5 ms·

It seems like most of these devices (example: https://hackaday.com/?p=683252 https://hackaday.com/?p=683252) have a fixed and unusual USB vendor+product ID that

It seems like most of these devices (example: https://hackaday.com/?p=683252 https://hackaday.com/?p=683252) have a fixed and unusual USB vendor+product ID that will surely come up in the system log.

lxgr 2y ago
People could always just use a device that physically moves the mouse around or pushes a button on the keyboard instead. Companies could of course react by requiring employees to solve a captcha every once in a while to work around that, but here's a controversial thought: What if the "proof of work" that companies use to "detect employee presence" was just... Actually accomplished work? That should solve the problem at the correct layer. Unless, of course, objectives and goals are poorly defined, and requiring that would result in huge embarrassment to the management layers above the mouse wigglers.
- vundercind 2y ago
  What’s going to happen is they’re gonna ask a Microsoft-provided AI whether employees’ data from Microsoft Recall looks like work, on a scale, and fire the lowest percent or two of that scale every quarter. If running it over data from all the days is too expensive, they’ll pick a day or two to sample and you’ll be retained or fired based on that. At least I’ve been assuming that’s the main reason MS is making Recall. To issue spyware-reports on corporate workers.
  - jdalgetty 2y ago
    Yikes!
  - recursive 2y ago
    You just to outspend at least 1% on your generative mouse-movement AI.
    - rozap 2y ago
      Now this is an exciting emerging market
- toss1 2y ago
  THIS is exactly right. I wish I could find it again, but I was impressed by a study of remote work around 2010 that showed that remote work improved the results of teams with good managers and (further) degraded the results of teams with bad managers. A key difference between the good vs bad managers was that the good managers focused on specific tasks being accomplished whereas the bad managers focused on the usual indicators of work such as as time of butts being in seats, which has only an incidental and often inverse relation to quantity or quantity of work produced. This 'tech' is simply attempting to enable bad management with remote workforces. A good sign that remote is winning the war and here to stay. A bad sign that bad mgt will make it suck more (at least until AI eliminates those jobs for good).
- TuringNYC 2y ago
  Not familiar with WF but I've seen large corps in the past where HR mandates N number of reportees to get promoted. It goes up the chain, so your boss is pushing you to hire and you are pushing to hire, etc. The problem is then, people get hired but there is no work for them, especially during lean times. There is a wide chasm between well-run large companies and poorly-run ones. You can look at sector-level revenue/employee to get an idea of efficiency/waste.
  - lxgr 2y ago
    Empire building is a well-known phenomenon in organizations, but that still begs the question: Does it matter whether employees largely hired and retained for payroll metrics actually wiggle their mice or not?
FooBarBizBazz 2y ago
One obvious alternative is to use a real optical mouse, but move the surface next to the sensor. This would be easy with any of several motorized mechanisms, e.g. a turntable -- but that would be bulky, most people do not actually have a turntable, and many cheap implementations would make noise. That makes me wonder what would happen with an optical mouse atop a tablet, displaying a slowly-panning video of a texture. However, there are a few obstacles. First, Google tells me that optical mice have frame rates of 1.5 to 5 kHz. That's very fast. It probably makes the optical flow problem easier, since there's less change between frames. But it's also much faster than the, say, 60 Hz refresh of a tablet's screen. I don't know what a bunch of "stairsteps" in the position signal would look like to the mouse. Second, there's the issue of wavelength. Google brings me optical mouse sensors with sensitivities in the 832 - 865 nm range, i.e., in the infrared. Visible red is around 740 nm. Looking at this chart [1] and extrapolating just a tiny bit, it seems that LCDs have almost no emissivity around 850 nm. So this probably would not work. Whether an LCD with an infrared backlight would work is left as a question for another time. What might be easier still, would be to somehow hack the USB device itself to spoof the vendor and product ID. I don't know how to do that myself, but it might be easy for anyone who's familiar with USB device development. Of course, this is a silly arms race that it wouldn't actually be worth participating in, but it's amusing to consider what you could do. I heard a story of an employee health plan that issued pedometers and issued cash rewards for higher step counts. In short order, one employee had put his on a clock pendulum, another had secured the thing to the collar of his hyperactive dog, and a third had hacked the thing to display all nines. [1] https://www.displaymate.com/Spectra_4.html https://www.displaymate.com/Spectra_4.html
- labcomputer 2y ago
  For what it’s worth, the max polling rate for low-speed USB devices (of which mice almost invariably are) is, IIRC, 100 Hz. I would worry less about stair stepping and more about making the speed and acceleration distributions and auto-correlation look natural: a real human-powered mouse won’t be constantly moving (you have to stop the think about the next step, use the keyboard, etc) and when it is moving will tend to follow a rapid move-pause-move pattern. Another telltale would be a mouse that moves naturally but never clicks. Vendor and device ID are the lowest hanging fruit, but I think there are a lot of other giveaways that a mouse jiggler is being used.

It seems like most of these devices (example: https://hackaday.com/?p=683252 https://hackaday.com/?p=683252) have a fixed and unusual USB vendor+product ID that will surely come up in the system log.

lxgr 2y ago

People could always just use a device that physically moves the mouse around or pushes a button on the keyboard instead. Companies could of course react by requiring employees to solve a captcha every once in a while to work around that, but here's a controversial thought: What if the "proof of work" that companies use to "detect employee presence" was just... Actually accomplished work? That should solve the problem at the correct layer. Unless, of course, objectives and goals are poorly defined, and requiring that would result in huge embarrassment to the management layers above the mouse wigglers.

vundercind 2y ago
What’s going to happen is they’re gonna ask a Microsoft-provided AI whether employees’ data from Microsoft Recall looks like work, on a scale, and fire the lowest percent or two of that scale every quarter. If running it over data from all the days is too expensive, they’ll pick a day or two to sample and you’ll be retained or fired based on that. At least I’ve been assuming that’s the main reason MS is making Recall. To issue spyware-reports on corporate workers.
- jdalgetty 2y ago
  Yikes!
- recursive 2y ago
  You just to outspend at least 1% on your generative mouse-movement AI.
  - rozap 2y ago
    Now this is an exciting emerging market
toss1 2y ago
THIS is exactly right. I wish I could find it again, but I was impressed by a study of remote work around 2010 that showed that remote work improved the results of teams with good managers and (further) degraded the results of teams with bad managers. A key difference between the good vs bad managers was that the good managers focused on specific tasks being accomplished whereas the bad managers focused on the usual indicators of work such as as time of butts being in seats, which has only an incidental and often inverse relation to quantity or quantity of work produced. This 'tech' is simply attempting to enable bad management with remote workforces. A good sign that remote is winning the war and here to stay. A bad sign that bad mgt will make it suck more (at least until AI eliminates those jobs for good).
TuringNYC 2y ago
Not familiar with WF but I've seen large corps in the past where HR mandates N number of reportees to get promoted. It goes up the chain, so your boss is pushing you to hire and you are pushing to hire, etc. The problem is then, people get hired but there is no work for them, especially during lean times. There is a wide chasm between well-run large companies and poorly-run ones. You can look at sector-level revenue/employee to get an idea of efficiency/waste.
- lxgr 2y ago
  Empire building is a well-known phenomenon in organizations, but that still begs the question: Does it matter whether employees largely hired and retained for payroll metrics actually wiggle their mice or not?

FooBarBizBazz 2y ago

One obvious alternative is to use a real optical mouse, but move the surface next to the sensor. This would be easy with any of several motorized mechanisms, e.g. a turntable -- but that would be bulky, most people do not actually have a turntable, and many cheap implementations would make noise. That makes me wonder what would happen with an optical mouse atop a tablet, displaying a slowly-panning video of a texture. However, there are a few obstacles. First, Google tells me that optical mice have frame rates of 1.5 to 5 kHz. That's very fast. It probably makes the optical flow problem easier, since there's less change between frames. But it's also much faster than the, say, 60 Hz refresh of a tablet's screen. I don't know what a bunch of "stairsteps" in the position signal would look like to the mouse. Second, there's the issue of wavelength. Google brings me optical mouse sensors with sensitivities in the 832 - 865 nm range, i.e., in the infrared. Visible red is around 740 nm. Looking at this chart [1] and extrapolating just a tiny bit, it seems that LCDs have almost no emissivity around 850 nm. So this probably would not work. Whether an LCD with an infrared backlight would work is left as a question for another time. What might be easier still, would be to somehow hack the USB device itself to spoof the vendor and product ID. I don't know how to do that myself, but it might be easy for anyone who's familiar with USB device development. Of course, this is a silly arms race that it wouldn't actually be worth participating in, but it's amusing to consider what you could do. I heard a story of an employee health plan that issued pedometers and issued cash rewards for higher step counts. In short order, one employee had put his on a clock pendulum, another had secured the thing to the collar of his hyperactive dog, and a third had hacked the thing to display all nines. [1] https://www.displaymate.com/Spectra_4.html https://www.displaymate.com/Spectra_4.html

labcomputer 2y ago
For what it’s worth, the max polling rate for low-speed USB devices (of which mice almost invariably are) is, IIRC, 100 Hz. I would worry less about stair stepping and more about making the speed and acceleration distributions and auto-correlation look natural: a real human-powered mouse won’t be constantly moving (you have to stop the think about the next step, use the keyboard, etc) and when it is moving will tend to follow a rapid move-pause-move pattern. Another telltale would be a mouse that moves naturally but never clicks. Vendor and device ID are the lowest hanging fruit, but I think there are a lot of other giveaways that a mouse jiggler is being used.