3 ms·
Watermarked ReRAM: A Technique to Prevent Counterfeit Memory Chips
- _3u10 5y agoWouldn’t counterfeit chips not work? Like if the RAM works at the advertised speed I don’t really care which fab made it.
- bb88 5y agoIt's not that counterfeit's won't work, just how well? And should Micron have to deal with warranty replacement for memory they didn't sell in the first place?
- yodon 5y agoThis is also a supply chain security issue, making sure the counterfeit chip doesn't have surveillance or other offensive capabilities added.
- _3u10 5y agoGenuine Cisco routers were backdoored for surveillance. I’m sure lots of genuine products have backdoors in them.
- joezydeco 5y agoUnless you're a shop that doesn't test 100% of your incoming parts to 100% completion. How would you know it's going to work at the advertised speed?
- Brian_K_White 5y agoTesting for "works" is non-trivial.
- _3u10 5y agoI guess if your skills are low enough that it could be difficult to test memory chips but you should test them anyway even if they are “genuine”. Personally I’ve never had much of an issue testing ram… lots of tools to do it.
- seabird 5y agoJust because a part passes memtest86 and a couple of hours of Prime95 does not mean that that part is interchangeable with the genuine article, especially as far as an OEM who is trying to meet requirements is concerned. With a counterfeit, you have no idea what the MBTF is, no idea what the specific electrical properties are, no idea what the errata are, etc. etc. Even moving from an obsolete/last-time-buy part to the manufacturer recommended replacement might not go entirely smooth.
- Brian_K_White 5y agoI think you're in a poor position from which to talk about low skills if you don't understand why vetting components is non-trivial.
- rowanG077 5y agoIf you use any ramtester software you test one combination under one environment. Maybe the counterfeit chips don't allow as long traces. Their setup or hold time are slightly different which might be an intermittent error. Or only an error when in an 50 degree celsius+ environment. It's trivial to test RAM under a single condition like you are proposing. It's almost impossible to check whether RAM performs as expected over all scenarios.
- belter 5y ago"...Electronic counterfeiting is a longstanding problem with adverse long-term effects for many sectors, remaining on the rise. This article presents a novel low-cost technique to embed watermarking in devices with resistive-RAM (ReRAM) by manipulating its analog physical characteristics through switching (set/reset) operation to prevent counterfeiting. We develop a system-level framework to control memory cells’ physical properties for imprinting irreversible watermarks into commercial ReRAMs that will be retrieved by sensing the changes in cells’ physical properties..." https://arxiv.org/pdf/2204.02104.pdf https://arxiv.org/pdf/2204.02104.pdf
- teraflop 5y agoI read the paper, and I really have a hard time understanding what kind of threat model the authors are proposing. If the details of the watermark are public, as they suggest, what would stop a counterfeit manufacturer from just writing their own watermark to their own devices? Wouldn't that be just as easy as printing a fake logo on the IC package? I suppose you could prevent this with some kind of scheme using digital signatures, but the paper never suggests or considers that. And if you had a way to generate a "certificate of authenticity" for each device, what would be the point of storing it as a wear pattern, instead of in an EEPROM or something? The authors only say that their proposal "does not require any hardware modification", which seems like a pretty flimsy justification.
- justsomehnguy 5y ago> scheme using digital signatures The simpliest one would be to have "CHIP S/N: XXXX MFG DATE: YY/MM/DD" signed by some CA and watermarked on each chip. That doesn't prevent for a counterfeit manufacturer from grabbing a handful of original chips and watermarking a series of their counterfeit chips with it. So it boils down to: a) no watermark on a chip what should be watermarked: obvious counterfeit b) one watermark on all chips: obvious counterfeit c) a valid watermarks on a shipment of chips: if you order a ton of them AND you bothered to setup both the tests and means to store and compare the results, so it would be obvious when you receive the shipment of chips with watermarks what you already has tested. for anyone else it would be c2) chips with a valid watermarks: not a counterfeit
- josephcsible 5y agotl;dr: ReRAM wears out and gets slower every time you write to it. The watermark is placed by intentionally repeatedly rewriting certain parts of it until the slowness from the wear becomes noticeable.
- mjevans 5y agoThus about as useful as selling an LCD already unevenly burned in with some ad or station logo already on it?
- arcticbull 5y agoI strongly suspect this will apply to a special region used exclusively for watermarking. Their imprinting rate is 0.6 bits per minute.
- Brian_K_White 5y agoSpotted that even though they tried so hard to hide it by saying "watermark" all over the title and article? It's invisible to the consumer, not like a pre-burned-in crt. It will apply to a dedicated set of cells, or it will be no different from existing natural wear which is just a fundamental property of the medium no matter what. It's no different than flash drives reserving some of the cells for wear replacement. You're buying a 5 gallon bucket, and in reality the design of the new better handle means it holds 5 gallons minus a drop. It's not good math to care about that. You would have to show that it robs 5% of the capacity or performance without also costing 5% less for it to make any sense to even think about it.