McAfee Claims Successful Insulin Pump Attack

judgecorp writes "Intel security subsidiary McAfee has claimed a successful wireless attack on insulin pumps that diabetics rely on to control blood sugar. While previous attempts to attack insulin pumps have met with mixed success, McAfee's Barnaby Jack says he has persuaded an insulin pump to deliver 45 days worth of insulin in one go, without triggering the pump's vibrating alert safety feature. All security experts still say that surgical implants are a benefit overall."

Comments

Re:McAfee for insulin pumps next

[ColdWetDog]

I'm not one to believe this sort of conspiracy theory

Says 'the eric conspiracy'.

Sometimes it is best to post as AC.

Re:Glaring errors in the techweek article.

[Guy Harris]

An insulin pump is NOT implanted inside the user's body

Except when it is, although you might have to live in Europe to get it.

Also, pump's cartridges to hold insulin typically range from 200-300 units. Contrary to the article's claims, this is not 45 days worth!

In an implanted pump, it probably would be a larger supply.

The BBC article also states "Mr Jack said diabetics typically needed a dose of 5-10 units of insulin after a heavy meal to help regulate blood sugar. Making the device empty its cartridge into a host's bloodstream would cause "deep trouble"."

This is very flawed as well. Typically, insulin is taken before a meal whenever possible, and how "heavy" the meal is, is irrelevant. What matters is the user's insulin to carb ratio (how much insulin they need to properly use a gram of carbs) and how many carbs the item they eat contains.

I suspect by "heavy meal" he meant "carb-heavy meal". It might have been clearer had he said "carb-heavy meal", so nobody thought that chowing down, say, a 16-ounce filet would require a large bolus. And, yes, your mileage may vary depending on the insulin/carbs ratio. I'm not sure either of those are severely bad oversimplifications, though.

Also, when a person relies on an insulin pump, they're not just adding insulin to their body during mealtimes, the vast majority will be using it to deliver a "basal" dose of insulin, or a small amount of insulin 24/7 to stay alive (as this is a function normal non-diabetic bodies perform.) They also use it to deliver corrections, or small doses of insulin in response to blood glucose levels that are higher than expected after meals or throughout the day. A pump is not just a device you use after a "heavy meal."

Again, a simplification, but I'm not sure it's a severe oversimplification in an article written for a general audience; it doesn't invalidate the point of the article.

While it is true that an insulin cartridge unwillingly emptied into a patient poses significant danger, even without an alarm, I suspect 99% of people would be able to quickly notice such a large dose of insulin being delivered. You can see and feel insulin being delivered that rapidly. And if they happened to miss it, that's what frequent monitoring of blood glucose (which is required for all insulin pump users) is for. Sure, taking 200-300 units more than you should have would be a world of suck, but if you had access to food to eat or a sweet drink or glucose tablets, it's very likely an experienced diabetic would survive that sort of incident... to say nothing of if the cartridge wasn't full.

Well, for an implanted pump, it could be a lot more than 300 units; how fast it takes action is another matter, so maybe spending a while with your local store's entire supply of orange juice might be sufficient.

If more security were implemented in an insulin pump, there would certainly be no "frequent surgeries to replace the batteries," as the battery is (like the entire pump) stored in an external pump.

Again, not for an implanted pump.

Re:Wow

[dgatwood]

Try essentially impossible to investigate. How many people do you walk within twenty feet of in any given week? Any given year? Now imagine that any one of those people might have been the person who injected code that waits a predetermined period of time, does something bad, and then erases the location where the time delay is stored so that the original value cannot be recovered after the fact.... Or worse, overwrites the time delay with a value that implicates someone else.