Crime novel trope: special passphrase deletes all data

Factual Questions
21

Modifying encryption algorithms generally makes them much, much weaker, especially if someone thinks they’re being clever with ‘subtle’ changes. It’s way too easy to introduce flaws if you don’t have peer reviews even if you’re an expert on crypto programming, and if you are an expert at crypto programming that probably came at the expense of extensive spy training, or can be used to make a ton of money at a legitimate or semi-legitimate job. Your efforts are pretty much always going to be way more productive if you work on having a solid plan for using encryption and managing devices and access that doesn’t rely on giant security holes like passwords written on paper (which was the case in the OP’s story).

This highlights one of the problems with security: you have to balance usability with security. “You must run this obscure item within 60 seconds” means you could lose all of your data if your system has issues on startup - there are windows updates that take longer than 60 seconds to give you back control, for example. If you have a program that you have to run every 20 minutes, it might kick in on a police raid - but also will kick in if your chinese food delivery arrives at the wrong time, or you get sidetracked while trying to work with your data. If the data is incriminating but has so little value to you that you’re not worried about losing it, you’re probably better off just not saving the data in the first place.

Side note: Any competent law enforcement will attach the disk to a duplicator and make a copy before anything else whether the disk is encrypted or not unless they’re trying to access the device without shutting it down. They can’t understand the data since the copy will just be encrypted data like the original, but they can mess with it without worrying about losing the original.

You could possibly thwart this with custom drive hardware that a disk duplicator won’t understand or would get wrong, but as far as I know even major power spies don’t bother with that, as it gets expensive and annoying (since spies in the field are generally going to treat disks as disposable) and isn’t really worth the effort.

22

As someone has already mentioned Ransomware will run in the background encrypting your files so you don’t even know it’s happening till its too late.

There could be a trigger that activates the Ransomware Virus in the background as soon as the someone tries to access the drive unless a code is activated first.

Another point is that its that it’s not possible to “Wipe” a disk, your information will always be on there.

“Wipe” programs do not “Wipe” a disk they overwrite your data, even then with the right equipment and enough determination the data your data may still be recoverable even if overwritten several times.

The only way to guarantee your data is gone for good is to physically destroy the disk.

Perhaps a phial of metal etching acid triggered to explode over the disk would do the trick :shushing_face:

23

How good are you at making a reliable trigger that will figure out that the cops are coming and won’t accidentally trigger from you dropping a notebook on the case of the PC? And how good are you at making something that will get enough acid to destroy the actual disk (not just damage the case) but also won’t accidentally hurt someone poking into the case? Because if you get the mechanics of the booby trap wrong, and end up splashing acid in a cop’s face or blowing off his hand, you’re looking at some slam-dunk violent felony charges, getting picked last in ‘who wants a deal for a lighter sentence’, and probably going to serve a sentence in a prison for violent criminals instead of white-collar criminals. An awful lot of data that you might want to hide from the police isn’t going to be a slam dunk, and isn’t going to get as harsh of a punishment as maiming or killing an LEO.

24

The scenario in the book seems like a figment of the author’s imagination. Never heard of such a capability. Typically sensitive data is protected by whole-disk encryption, at least corporate data (I don’t know what military and spies do). If I wanted to analysis of a hard drive and didn’t know anything about the state of the computer, I would not even turn on the computer. I would mount the drive on another machine that has all my forensic analysis bells and whistles.

25

Right. The novel is a writers imagination. But the trigger scenario is valid. With my limited knowledge I could make an account that, if logged in, would run a startup program, script or batch file that starts erasing data files. Forget about wiping the entire disk. I could set the startup material to just delete known sensitive data.

My point is that you cannot do it (at least under Windows 7) with my normal login account. It has to be some other account.

26

Not so. As @si_blakely said way back at post #3, there are customization points even in the latest Windows where a competent developer could insert custom activity into a login attempt with the right account and the trigger-as-password.

Could Joe Average computer user program that from scratch? No. Could any spy agency or Fortune 1000 company have its dev team whip one up? That’d be trivial. Could such a tweak already be available for easy free download and install from the quasi-criminal part of the web right now? It certainly could. Is it? I haven’t searched & don’t intend to.

As has been said several times above, the way you “wipe” data in the modern era isn’t to erase the sensitive bulk data. It’s to erase the tiny smidgen of key data that is necessary to decrypt your sensitive bulk data that you’ve stored encrypted from the git-go. In effect that wipes all the meaning from all the sensitive data without wiping any of the bits of the same sensitive data. And does so substantially instantly.

I recall reading how Edward Snowden of NSA spying whistleblower fame protected the data he was stealing from NSA. In his explanation he repeatedly encrypted the data, like 20 times, using a different key at each level and with each key stored in a different way on a different medium. Some on disk, some on USB sticks, a couple in his head. Correct decryption required having all the keys and knowing which order to apply them in. If any one key got lost or was destroyed, the data was irretrievably gone. He was pretty confident even the NSA wasn’t getting into his data no matter how hard they tried. And he could “wipe” it real easily by losing or destroying any of the keys.

27

Has there ever been any evidence of this being actually done? If it were really possible to read a disk that had been overwritten n times, then you could use that same technology to make a hard drive with n times the capacity.

28

As I expect @Chronos remembers from prior threads on topic, IMO @chickenwrangler is repeating a well-known bit of PC / IT lore that upon further investigation has no actual basis in fact. It’s just been repeated by many articles and many people, including myself before I knew better, over and over for 30+ years now.

It’s widely believed, but simply false. Though @chickenwrangler gives himself an out with “… may be recoverable …”. “May” includes as a degenerate case “Can’t really, but also can’t be disproven since negatives can’t ever be proven.”

Although this particular UL got started long before wiki was a thing, see also:

29

Here’s the last thread on point with some excellent refs to real experts debunking of this UL:

30

I have not read that book; is that related to the—extremely relevant to this thread—technique where, upon entering an initial password, one password boots into the real system while another password boots into a plausible decoy system? Repeat with multiple layers if desired.

Merely splitting a secret key into 20 essential parts is straightforward enough.

31

A master programmer can pretty much get a computer to do anything he wants. Many years ago, there was a famous computer virus called, “Michaelangelo”. The virus was designed to infect DOS systems, but did not engage the operating system or make any OS calls. Michelangelo, like all boot sector viruses, operated at the BIOS level. Each year, the virus remained dormant until March 6, the birthday of Michelangelo.

32

For sure, if your hardware has been tampered with/rooted/backdoored/exploited then it’s probably too late to save yourself by wiping things out.

33

Not that I know of. His purpose wasn’t to fool anyone who had captured his hardware. His purpose was simply to ensure that capturing his data (with or without the underlying hardware) wouldn’t reveal what a treasure trove it was.

Yes and no.

Brute-force-ability is related to the length of the key. If a long-enough key is split into a bunch of smaller hunks, some of which are revealed, the brute-force-ability of what’s left gets exponentially much easier. By simply using 20 full length keys the brute-force-ablility of any one of them is uncompromised.

There wasn’t any sense that twice encrypted was “more or better encrypted” than once. It was just a matter of throwing up a large number of discrete obstacles, each of large difficulty. Heck, even armed with all 20 keys but no knowledge of their correct sequence you’d have to try all 2^20 ~=1 million sequences of keys.

In any case, I’m not defending his technique as ideal or even necessary; I’m just relating his tale for whatever that’s worth.

His book is a good read; I recommend it. He’s of course convinced of the righteousness of his cause, but he’s far from alone in that.

34

I don’t know of any evidence of recovering data after a deliberate wipe by overwriting the disk three or seven times (what ‘wipe’ programs typically do). It is definitely possible to recover after a smaller amount of overwrites:

Once I was working support at a company that created an office automation package for offices, including small offices. They were supposed to have their own IT, and the group I was in was the group that handled the interaction between our software and the OS/hardware (and often did billable work fixing stuff that their IT should fix). One office had hired a local IT guy to handle backups among other things, and one day they had a disk failure in their RAID. He replaced the wrong disk and when the RAID complained, told it to reinitialize which overwrote at least some data and turned it into a mess. When the office got some new drives, installed a fresh OS, and came to us to restore from backup, it turned out their IT guy had never run a single backup, and they’d be in serious regulatory trouble if they didn’t have their mandatory records.

They ended up sending the drives to a data recovery place, who were able to get the entire raid working again with the original data, for around $20,000 (in the early 2000s). There hadn’t been a deliberate ‘overwrite seven times’ kind of delete, but some of the data had definitely been overwritten and all of the metadata was overwritten, but they were able to make a full recovery. I never heard if they sued the IT guy, but they certainly should have unless we weren’t getting the whole story.

The fact that something possible to do and should be planned for in a worst case scenario doesn’t mean that it’s accurate enough for day-to-day data storage - if there’s only a 50-50 shot of recovering the data after overwriting, you don’t want to flip a coin to see if your data is still available every day, but you don’t want to rely on that wipe when it’s a 50-50 chance of going to prison if the wipe doesn’t work. Similarly, the fact that something is possible doesn’t mean that it’s economical - data recovery places often disassemble the disk and use more sensitive instruments to read the extra data at the ‘edge’ of the disk, but if the more sensitive instruments to get five times the capacity are much more expensive and bulkier than a stack of ten regular disks and a raid controller for them, no one is going to want to spend the money and space on superdisks when they could get more storage and reliability for less money using mass produced disks.

35

Not what I meant. Let K be the important secret key. Choose 19 random numbers K1, …, Kn of the same length as K, and just pick K20 = K ⊕ K1 ⊕ K2 ⊕ … ⊕ K19. Now destroy K and distribute the pieces, any of which could naturally be further encrypted or password-protected or whatever.

How is brute-force-ability compromised?

36

I misunderstood your point. I agree with your interpretation / explanation.

37

The whole point of RAID is to be able to recover data after some of it is irretrievably overwritten, so it is not stretching credibility that for $20000 + all the drives + disassembly/instrumented low-level imaging of the failed drive you could get your data back.

38

I should have been clearer on part of my point - I think that recovering data that’s been actually wiped (like running a program that deliberately overwrites multiple times) is impossible, and I’ve also never seen any evidence that someone can actually recover data from that. Incidental overwrites are different, and data is often not actually overwritten, and even if it is in some cases can be reconstructed or recovered with advanced analysis. I was disagreeing with Chronos’s argument that if you could do that you could use the technique to make larger drives, but just that specific argument.

Also in my example I’m pretty sure that the IT guy had already gotten rid of the failed drive before they tried reconstructing the RAID, but it was around 15 years ago so I could be wrong. That guy was really incredibly incompetent.

39

As I understood it - based on technical discussions from 10 or more years ago when disks were much smaller, so the recording tracks much bigger… A disk is a series of circular recording tracks; the “write” head is wider than the “read” head to ensure a strong enough signal for the read. Plus, due to the vagaries of mechanics, the write head might “wander” a fraction from the true path. Think of the write as the road including gravel shoulders and the read as the paved portion.

First there’s the allegation that recording over something leaves a “ghost” residue of the previously recorded data (any reel-to-reel tape enthusiasts here?). So re-write multiple times, varying 1’s and 0’s, ghost recording isn’t that robust. Second, there’s the thought that some of the data might be picked up by analyzing the edges of the recording, looking for the “shoulders” where the overwrite missed the previous write. Hence, again, multiple overwrites.

The problem is - ghost analysis will fail if the track is overwritten multiple times. Examining the edges may work, but this technique may get you fragments of the data - not a complete record. If it’s fragments of clear text then great - you may recover enough to say “look your honor, this is part of a bomb-making manual and this is a partial list of diplomatic personnel”. You likely wouldn’t get enough to have a truly decryptable file or disk.

I will agree with you, it has been demonstrated in real life that clever people who try to make their own encryption or modify algorithms will not create decently robust algorithms; it’s a very specialized field. I would think though, there are some public domain encryption programs, with source. You could modify these so after applying the robust encryption, it then applies an additional encryption. It might be an elementary trick, but if the NSA/CIA assumes you are using the off-the-shelf version it may confuse them, if you erase your altered copy (or keep it separate on a USB).

In all such cases, the question is - how easy is it to decrypt, and how hard do they want to read your data? If you think the data cannot be decrypted, why destroy it?

Additional note - there’s a new high density data drive type called “Shingled”, IIRC. It writes in groups of tracks so close together that the “write” tracks overlap, so writing any data in that track group means re-writing the entire group - write track 1, slightly overlap 1 writing track 2, slightly overlap 2 writing 3, etc. The advantage is higher data density, no blank areas between tracks; the downside is much slower random writes - so ideal for archives and backups, less so for busy servers. And… makes above mentioned “ghost” data recovery techniques less likely to work.

Another technique (which I assume would work) is using an SSD drive and rigging it to experience a pulse of high voltage to effectively destroy the storage. If necessary, rig up a battery-powered device to do this to prevent “just pull the power cord”. You could even rig this to be software controlled also.

(I’m imagining testing an SSD killer device could get expensive)

There’s no end to techniques. Just use imagination.

And finally - using obscurity doesn’t work. “I’ll just hide the file in this folder and rename it” yields to basic search programs and others that detect anomalous files - “error - file is not a jpg”

40

A sort-of way to do this under a normal account would be to have the normal account start the process on login, but with enough of a delay that the user could stop it by running some other program and entering a special password or something.

An attacker who was careful could probably still figure out how to stop this, but if it were a bespoke security system it seems plausible that he’d be unsuccessful.

But, yeah, that’s why you clone the drive first.