The Linkielist

Linking ideas with the world

The Linkielist

Final Fantasy VII background graphics upscaled 4x by AI

Posted on by
The Remako HD Graphics Mod is a mod that completely revamps the pre-rendered backgrounds of the classic JRPG Final Fantasy VII. All of the backgrounds now have 4 times the resolution of the original.

Using state of the art AI neural networks, this upscaling tries to emulate the detail the original renders would have had. This helps the new visuals to come as close to a higher resolution re-rendering of the original as possible with current technology.

What does it look like?

Bbelow are two trailers. One is a comparison of the raw images, while the other shows off the mod in action.
If that’s still not enough, then please check out the screenshot gallery here.

Source: FF7 Remako HD Graphics Mod Beta Released

Custom firmware for lights allows you to control them with Homeassistant and more controllers

Posted on by

Sonoff B1, lights and shades

12 Replies

Six months ago I was reviewing the AiThinker AiLight, a great looking light bulb with an embedded ESP8266EX microcontroller, driven by a MY9291 LED driver. Just before summer IteadStudio released it’s Sonoff B1 [Itead.cc] light bulb, heavily inspired (probably same manufacturer) by the AiLight, at least on the design.

Now that IteadStudio has become popular between the home automation community you can also find the Sonoff B1 on global marketplaces like Ebay or Aliexpress for around 13€.

A closer look at the B1 uncovers some important differences. But before going deeper into the details let me first say that this post will probably look more like a review, at least more than I use to write. And second, yes: ESPurna supports the Sonoff B1 🙂

An unboxing?

Not quite so. I leave that to other people with better skills on the video editing world. Let me just tell you than the “box” is somewhat different from what I expected. You might recall the AiLight box: a simple beige drawer-like box with a “WiFi Light” text and a simple icon. No colors, pictures, specifications,… nothing.

Instead, the Sonoff B1 I received from IteadStudio comes in a colorful box, with the usual pictures and data you can find in retail products.

Inside the box the light bulb is comfy housed in a polyethylene foam, along with a quality control certification and a small “getting started” manual in English and Chinese.

A heat sink?

Don’t think so. The first thing I noticed when I opened the box was that the bulb was very similar to the AiLight, the second the only visual difference. It certainly looks like a big heat sink. I almost fear touching it while connected. But how much heat can you generate if the light is rated 6W? The bulb body houses a basic AC/DC power supply (90-250VAC to 12VDC) and is accessible unscrewing the metal frame (the heat-sink part from the smooth part with the “sonoff” logo).

The AiLight is also 6W and you can safely touch it, even when it has been at full power for a lot of time. The Sonoff B1 shouldn’t be different. So I’m lean towards thinking it’s an aesthetic decision. Unless there are some beefy power LEDs inside.

Power LEDs?

Not all of them. Anyway I think this is the aspect where the B1 clearly differentiates from the AiLight. The later has 8 cold white power LEDs, as well as 6 red, 4 green and 4 blue power LEDs. The Sonoff B1 also has 8 cold white ones. But then it features 8 warm white power LEDs and 3 5050 RGB LEDs!

I don’t have a luximeter but the difference when fully white between the two is hard to spot. But the warm white color really makes the difference in favor of the Sonoff bulb. On the other hand, the 3 5050 SMD LEDs are clearly not enough. Even more: since the RGB LEDs are closer to the center of the round PCB, just around the WiFi antenna, the shadow of the antenna is very noticeable if you are using a colored light.

Hard to tell which one is brighter for the naked eye…

The pic does not justice the difference. The right on is the AiLight with the white power LEDs at full duty. The left on is the Sonoff B1 using the warm white power LEDs (you can see the yellowish color in the wall). The cold white LEDs are brighter but, depending on the room, the warm white LEDs could be more suitable.

Both bulbs again, now with the red channel at full duty. No need for words.

3 5050 RGB LEDs, 3 shadows of the antenna

A view without the cap, red LEDs are at 100% duty cycle, white LEDs are only at 10%…

I think the Sonoff B1 could be a better choice when used to illuminate with a warm white light your living room or your bedroom than the AiLight. If you need a colorful illumination, discotheque moods or a nice cold white for your kitchen, use the AiLight. Another possible (and interesting) use for Sonoff B1 would be as a notification light using traffic light color code, for instance. Clearly visible but not disturbing colors.

The controller?

Not the same. It is actually an ESP8285. In practice, you can talk to it like if it was an ESP2866 with a 1Mb embedded flash using DOUT flash mode. So that’s my recommended configuration.

The ESP8285 and required components with the 5050 RGB LEDs

As you can see in the pictures, the PCB is actually 2 PCB, one for the power LEDs and the other one for the microcontroller, some components and the 5050 on the front, a buck converter (12VDC to 3.3VDC for the ESP8285) and the LED driver on the back. The two PCBs are soldered together and glued to the underneath support.

In the AiLight the LED driver is a MY9291 [datasheet, PDF] by My-Semi. The Sonoff B1 uses another My-Semi driver, the MY9231 [datasheet, PDF]. The MY9291 is a 4 channels LED driver but the MY9231 is just 3 channels… so how is it possible to do RGB plus white and warm? Well actually these ICs are daisy chainable, so there are two MY9231 controllers in the Sonoff B1, the first one controlling the white power LEDs and the second the 5050 RGB LEDs.

I did not want to remove the glue under the PCB. But you can glimpse one My-Semi controller through the bottom hole.

ESPurna?

The ESPurna firmware is released as free open software and can be checked out at my Espurna repository on GitHub.

Sure! You can flash the Sonoff B1 following the same procedure of the AiLight. There are 6 pads on the PCB labelled 3V3, RX, TX, GND, GPIO0 and SDA. You will need to wire the first 5 (tin you cable, apply a small drop on the pad and then heat them together). Connect RX to TX, TX to RX, GND to GND, GPIO0 to GND and finally 3V3 to the 3V3 power source of your programmer. It will then enter into flash mode (GPIO0 is grounded). You can either flash the bin file from the ESPurna downloads section or build your own image (check the ESPurna wiki for docs).

Wired flashing of the Sonoff B1

Since ESPurna version 1.9.0 you define and control any number of dimming channels, you can also define the first three to be RGB channels. If you do, the web UI will show you a colorpicker to select the color.

You can also control it via MQTT. It supports CSS notation, comma separated or color temperature, as well as brightness and status, of course.

1
2
3
4
5
6
7
8
9
10
11
// 100% red
mosquitto_pub -t /home/study/light/color/set -m "#FF0000";
// 100% warm white
mosquitto_pub -t /home/study/light/color/set -m "0,0,0,0,255";
// 300 mired color temperature
mosquitto_pub -t /home/study/light/color/set -m "M300";
// 4000 kelvin color temperature
mosquitto_pub -t /home/study/light/color/set -m "K4000";

Of course you can also use Home Assistant MQTT Light component. The configuration would look like this:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
light:
  - platform: mqtt
    name: 'AI Light TEST'
    state_topic: '/home/study/light/relay/0'
    command_topic: '/home/study/light/relay/0/set'
    payload_on: 1
    payload_off: 0
    rgb_state_topic: '/home/study/light/color'
    rgb_command_topic: '/home/study/light/color/set'
    rgb: true
    optimistic: false
    color_temp: true
    color_temp_command_topic: '/home/study/light/mired/set'
    brightness: true
    brightness_command_topic: '/home/study/light/brightness/set'
    brightness_state_topic: '/home/study/light/brightness'
    white_value: true
    white_value_command_topic: '/home/study/light/channel/3/set'
    white_value_state_topic: '/home/study/light/channel/3'

Either way, flashing custom firmware like ESPurna on a 13€ Sonoff B1 [Ebay] device allows you to first fully control your device (no connections outside your home network if you don’t want to) and second, make it interoperate with other services like Home Assistant, Domoticz, Node-RED or any other MQTT o REST capable services.

After all, I’m talking about Technological Sovereignty.

Source: Sonoff B1, lights and shades – Tinkerman

Don’t Toss That Bulb, It Knows Your Password

Posted on by

As it turns out, giving every gadget you own access to your personal information and Internet connection can lead to unintended consequences. Who knew, right? But if you need yet another example of why trusting your home appliances with your secrets is potentially a bad idea, [Limited Results] is here to make sure you spend the next few hours doubting your recent tech purchases.

In a series of posts on the [Limited Results] blog, low-cost “smart” bulbs are cracked open and investigated to see what kind of knowledge they’ve managed to collect about their owners. Not only was it discovered that bulbs manufactured by Xiaomi, LIFX, and Tuya stored the WiFi SSID and encryption key in plain-text, but that recovering said information from the bulbs was actually quite simple. So next time one of those cheapo smart bulb starts flickering, you might want to take a hammer to it before tossing it in the trash can; you never know where it, and the knowledge it has of your network, might end up.

Regardless of the manufacturer of the bulb, the process to get one of these devices on your network is more or less the same. An application on your smartphone connects to the bulb and provides it with the network SSID and encryption key. The bulb then disconnects from the phone and reconnects to your home network with the new information. It’s a process that at this point we’re all probably familiar with, and there’s nothing inherently wrong with it.

The trouble comes when the bulb needs to store the connection information it was provided. Rather than obfuscating it in some way, the SSID and encryption key are simply stored in plain-text on the bulb’s WiFi module. Recovering that information is just a process of finding the correct traces on the bulb’s PCB (often there are test points which make this very easy), and dumping the chip’s contents to the computer for analysis.

It’s not uncommon for smart bulbs like these to use the ESP8266 or ESP32, and [Limited Results] found that to be the case here. With the wealth of information and software available for these very popular WiFi modules, dumping the firmware binary was no problem. Once the binary was in hand, a little snooping around with a hex editor was all it took to identify the network login information. The firmware dumps also contained information such as the unique hardware IDs used by the “cloud” platforms the bulbs connect to, and in at least one case, the root certificate and RSA private key were found.

On the plus side, being able to buy cheap smart devices that are running easily hackable modules like the ESP makes it easier for us to create custom firmware for them. Hopefully the community can come up with slightly less suspect software, but really just keeping the things from connecting to anything outside the local network would be a step in the right direction.

Source: Don’t Toss That Bulb, It Knows Your Password | Hackaday

Towards reconstructing intelligible speech from the human auditory cortex

Posted on by

To advance the state-of-the-art in speech neuroprosthesis, we combined the recent advances in deep learning with the latest innovations in speech synthesis technologies to reconstruct closed-set intelligible speech from the human auditory cortex. We investigated the dependence of reconstruction accuracy on linear and nonlinear (deep neural network) regression methods and the acoustic representation that is used as the target of reconstruction, including auditory spectrogram and speech synthesis parameters. In addition, we compared the reconstruction accuracy from low and high neural frequency ranges. Our results show that a deep neural network model that directly estimates the parameters of a speech synthesizer from all neural frequencies achieves the highest subjective and objective scores on a digit recognition task, improving the intelligibility by 65% over the baseline method which used linear regression to reconstruct the auditory spectrogram

Source: Towards reconstructing intelligible speech from the human auditory cortex | Scientific Reports

Data Leak in Singapore Exposes HIV Status of 14,000 Locals and Foreign Visitors

Posted on by

Medical records and contact information belonging to thousands of HIV-positive Singaporeans and foreign visitors to the southeast Asian city state have been leaked online, according to an alert issued by the country’s Ministry of Health (MOH).

In a statement on its website, the ministry said the confidential health information of some 14,200 individuals diagnosed with HIV had been exposed.

“The information has been illegally disclosed online,” it said. “We have worked with the relevant parties to disable access to the information.”

Source: Data Leak in Singapore Exposes HIV Status of 14,000 Locals and Foreign Visitors

This is why we don’t like centralised medical databases

Apple: You can’t sue us for slowing down your iPhones because we’re like a contractor in your house

Posted on by

Apple is like a building contractor you hire to redo your kitchen, the tech giant has argued in an attempt to explain why it shouldn’t have to pay customers for slowing down their iPhones.

Addressing a bunch of people trying to sue it for damages, the iGiant’s lawyers told [PDF] a California court this month: “Plaintiffs are like homeowners who have let a building contractor into their homes to upgrade their kitchens, thus giving permission for the contractor to demolish and change parts of the houses.”

They went on: “Any claim that the contractor caused excessive damage in the process sounds in contract, not trespass.”

[…]

In this particular case in the US, the plaintiffs argue that Apple damaged their phones by effectively forcing them to install software updates that were intended to fix the battery issues. They may have “chosen” to install the updates by tapping on the relevant buttons, but they did so after reading misleading statements about what the updates were and what they would do, the lawsuit claims.

Nonsense! says Apple. You invited us into your house. We did some work. Sorry you don’t like the fact that we knocked down the wall to the lounge and installed a new air vent through the ceiling, but that’s just how it is.

[…]

But that’s not the only disturbing image to emerge from this lawsuit. When it was accused of damaging people’s property by ruining their batteries, Apple argued – successfully – in court that consumers can’t reasonably expect their iPhone batteries to last longer than a year, given that its battery warranty runs out after 12 months. That would likely come as news to iPhone owners who don’t typically expect to spend $1,000 on a phone and have it die on them a year later.

Call of Duty

Apple has also argued that it’s not under any obligation to tell people buying its products about how well its batteries and software function. An entire section of the company’s motion to dismiss this latest lawsuit is titled: “Apple had no duty to disclose the facts regarding software capability and battery capacity.”

Of course, the truth is that Apple knows that it screwed up – and screwed up badly. Which is why last year it offered replacement batteries for just $29 rather than the usual $79. Uptake of the “program” was so popular that analysts say it has accounted for a significant drop-off in new iPhone purchases.

[…]

Ultimately of course, Apple remains convinced that it’s not really your phone at all: Cupertino has been good enough to allow you to use its amazing technology, and all you had to do was pay it a relatively small amount of money.

We should all be grateful that Apple lets us use our iPhones at all. And if it wants to slow them down, it can damn well slow them down without having to tell you because you wouldn’t understand the reasons why even if it bothered to explain them to you.

Source: Apple: You can’t sue us for slowing down your iPhones because you, er, invited us into, uh, your home… we can explain • The Register

This kind of reasoning beggars belief

Apple temporarily disables group FaceTime to fix a bug that lets you eavesdrop on your contacts

Posted on by

There was chaos on the internet late last night after 9to5Mac discovered a bug in Apple’s FaceTime video calling app that let you hear other person’s voice even before they answered your call. According to the report, a user running iOS 12.1 could potentially exploit the vulnerability to eavesdrop on others through a group FaceTime call.

What’s more, The Verge noted if the recipient ignored or dismissed the call using the power button, their video feed was streamed to the caller.

Source: Apple temporarily disables group FaceTime to fix a bug that lets you eavesdrop on your contacts

Google’s Sidewalk Labs Plans to Package and Sell Location Data on Millions of Cellphones

Posted on by

Most of the data collected by urban planners is messy, complex, and difficult to represent. It looks nothing like the smooth graphs and clean charts of city life in urban simulator games like “SimCity.” A new initiative from Sidewalk Labs, the city-building subsidiary of Google’s parent company Alphabet, has set out to change that.

The program, known as Replica, offers planning agencies the ability to model an entire city’s patterns of movement. Like “SimCity,” Replica’s “user-friendly” tool deploys statistical simulations to give a comprehensive view of how, when, and where people travel in urban areas. It’s an appealing prospect for planners making critical decisions about transportation and land use. In recent months, transportation authorities in Kansas City, Portland, and the Chicago area have signed up to glean its insights. The only catch: They’re not completely sure where the data is coming from.

Typical urban planners rely on processes like surveys and trip counters that are often time-consuming, labor-intensive, and outdated. Replica, instead, uses real-time mobile location data. As Nick Bowden of Sidewalk Labs has explained, “Replica provides a full set of baseline travel measures that are very difficult to gather and maintain today, including the total number of people on a highway or local street network, what mode they’re using (car, transit, bike, or foot), and their trip purpose (commuting to work, going shopping, heading to school).”

To make these measurements, the program gathers and de-identifies the location of cellphone users, which it obtains from unspecified third-party vendors. It then models this anonymized data in simulations — creating a synthetic population that faithfully replicates a city’s real-world patterns but that “obscures the real-world travel habits of individual people,” as Bowden told The Intercept.

The program comes at a time of growing unease with how tech companies use and share our personal data — and raises new questions about Google’s encroachment on the physical world.

If Sidewalk Labs has access to people’s unique paths of movement prior to making its synthetic models, wouldn’t it be possible to figure out who they are, based on where they go to sleep or work?

Last month, the New York Times revealed how sensitive location data is harvested by third parties from our smartphones — often with weak or nonexistent consent provisions. A Motherboard investigation in early January further demonstrated how cell companies sell our locations to stalkers and bounty hunters willing to pay the price.

For some, the Google sibling’s plans to gather and commodify real-time location data from millions of cellphones adds to these concerns. “The privacy concerns are pretty extreme,” Ben Green, an urban technology expert and author of “The Smart Enough City,” wrote in an email to The Intercept. “Mobile phone location data is extremely sensitive.” These privacy concerns have been far from theoretical. An Associated Press investigation showed that Google’s apps and website track people even after they have disabled the location history on their phones. Quartz found that Google was tracking Android users by collecting the addresses of nearby cellphone towers even if all location services were turned off. The company has also been caught using its Street View vehicles to collect the Wi-Fi location data from phones and computers.

This is why Sidewalk Labs has instituted significant protections to safeguard privacy, before it even begins creating a synthetic population. Any location data that Sidewalk Labs receives is already de-identified (using methods such as aggregation, differential privacy techniques, or outright removal of unique behaviors). Bowden explained that the data obtained by Replica does not include a device’s unique identifiers, which can be used to uncover someone’s unique identity.

However, some urban planners and technologists, while emphasizing the elegance and novelty of the program’s concept, remain skeptical about these privacy protections, asking how Sidewalk Labs defines personally identifiable information. Tamir Israel, a staff lawyer at the Canadian Internet Policy & Public Interest Clinic, warns that re-identification is a rapidly moving target. If Sidewalk Labs has access to people’s unique paths of movement prior to making its synthetic models, wouldn’t it be possible to figure out who they are, based on where they go to sleep or work? “We see a lot of companies erring on the side of collecting it and doing coarse de-identifications, even though, more than any other type of data, location data has been shown to be highly re-identifiable,” he added. “It’s obvious what home people leave and return to every night and what office they stop at every day from 9 to 5 p.m.” A landmark study uncovered the extent to which people could be re-identified from seemingly-anonymous data using just four time-stamped data points of where they’ve previously been.

Source: Google’s Sidewalk Labs Plans to Package and Sell Location Data on Millions of Cellphones

Firefox cracks down on creepy web trackers, holds supercookies over fire whilst Chrome kills ad blockers

Posted on by

The Mozilla Foundation has announced its intent to reduce the ability of websites and other online services to track users of its Firefox browser around the internet.

At this stage, Moz’s actions are baby steps. In support of its decision in late 2018 to reduce the amount of tracking it permits, the organisation has now published a tracking policy to tell people what it will block.

Moz said the focus of the policy is to bring the curtain down on tracking techniques that “cannot be meaningfully understood or controlled by users”.

Notoriously intrusive tracking techniques allow users to be followed and profiled around the web. Facebook planting trackers wherever a site has a “Like” button is a good example. A user without a Facebook account can still be tracked as a unique individual as they visit different news sites.

Mozilla’s policy said these “stateful identifiers are often used by third parties to associate browsing across multiple websites with the same user and to build profiles of those users, in violation of the user’s expectation”. So, out they go.

Source: Mozilla security policy cracks down on creepy web trackers, holds supercookies over fire • The Register

I’m pretty sure which browser you should be using

94% of Dutch worried about their privacy

Posted on by

Bescherming van de privacy is een breed gedeelde zorg. Maar liefst 94 procent van de Nederlands maakt zich zorgen over de bescherming van zijn persoonsgegevens. Een op drie mensen maakt zich zelfs veel of zeer veel zorgen. Dat blijkt uit onderzoek dat de Autoriteit Persoonsgegevens (AP) liet doen in het kader van de Dag van de Privacy.

Er zijn vooral zorgen over misbruik van (een kopie van) het identiteitsbewijs, organisaties die hun online zoekgedrag volgen en hen volgen via het wifi-signaal van hun mobiele telefoon.

Slechts 12 procent zegt wel eens gebruik te hebben gemaakt van een privacyrecht. Mensen weten volgens de toezichthouder niet hoe ze dat moeten doen, vinden het gedoe of niet belangrijk genoeg. Het recht op dataportabiliteit en het recht op een menselijke blik bij geautomatiseerde besluiten zijn de minst bekende rechten.

Gevraagd wat mensen doen als hun rechten worden geschonden, zegt 62 procent eerst contact op te nemen met de organisaties, 59 procent van de ondervraagden zegt een klacht in te dienen bij de AP.

Source: ‘Nederland maakt zich zorgen over privacy’ – Emerce

South Korea rules pre-installed phone bloatware must be deletable

Posted on by

Smartphone users in South Korea will soon be able to have the option of deleting unnecessary pre-installed bloatware, thanks to new industry guidelines commencing in April.

“The move aims to rectify an abnormal practice that causes inconvenience to smartphone users and causes unfair competition among industry players,” said the Ministry of Science, ICT and Future Planning, in a press release.

The measure will also help give users more data storage and improve battery life, said the ministry.

Under the new guidelines, telcos are required to make most of their pre-installed apps deletable except for four necessary items related to Wi-Fi connectivity, near-field communication (NFC), the customer service center and the app store.

For example, Samsung’s Galaxy S4 released by SK Telecom has a total of 80 apps pre-installed, including 25 apps loaded by the telco, 39 by Samsung and 16 by the OS provider Google, noted Yonhap News. When the new guidelines kick in, at least half of those apps can be deleted, it added.

Source: South Korea rules pre-installed phone bloatware must be deletable | ZDNet

This 3D Printing Technique Is 100 Times Faster Than Standard 3D Printers

Posted on by

A new 3D-printing technique could render a three-dimensional object in minutes instead of hours—at up to 100 times current speeds. The experimental approach uses a vat of resin and some clever tricks with UV and blue LED lights (no lasers needed) to accelerate the printing process.

The technique looks almost like a time-reverse film loop of an object dissolving in a reservoir of acid. But instead of acid, this reservoir contains a specially-designed resin that hardens when exposed to a particular shade of blue light. Crucially, that hardening (the technical term is polymerization) does not take place in the presence of a certain wavelength of UV light.

The resin is also particularly absorbent at the wavelengths of both the blue and UV light. So the intensity of UV or blue light going in translates directly to the depth to which light will penetrate into the resin bath. The brighter the light beam, the further it penetrates and the further its effects (whether inhibiting polymerization in the case of UV light, or causing it in the case of blue light) will be felt in the bath along that particular light path.

Timothy Scott, associate professor of chemical engineering at the University of Michigan, says the way to get a 3D-printed object out of this process is to send UV light through a glass-bottomed basin of resin. Then, at the same time, through that same glass window, send patterns of bright and dim blue light.

If this printing process used only the blue light, it would immediately harden the first bit of resin it encounters in the basin—the stuff just inside the glass. And so each successive layer of the object to be printed would need to be scraped or pulled off the window’s surface—a time-consuming and potentially destructive process.

“We use the [UV] wavelength to prevent the resin from polymerizing against the projection window,” Scott says. “But we can change the intensity of the inhibiting wavelength, that in turn can thicken up…the region that doesn’t polymerize. We can go to hundreds of microns comfortably, approaching or even exceeding a millimeter, so that’s getting quite thick. We can do that across not only the entire region of our bath, but we can do it selectively. By, again, patterning the intensity that we’re projecting into the vat.”

Which is why the UV light, perhaps the key innovation of the new research, potentially streamlines the entire light-resin 3D-printing process, also called 3D stereolithography.

To be clear, other 3D-stereolithography printing processes and even startup companies are out there in the world. What’s new with the Michigan group’s research (published in Science Advances earlier this month) is the UV light inhibitor that not only prevents the hardened resin from sticking to the window but also can be used in concert with the blue light to sculpt 3D surfaces and contours of hardened resin in the bath.

In a sense, Scott says, the new stereolithography process is really one of the very first truly 3D printing processes—in that it prints not just a series of single 2D layers but rather entire 3D wedges of material in one pass.

“That is straight-up unique, the ability to pattern a volume,” Scott says. “Patterning in 2D is easy, patterning in 3D is nontrivial.”

Source: This 3D Printing Technique Is 100 Times Faster Than Standard 3D Printers – IEEE Spectrum

Deepmind AI beats top Starcraft pros

Posted on by

Now, we introduce our StarCraft II program AlphaStar, the first Artificial Intelligence to defeat a top professional player. In a series of test matches held on 19 December, AlphaStar decisively beat Team Liquid’s Grzegorz “MaNa” Komincz, one of the world’s strongest professional StarCraft players, 5-0, following a successful benchmark match against his team-mate Dario “TLO” Wünsch. The matches took place under professional match conditions on a competitive ladder map and without any game restrictions.

Although there have been significant successes in video games such as Atari, Mario, Quake III Arena Capture the Flag, and Dota 2, until now, AI techniques have struggled to cope with the complexity of StarCraft. The best results were made possible by hand-crafting major elements of the system, imposing significant restrictions on the game rules, giving systems superhuman capabilities, or by playing on simplified maps. Even with these modifications, no system has come anywhere close to rivalling the skill of professional players. In contrast, AlphaStar plays the full game of StarCraft II, using a deep neural network that is trained directly from raw game data by supervised learning and reinforcement learning.

Demonstration game against MaNa

Source: AlphaStar: Mastering the Real-Time Strategy Game StarCraft II | DeepMind

Just keep slurping: HMRC adds two million taxpayers’ voices to biometric database – but people are starting to opt-out, now that they can

Posted on by

HMRC’s database of Brits’ voiceprints has grown by 2 million since June – but campaign group Big Brother Watch has claimed success as 160,000 people turned the taxman’s requests down.

The Voice ID scheme, which requires taxpayers to say a key phrase that is recorded to create a digital signature, was introduced in January 2017. In the 18 months that followed, HMRC scooped up some 5.1 million people’s voiceprints this way.

Since then, another 2 million records have been collected, according to a Freedom of Information request from Big Brother Watch.

That is despite the group having challenged the lawfulness of the system in June 2018, arguing that users hadn’t been given enough information on the scheme, how to opt in or out, or details on when or how their data would be deleted.

Under the GDPR, there are certain demands on organisations that process biometric data. These require a person to give “explicit consent” that is “freely given, specific, informed and unambiguous”.

Off the back of the complaint, the Information Commissioner’s Office launched an investigation, and Big Brother Watch said the body would soon announce what action it will take.

Meanwhile, HMRC has rejigged the recording so it offers callers a clear way to opt out of the scheme – previously, as perm sec Jon Thompson admitted in September, it was not clear how users could do this.

Big Brother Watch said that this, and the publicity around the VoiceID scheme, has led to a “backlash” as people call on HMRC to delete their Voice IDs. FoI responses show 162,185 people have done so to date.

“It is a great success for us that HMRC has finally allowed taxpayers to delete their voiceprints and that so many thousands of people are reclaiming their rights by getting their Voice IDs deleted,” said the group’s director, Silkie Carlo.

Source: Just keep slurping: HMRC adds two million taxpayers’ voices to biometric database • The Register

Program allows ordinary digital camera to see round corners

Posted on by

In a demonstration of “computational periscopy” a US team at Boston University showed they could see details of objects hidden from view by analysing shadows they cast on a nearby wall.

Vivek Goyal, an electrical engineer at the university, said that while the work had clear implications for surveillance he hoped it would lead to robots that could navigate better and boost the safety of driverless cars.

He said: “I’m not especially excited by surveillance, I don’t want to be doing creepy things, but being able to see that there’s a child on the other side of a parked car, or see a little bit around the corner of an intersection could have a significant impact on safety.”

The problem of how to see round corners has occupied modern researchers for at least a decade. And while scientists have made good progress in the field, the equipment used so far has been highly specialised and expensive.

In the latest feat, Goyal and his team used a standard digital camera and a mid-range laptop. “We didn’t use any sophisticated hardware. This is just an ordinary camera and we are all carrying these around in our pockets,” he said.

The researchers, writing in the journal Nature, describe how they pieced together hidden scenes by pointing the digital camera at the vague shadows they cast on a nearby wall. If the wall had been a mirror the task would have been easy, but a matt wall scatters light in all directions, so the reflected image is nothing but a blur. Goyal said: “In essence, computation can turn a matt wall into a mirror.”

They found that when an object blocked part of the hidden scene, their algorithms could use the combination of light and shade at different points on the wall to reconstruct what lay round the corner. In tests, the program pieced together hidden images of video game characters – including details such as their eyes and mouths – along with coloured strips and the letters “BU”.

Given the relative simplicity of the program and equipment, Goyal believes it could be possible for humans to learn the same trick. In a draft blog written for Nature, he said: “It is even conceivable for humans to be able to learn to see around corners with their own eyes; it does not require anything superhuman.”

Source: Program allows ordinary digital camera to see round corners | Science | The Guardian

We may finally know what causes Alzheimer’s – and how to stop it

Posted on by

If you bled when you brushed your teeth this morning, you might want to get that seen to. We may finally have found the long-elusive cause of Alzheimer’s disease: Porphyromonas gingivalis, the key bacteria in chronic gum disease.

That’s bad, as gum disease affects around a third of all people. But the good news is that a drug that blocks the main toxins of P. gingivalis is entering major clinical trials this year, and research published today shows it might stop and even reverse Alzheimer’s. There could even be a vaccine.

Alzheimer’s is one of the biggest mysteries in medicine. As populations have aged, dementia has skyrocketed to become the fifth biggest cause of death worldwide. Alzheimer’s constitutes some 70 per cent of these cases and yet, we don’t know what causes it.

Bacteria in the brain

The disease often involves the accumulation of proteins called amyloid and tau in the brain, and the leading hypothesis has been that the disease arises from defective control of these two proteins.

But research in recent years has revealed that people can have amyloid plaques without having dementia. So many efforts to treat Alzheimer’s by moderating these proteins have failed that the hypothesis has been seriously questioned.

Read more: Here’s how to avoid gum disease

However evidence has been growing that the function of amyloid proteins may be as a defence against bacteria, leading to a spate of recent studies looking at bacteria in Alzheimer’s, particularly those that cause gum disease, which is known to be a major risk factor for the condition.

Bacteria involved in gum disease and other illnesses have been found after death in the brains of people who had Alzheimer’s, but until now, it hasn’t been clear whether these bacteria caused the disease or simply got in via brain damage caused by the condition.

Gum disease link

Multiple research teams have been investigating P. gingivalis, and have so far found that it invades and inflames brain regions affected by Alzheimer’s; that gum infections can worsen symptoms in mice genetically engineered to have Alzheimer’s; and that it can cause Alzheimer’s-like brain inflammation, neural damage, and amyloid plaques in healthy mice.

“When science converges from multiple independent laboratories like this, it is very compelling,” says Casey Lynch of Cortexyme, a pharmaceutical firm in San Francisco, California.

In new study, Cortexyme have now reported finding the toxic enzymes – called gingipains – that P. gingivalis uses to feed on human tissue in 96 per cent of the 54 Alzheimer’s brain samples they looked at, and found the bacteria themselves in all three Alzheimer’s brains whose DNA they examined.

The bacteria and its enzymes were found at higher levels in those who had experienced worse cognitive decline, and had more amyloid and tau accumulations. The team also found the bacteria in the spinal fluid of living people with Alzheimer’s, suggesting that this technique may provide a long-sought after method of diagnosing the disease.

Source: We may finally know what causes Alzheimer’s – and how to stop it | New Scientist

How to replicate an $86 million license plate recognition and stolen check project in 57 lines of code

Posted on by

The Victoria Police are the primary law enforcement agency of Victoria, Australia. With over 16,000 vehicles stolen in Victoria this past year — at a cost of about $170 million — the police department is experimenting with a variety of technology-driven solutions to crackdown on car theft. They call this system BlueNet.

To help prevent fraudulent sales of stolen vehicles, there is already a VicRoads web-based service for checking the status of vehicle registrations. The department has also invested in a stationary license plate scanner — a fixed tripod camera which scans passing traffic to automatically identify stolen vehicles.

Don’t ask me why, but one afternoon I had the desire to prototype a vehicle-mounted license plate scanner that would automatically notify you if a vehicle had been stolen or was unregistered. Understanding that these individual components existed, I wondered how difficult it would be to wire them together.

But it was after a bit of googling that I discovered the Victoria Police had recently undergone a trial of a similar device, and the estimated cost of roll out was somewhere in the vicinity of $86,000,000. One astute commenter pointed out that the $86M cost to fit out 220 vehicles comes in at a rather thirsty $390,909 per vehicle.

Surely we can do a bit better than that.

Existing stationary license plate recognition systems

The Success Criteria

Before getting started, I outlined a few key requirements for product design.

Requirement #1: The image processing must be performed locally

Streaming live video to a central processing warehouse seemed the least efficient approach to solving this problem. Besides the whopping bill for data traffic, you’re also introducing network latency into a process which may already be quite slow.

Although a centralized machine learning algorithm is only going to get more accurate over time, I wanted to learn if an local on-device implementation would be “good enough”.

Requirement #2: It must work with low quality images

Since I don’t have a Raspberry Pi camera or USB webcam, so I’ll be using dashcam footage — it’s readily available and an ideal source of sample data. As an added bonus, dashcam video represents the overall quality of footage you’d expect from vehicle mounted cameras.

Requirement #3: It needs to be built using open source technology

Relying upon a proprietary software means you’ll get stung every time you request a change or enhancement — and the stinging will continue for every request made thereafter. Using open source technology is a no-brainer.

My solution

At a high level, my solution takes an image from a dashcam video, pumps it through an open source license plate recognition system installed locally on the device, queries the registration check service, and then returns the results for display.

The data returned to the device installed in the law enforcement vehicle includes the vehicle’s make and model (which it only uses to verify whether the plates have been stolen), the registration status, and any notifications of the vehicle being reported stolen.

If that sounds rather simple, it’s because it really is. For example, the image processing can all be handled by the openalpr library.

This is really all that’s involved to recognize the characters on a license plate:

A Minor Caveat
Public access to the VicRoads APIs is not available, so license plate checks occur via web scraping for this prototype. While generally frowned upon — this is a proof of concept and I’m not slamming anyone’s servers.

Here’s what the dirtiness of my proof-of-concept scraping looks like:

Results

I must say I was pleasantly surprised.

I expected the open source license plate recognition to be pretty rubbish. Additionally, the image recognition algorithms are probably not optimised for Australian license plates.

The solution was able to recognise license plates in a wide field of view.

Annotations added for effect. Number plate identified despite reflections and lens distortion.

Although, the solution would occasionally have issues with particular letters.

Incorrect reading of plate, mistook the M for an H

But … the solution would eventually get them correct.

A few frames later, the M is correctly identified and at a higher confidence rating

As you can see in the above two images, processing the image a couple of frames later jumped from a confidence rating of 87% to a hair over 91%.

I’m confident, pardon the pun, that the accuracy could be improved by increasing the sample rate, and then sorting by the highest confidence rating. Alternatively a threshold could be set that only accepts a confidence of greater than 90% before going on to validate the registration number.

Those are very straight forward code-first fixes, and don’t preclude the training of the license plate recognition software with a local data set.

The $86,000,000 Question

To be fair, I have absolutely no clue what the $86M figure includes — nor can I speak to the accuracy of my open source tool with no localized training vs. the pilot BlueNet system.

I would expect part of that budget includes the replacement of several legacy databases and software applications to support the high frequency, low latency querying of license plates several times per second, per vehicle.

On the other hand, the cost of ~$391k per vehicle seems pretty rich — especially if the BlueNet isn’t particularly accurate and there are no large scale IT projects to decommission or upgrade dependent systems.

Future Applications

While it’s easy to get caught up in the Orwellian nature of an “always on” network of license plate snitchers, there are many positive applications of this technology. Imagine a passive system scanning fellow motorists for an abductors car that automatically alerts authorities and family members to their current location and direction.

Teslas vehicles are already brimming with cameras and sensors with the ability to receive OTA updates — imagine turning these into a fleet of virtual good samaritans. Ubers and Lyft drivers could also be outfitted with these devices to dramatically increase the coverage area.

Using open source technology and existing components, it seems possible to offer a solution that provides a much higher rate of return — for an investment much less than $86M.

Source: How I replicated an $86 million project in 57 lines of code

Tikkie: IBAN-numbers users exposed (Dutch)

Posted on by

De populaire betaal-app Tikkie biedt de mogelijkheid om geld over te boeken naar andere Tikkie-gebruikers op basis van hun 06-nummer. Daardoor was het mogelijk om de IBAN-nummers van vele nietsvermoedende Tikkie-gebruikers te achterhalen, met het gevaar voor identiteitsfraude en phishing.

Dat blijkt uit onderzoek van RTL Nieuws. ABN Amro bevestigt de kwetsbaarheid en heeft de nieuwe functie, Tikkie Pay, tijdelijk offline gehaald. “Bedankt voor de oplettendheid”, aldus de woordvoerder.

IBAN-nummers

Tikkie, dat 4 miljoen gebruikers heeft, toonde met zijn nieuwe functie alle gebruikers uit jouw contactenlijst die hun 06-nummer aan Tikkie hebben gekoppeld. Je kon op een naam drukken, vervolgens een bedrag overmaken en net voor de overboeking de Tikkie annuleren. In de omschrijving van de overboeking zag je dan het IBAN-nummer van de ontvanger, zonder dat diegene daar weet van heeft.

Source: Privacylek bij Tikkie: IBAN-nummers gebruikers in te zien | RTL Nieuws

Smartians – turn old tech things into smart things by pushing, pulling, prodding and turning: looking for funding

Posted on by

Why replace your things just because they’re not state-of-the-art? Smartians are cloud-connected motors that breathe new life into the things around you.

<iframe src=”https://player.vimeo.com/video/294150597″ width=”640″ height=”360″ frameborder=”0″ allowfullscreen></iframe>

Very clever, I hope they get funded!

 

Source: Smartians | FROLIC studio |

Heads up: Debian’s package manager is APT for root-level malware injection… Fix out now to thwart MITM hijacks

Posted on by

The Debian Project has patched a security flaw in its software manager Apt that can be exploited by network snoops to execute commands as root on victims’ boxes as they update or install packages.

The Linux distro’s curators have pushed out an fix to address CVE-2019-3462, a vulnerability uncovered and reported by researcher Max Justicz.

The flaw is related to the way Apt and apt-get handle HTTP redirects when downloading packages. Apt fetches packages over plain-old HTTP, rather than a more secure HTTPS connection, and uses cryptographic signatures to check whether the downloaded contents are legit and haven’t been tampered with.

This unfortunately means a man-in-the-middle (MITM) miscreant who was able to intercept and tamper with a victim’s network connection could potentially inject a redirect into the HTTP headers to change the URL used to fetch the package.

And the hacker would be able to control the hashes used by Apt to check the downloaded package, passing the package manager legit values to masquerade the fetched malware as sanctioned software.

All in all, users can be fed malware that’s run as root during installation, allowing it to commandeer the machine.

[…]

As an added wrinkle, Apt is updated by Apt itself. And seeing as the update mechanism is insecure, folks need to take extra steps to install the security fix securely. Admins will want to first disable redirects (see below) and then go through the usual apt update and upgrade steps.

$ sudo apt update -o Acquire::http::AllowRedirect=false
$ sudo apt upgrade -o Acquire::http::AllowRedirect=false

Source: Heads up: Debian’s package manager is APT for root-level malware injection… Fix out now to thwart MITM hijacks • The Register

Wow, fancy that. Web ad giant Google to block ad-blockers in Chrome. For safety, apparently

Posted on by

Google engineers have proposed changes to the open-source Chromium browser that will break content-blocking extensions, including various ad blockers.

Adblock Plus will most likely not be affected, though similar third-party plugins will, for reasons we will explain. The drafted changes will also limit the capabilities available to extension developers, ostensibly for the sake of speed and safety. Chromium forms the central core of Google Chrome, and, soon, Microsoft Edge.

In a note posted Tuesday to the Chromium bug tracker, Raymond Hill, the developer behind uBlock Origin and uMatrix, said the changes contemplated by the Manifest v3 proposal will ruin his ad and content blocking extensions, and take control of content away from users.

Content blockers may be used to hide or black-hole ads, but they have broader applications. They’re predicated on the notion that users, rather than anyone else, should be able to control how their browser presents and interacts with remote resources.

Manifest v3 refers to the specification for browser extension manifest files, which enumerate the resources and capabilities available to browser extensions. Google’s stated rationale for making the proposed changes, cutting off blocking plugins, is to improve security, privacy and performance, and supposedly to enhance user control.

Source: Wow, fancy that. Web ad giant Google to block ad-blockers in Chrome. For safety, apparently • The Register

uBlock origin is not only an ad blocker but also an important privacy and security tool

Too Good To Go app – anti supermarket food wastage, in 9 EU countries, including NL

Posted on by

Supermarkets create cheap “magic boxes” with end of life food in them. You can see where to pick them up on the app. Jumbo NL has started a pilot in 13 shops.

 

Het van oorsprong Deense initiatief Too Good To Go heeft na één jaar in Nederland meer dan 200.000 maaltijden gered van de vuilnisbak. De gelijknamige app heeft ondertussen al meer dan 250.000 geregistreerde gebruikers en meer dan 1000 partners met dekking in alle provincies in Nederland.

Op de kaart of in de lijst in de app kunnen consumenten bekijken welke locaties iets lekkers voor ze klaar hebben liggen tegen sluitingstijd. Vervolgens bestellen en betalen zij direct in de app.

Sinds gisteren is bij Jumbo een pilot met Too Goo To Go in 13 winkels gestart. De pilot duurt een maand en is de eerste stap op weg naar een mogelijke landelijke uitrol.

Gebruikers zien in de Too Good To Go app welke Jumbo winkels een Magic Box aanbieden. Ze rekenen deze vervolgens af via de app en kunnen de verrassingsbox binnen een afgesproken tijdsslot ophalen in de winkel. De prijs is altijd een derde van de daadwerkelijke waarde: een box met een waarde van 15 euro kost dus slechts 5 euro.

Deelnemers aan de pilot zijn elf winkels in Amsterdam – waaronder de City winkels – en Foodmarkt Amsterdam en een City in Groningen.

Winkels bepalen zelf hoe ze de box samenstellen, waarbij beschikbaarheid en variatie belangrijke criteria zijn.

Vanaf vandaag is de stad Wageningen ook als locatie toegevoegd aan de app. Om de impact van de app van Too Good To Go op het consumentengedrag te meten en om te bepalen wat de volgende stukjes van de puzzel moeten worden, start Too Good To Go in samenwerking met Wageningen University & Research een onderzoek naar de verandering in bewustwording en het gedrag rond voedselverspilling.

Too Good To Go is al actief in negen Europese landen.

Source: 250.000 gebruikers voor app Too Good To Go – Emerce

Massive Oklahoma Government Data Leak Exposes 7 Years of FBI Investigations – unsecured rsync

Posted on by

Last December, a whopping 3 terabytes of unprotected data from the Oklahoma Securities Commission was uncovered by Greg Pollock, a researcher with cybersecurity firm UpGuard. It amounted to millions of files, many on sensitive FBI investigations, all of which were left wide open on a server with no password, accessible to anyone with an internet connection, Forbes can reveal.

“It represents a compromise of the entire integrity of the Oklahoma department of securities’ network,” said Chris Vickery, head of research at UpGuard, which is revealing its technical findings on Wednesday. “It affects an entire state level agency. … It’s massively noteworthy.”

A breach back to the ’80s

The Oklahoma department regulates all financial securities business happening in the state. It may be little surprise there was leaked information on FBI cases. But the amount and variety of data astonished Vickery and Pollock.

Vickery said the FBI files contained “all sorts of archive enforcement actions” dating back seven years (the earliest file creation date was 2012). The documents included spreadsheets with agent-filled timelines of interviews related to investigations, emails from parties involved in myriad cases and bank transaction histories. There were also copies of letters from subjects, witnesses and other parties involved in FBI investigations.

[…]

Just as concerning, the leak also included email archives stretching back 17 years, thousands of social security numbers and data from the 1980s onwards.

[…]

After Vickery and Pollock disclosed the breach, they informed the commission it had mistakenly left open what’s known as an rsync server. Such servers are typically used to back up large batches of data and, if that information is supposed to be secure, should be protected by a username and password.

There were other signs of poor security within the leaked data. For instance, passwords for computers on the Oklahoma government’s network were also revealed. They were “not complicated,” quipped Chris Vickery, head of research on the UpGuard team. In one of the more absurd choices made by the department, it had stored an encrypted version of one document in the same file folder as a decrypted version. Passwords for remote access to agency computers were also leaked.

This is the latest in a series of incidents involving rsync servers. In December, UpGuard revealed that Level One Robotics, a car manufacturing supply chain company, was exposing information in the same way as the Oklahoma government division. Companies with data exposed in that event included Volkswagen, Chrysler, Ford, Toyota, General Motors and Tesla.

For whatever reason, governments and corporate giants alike still aren’t aware how easy it is for hackers to constantly scan the Web for such leaks. Starting with basics like passwords would help them keep their secrets secure.

Source: Massive Oklahoma Government Data Leak Exposes 7 Years of FBI Investigations

Let’s Encrypt ends TLS-SNI-01 validation support

Posted on by

Let’s Encrypt allows subscribers to validate domain control using any one of a few different validation methods. For much of the time Let’s Encrypt has been operating, the options were “DNS-01”, “HTTP-01”, and “TLS-SNI-01”. We recently introduced the “TLS-ALPN-01” method. Today we are announcing that we will end all support for the TLS-SNI-01 validation method on February 13, 2019.

In January of 2018 we disabled the TLS-SNI-01 domain validation method for most subscribers due to a vulnerability enabled by some shared hosting infrastructure 1.1k. We provided temporary exceptions for renewals and for a small handful of hosting providers in order to smooth the transition to DNS-01 and HTTP-01 validation methods. Most subscribers are now using DNS-01 or HTTP-01.

If you’re still using TLS-SNI-01, please switch to one of the other validation methods as soon as possible. We will also attempt to contact subscribers who are still using TLS-SNI-01, if they provided contact information.

We apologize for any inconvenience but we believe this is the right thing to do for the integrity of the Web PKI.

https://community.letsencrypt.org/t/february-13-2019-end-of-life-for-all-tls-sni-01-validation-support/74209