination against those with mental illness, often the questions are phrased in intricate ways that are harder to game - but also harder to answer without self-incrimination. It's estimated that as many as 72% of CVs are never seen by human eyes; instead, they…
Content Type: Examples
As speech recognition and language-processing software continue to improve, the potential exists for digital personal assistants - Apple's Siri, Amazon's Alexa, and Google Assistant - to amass deeper profiles of customers than has ever been possible before. A new level of competition arrived in 2016, when Google launched its Home wireless speaker into a market that already included the Amazon Echo, launched in 2014. It remained unclear how much people would use these assistants and how these…
Content Type: Examples
The price of using voice search is that Google records many of the conversations that take place in their presence. Users wishing to understand what Google has captured can do so by accessing the portal the company introduced in 2015. Their personal history pages on the site include both a page showing activity on the web and a separate specific audio page that lists the captured recordings. The information made available there includes when and how by what device or app the sound was recorded…
Content Type: Examples
In 2015, a small number of Silicon Valley start-ups began experimenting with assessing prospective borrowers in developing countries such as Kenya by inspecting their smartphones. Doing so, they claimed, enabled them to charge less in interest than more traditional microlenders, since many of their target customers lack traditional credit ratings. The amount of data on phones - GPS coordinates, texts, emails, app data, and more obscure details such as how often the user recharges the battery,…
Content Type: Examples
In the past, car insurers relied on drivers to report their annual mileage when renewing their policies. Increasingly, however, insurers are turning to other methods. In 2016, State Farm, a US insurance company, acknowledged that the company verifies the mileage driven by the cars it insures in a variety of ways: directly from customer reports; from monitoring devices plugged into their cars; and in some cases from third-party vendors. These third-party vendors may include car dealers and also…
Content Type: Examples
In 2015, Ant Financial, a Chinese company affiliated with Alibaba, began deploying "Sesame Credit scores". Ant, which also runs the popular payment app Alipay, claims that it uses both online and offline purchasing and spending habits to calculate a credit worthiness score for each of its more than 350 million users. Users who meet various thresholds qualify for bonuses such as small loans or waivers on deposits. Almost as soon as the system went live, users of popular social media apps such as…
Content Type: Examples
In 2016 researchers in China claimed an experimental algorithm could correctly identify criminals based on images of their faces 89% of the time. The research involved training an algorithm on 90% of a dataset of 1,856 photos of Chinese males between 18 and 55 with no facial hair or markings. Among them were 730 ID pictures of convicted criminals or suspects wanted by the Ministry of Public Security. However, criminology experts warned that the results may merely reflect bias in the criminal…
Content Type: Examples
In 2016 researchers at the University of Texas at Austin and Cornell University demonstrated that a neural network trained on image datasets can successfully identify faces and objects that have been blurred, pixellated, or obscured by the P3 privacy system. In some cases, the algorithm performed more consistently than humans.
https://www.schneier.com/blog/archives/2016/09/using_neural_ne.html
Writer: Bruce Schneier
Publication: Schneier on Security
Publication date:
Content Type: Examples
In 2015, Facebook's AI lab announced that its researchers had devised an experimental algorithm that could recognise people in photographs even when their faces are hidden or turned away. The researchers trained a sophisticated neural network on a dataset of 40,000 photographs taken from Flickr, some full-face and others facing away and were able to reach 83% accuracy. The company hopes an algorithm like this one could help apps like Facebook's then-new Moments sort photos into groups. …
Content Type: Examples
In 2016, Facebook gave conflicting accounts of whether the service uses location data in order to recommend prospective Friends in its "People You May Know" feature. When the company first admitted - for publication - that location data was indeed one of the signals it used, many users felt this explained some of the PYMK recommendations they were seeing and experts pointed out the similarity to the NSA's techniques as were exposed in 2013 by the Snowden revelations. A few days later,…
Content Type: Examples
In September 2016, the US Federal Trade Commission hosted a workshop to study the impact of big data analysis on poor people, whose efforts to escape poverty may be hindered by the extensive amounts of data being gathered about them. Among those who intensively surveil low-income communities are police, public benefits programmes, child welfare systems, and monitoring programmes for domestic abuse offenders. Some areas require applicants for food stamps and other public benefits to undergo…
Content Type: Examples
In 2016, Facebook and Google began introducing ways to measure the effectiveness of online ads by linking them to offline sales and in-store visits. Facebook's measurement tools are intended to allow stores to see how many people visit in person after seeing a Facebook campaign, and the company offered real-time updates and ad optimisation. Facebook noted that information will only be collected from people who have turned on location services on their phones. The company also offered an Offline…
Content Type: Examples
In 2016, researchers affiliated with Verto Analytics and the Qatar Computing Research Institute published work in which they analysed the app usage and demographics of more than 3,700 people in order to find correlations. Based on the models they developed, they found they could predict a user's gender, age, marital status, and income with between 61% and 82% accuracy.
https://www.washingtonpost.com/news/the-intersect/wp/2016/03/03/quiz-can-we-guess-your-age-and-income-based-solely-on-…
Content Type: Examples
In 2016, supporters of Ted Cruz and Rand Paul for president were surprised to begin getting emails from the Trump campaign soon after their candidates dropped out of the race for the Republican nomination. In an investigation, CNNMoney found that nearly every failed 2016 presidential candidate sold, rented, or loaned their supporters' email addresses to other candidates, marketers, charities, and private companies. From analysing thousands of emails and Federal Election Commission records,…
Content Type: Examples
In 2016, Verbraucherzentrale NRW, a consumer protection organisation in the German state of North Rhine-Westphalia accused Samsung of harvesting data and sending it back to the company over the internet without informing users as soon as its smart televisions are connected to the internet. The complaint specifies the model UE40H6270. The group wants Samsung to default to an anonymous setting rather than transmit information including the show being watched, the time of date, and the user's IP…
Content Type: Examples
In 2016, VICE News discovered that the confidential and "shadowy" World-Check database, which has wrongly linked individuals to terrorist activity, was being widely used by British police and intelligence. Also a customer is the Charity Commission, which uses it to screen charities and aid beneficiaries. Owned by Thomson-Reuters, World-Check profiles individuals using "open-source" information to uncover their "hidden risk" for government agencies and banks and is part of a rapidly growing by…
Content Type: Examples
In 2016, the US's third-largest property and casualty insurer, Liberty Mutual, announced it would partner with Subaru to enable drivers who have bought Subaru's Starlink infotainment system to download a car app that will notify them if they are accelerating too aggressively or braking too hard. The insurer's RightTrack programme, which began in 2012 and of which the app is a part, gives drivers a 5% discount for enrolling and further discounts of up to 30% for following the app's instructions…
Content Type: Examples
In 2017 two Metropolitan police officers were jailed for five years for hacking into the force's intelligence database and leaking the identity of a protected witness to the defence lawyer in a 2011 murder trial that eventually saw the defendant, Leon De St Aubin, convicted. While the trial was in progress, the St Aubin's girlfriend, Lydia Laura, got a job as a civilian custody officer at Hammersmith police station without disclosing her relationship. In return for sexual favours, police…
Content Type: Examples
In a 2017 study of patterns of postings on Chinese social media, three Harvard researchers disagreed with the widespread claim that the government's strategy is to post "50c party" posts that argue for the government's side in policy and political debates. Instead, the researchers estimated that the Chinese government fabricates as many as 448 million social media comments a year but found that these avoid controversial issues and arguments with government skeptics. Instead, the researchers…
Content Type: Examples
In early 2016 Libreville, the capital of Gabon, signed up for Microsoft's CityNext programme, which is intended to supply innovative "smart city" solutions in eight key areas: health, social services, infrastructure, water, electricity, justice, culture, and education. Applications in each area will allow the city to manage traffic and urban transport, govern and collect taxes, and provide citizens with electronic access to health, citizen, police, and emergency services, as well as make it…
Content Type: Examples
In 2016 Jonathan Evans, the former head of Britain's MI5 warned that private firms are analysing "open source" - that is, publicly posted - information to create profiles that are just as intrusive as anything Britain's intelligence agencies deploy and that the gap is closing between open information and secret intelligence. Evans suggested that, compared to telephone companies and banks, giant internet firms were failing in their "moral duty" to protect their users. He drew the line at…
Content Type: Examples
By 2016, numerous examples had surfaced of bias in facial recognition systems that meant they failed to recognise non-white faces, labelled non-white people as "gorillas", "animals", or "apes" (Google, Flickr), told Asian users their eyes were closed when taking photographs (Nikon), or tracked white faces but couldn't see black ones (HP). The consequences are endemic unfairness and a system that demoralises those who don't fit the "standard". Some possible remedies include ensuring diversity in…
Content Type: Examples
In 2016, when security expert Matthew Garrett stayed in a London hotel where the light switches had been replaced by Android tablets, it took him only a few hours to gain access to all of the room's electronics. The steps he followed: plug his laptop into a link in place of one of the tablets; set up a transparent bridge; analyse the data traffic with WireShark to identify the protocol in use; then exploit that protocol. That protocol was Modbus, an old protocol with no authentication. Once…
Content Type: Examples
By 2016, a logical direction for data-driven personalisation efforts to go was toward the "Internet of Emotions": equipping devices with facial, vocal, and biometric sensors that use affective computing to analyse and influence the feelings of device owners. Of particular concern is the potential for using subtle cues to manipulate people in a more nuanced way than is presently discussed. The beginnings of this are already visible in the example of an Amazon Echo that displayed the items a…
Content Type: Examples
The discovery in 2016 of previous hacker break-ins such as the 2013 theft of 360 million old MySpace accounts and the 2012 hack of LinkedIn suggest that although websites come and go and "linkrot" means web pages have a short half-life, user data lives on for a deceptively long time. This is especially true of user names, passwords, profiles, and other personal data that is both commercially valuable and useful to criminals over a long period. The email address that someone used to open a…
Content Type: Examples
In 2015, Chinese authorities banned the 1.6 million members of the country's People's Liberation Army from using smartwatches and other wearable technology in order to prevent security breaches. Army leaders announced the decision after a soldier in the city of Nanjing was reported for trying to use a smartwatch he had been given to take pictures of his comrades because automatic connections could mean the devices uploaded voice and video to the cloud in violation of national security…
Content Type: Examples
In 2015, ABI Research discovered that the power light on the front of Alphabet's Nest Cam was deceptive: even when users had used the associated app to power down the camera and the power light went off, the device continued to monitor its surroundings, noting sound, movement, and other activities. The proof lay in the fact that the device's power drain diminished by an amount consistent with only turning off the LED light. Alphabet explained the reason was that the camera had to be ready to be…
Content Type: Examples
In 2016, researchers at Dalhousie University in Canada and the Weizman Institute of Science in Israel developed a proof-of-concept attack that allowed them to take control of LED light bulbs from a distance of up to 400 metres by exploiting a flaw in the Zigbee protocol implementation used in the Philips Hue system. Because the same key was used in every bulb, once it had been extracted from one bulb it could be reused on all of them. Writing a new operating system to one of the bulbs…
Content Type: Examples
At the 2016 Usenix conference, MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) presented a system called Chronos that could use wifi signals to detect the position of a person or object inside a room to within tens of centimetres. MIT claimed Chronos was 20 times more accurate than previous wifi-based tracking systems, 94% successful in detecting which room a person was in, and 97% accurate in determining if a customer was inside or outside a shop. The researchers hoped…