WEBVTT 00:00.000 --> 00:04.740 Your password probably sucks, but that's not the real problem here. 00:05.180 --> 00:13.280 The real problem is the companies that store it like amateurs, making even stronger passwords 00:13.280 --> 00:15.720 crack in seconds sometimes. 00:16.440 --> 00:25.660 Key stretching changes this equation by making every wrong guess cost actual money in electricity 00:25.660 --> 00:27.660 and hardware and time. 00:28.460 --> 00:33.140 Picture your password as a cheap lock on a storage unit, for example. 00:33.680 --> 00:37.000 Anyone with bolt cutters can walk right through it, essentially. 00:37.580 --> 00:44.780 Key stretching wraps that lock in layers of computational concrete, like suddenly those 00:44.780 --> 00:50.920 bolt cutters need to become like industrial mining equipment that burns thousands of 00:50.920 --> 00:56.000 dollars in electricity just to try to get it open. 00:56.420 --> 01:01.600 Most people think password strength comes from randomness and length, and that's part 01:01.600 --> 01:03.280 of it, but there's more to it. 01:03.420 --> 01:07.260 What actually matters is how expensive your password becomes to guess. 01:07.980 --> 01:15.320 For example, a character password properly stretched with modern algorithms resists attacks better 01:15.320 --> 01:20.360 than a 16 character password stored with basic protection. 01:21.400 --> 01:24.820 Here's what nobody actually explains about password storage. 01:25.260 --> 01:30.620 When you create an account, the website doesn't actually save your password, right? 01:31.000 --> 01:38.640 They run it through a mathematical function that creates a unique fingerprint called a hash. 01:39.220 --> 01:46.200 Attackers who steal these ashes need to try to guess the passwords and check if they 01:46.200 --> 01:49.820 can create the same type of fingerprint that was created. 01:49.820 --> 01:56.980 Now, without key stretching, modern graphics cards check billions of guesses per second, so 01:57.290 --> 02:03.480 your super secure one, two, three password falls in basically microseconds. 02:03.900 --> 02:11.080 With proper stretching, using something like Argon2, each guess takes half a second and 02:11.080 --> 02:13.060 eats a gigabyte of RAM. 02:13.540 --> 02:19.140 Now your attacker went from billions of attempts per second down to maybe one or two. 02:19.140 --> 02:25.400 The math becomes kind of hilarious when you scale it up, so testing one million password 02:25.400 --> 02:32.160 candidates at two guesses per second requires almost six days of continuous computation, 02:32.520 --> 02:34.060 which is a lot more effort, right? 02:34.580 --> 02:41.200 Obviously, testing one billion, though, requires nearly 16 years, and that assumes perfect 02:41.200 --> 02:48.060 efficiency with zero downtime, like no hardware failures, no electrical cost piling up. 02:48.060 --> 02:51.820 Think about what that actually means for mass surveillance, for example. 02:52.420 --> 02:58.480 The NSA can't just vacuum up everyone's data and crack it later when key stretching is 02:58.480 --> 02:59.620 actually done right. 02:59.780 --> 03:06.260 They need to try to decide if cracking your encrypted file is worth burning 50 grand 03:06.700 --> 03:12.880 in computational resources, and most people really aren't worth that level of investment. 03:13.300 --> 03:18.980 Early UNIX engineers basically understood this principle back when computers filled 03:18.980 --> 03:22.120 entire rooms because they had to. 03:22.740 --> 03:25.460 You have to be more efficient when you have less. 03:26.000 --> 03:31.360 So they used multiple rounds of encryption, DES specifically, not to hide passwords better, 03:31.460 --> 03:36.460 but to make each guess hurt, and modern algorithms took this concept and basically 03:36.460 --> 03:40.900 weaponized it into economic warfare against attackers. 03:40.900 --> 03:47.820 PBKDF2 runs your password through thousands or even millions of hash iterations, 03:48.260 --> 03:55.200 and each feeding into the next, kind of like a computational chain that can't be shortcut. 03:56.020 --> 04:02.640 Bcrypt uses something called Blowfish Cipher to create deliberate complexity that really 04:02.640 --> 04:06.780 neutralizes GPU advantages to a large degree. 04:06.780 --> 04:11.480 Those expensive graphics cards that hackers love suddenly become inefficient, 04:11.900 --> 04:14.740 paperweights basically, or space heaters. 04:15.060 --> 04:22.040 Xcrypt goes nuclear on memory requirements by demanding huge amounts of RAM for each guess, 04:22.240 --> 04:27.280 and it makes specialized cracking hardware just economically stupid. 04:27.700 --> 04:34.360 Your attacker needs actual computers with real memory, not just raw processing power. 04:34.360 --> 04:41.540 Argon2 combines all of these approaches with adjustable parameters that let defenders adapt 04:41.540 --> 04:44.820 to evolving threats when they evolve. 04:45.120 --> 04:47.080 Have you heard of Keystretching before this video? 04:47.620 --> 04:49.480 Like most people haven't, that's why I'm asking. 04:50.240 --> 04:53.360 Because companies really, on one degree, don't really want you to understand 04:53.360 --> 04:55.460 how badly they're failing you. 04:55.760 --> 04:59.520 I'd definitely like to hear in the comments if this is new information to you. 04:59.520 --> 05:04.040 I'm genuinely curious how well this knowledge is actually spread 05:04.040 --> 05:07.260 outside of the kind of technical circles that exist. 05:07.940 --> 05:12.280 And here's where companies, some degree, constantly screw you over. 05:12.400 --> 05:18.360 Like your bank probably protects your password worse than a properly configured WordPress blog 05:18.360 --> 05:19.080 does. 05:19.520 --> 05:25.240 They chose speed, like over security, because they'd rather get hacked than 05:25.240 --> 05:27.140 weigh an extra half a second. 05:27.140 --> 05:28.980 And that's just being competent. 05:29.300 --> 05:36.900 The implementation failures definitely go even further than just being cheap, though. 05:37.000 --> 05:39.200 We can't just put it all on that. 05:39.580 --> 05:42.020 Keystretching needs something called the salt. 05:42.260 --> 05:44.860 And basically it's a random value. 05:45.100 --> 05:50.840 You need to each user that gets mixed with your password before it gets stretched. 05:51.280 --> 05:56.360 And without proper salts, attackers can basically pre-compute answers 05:56.360 --> 05:58.020 or optimize their approach. 05:58.760 --> 06:03.760 Yet company after company reuses salts or doesn't use them at all. 06:03.940 --> 06:10.660 Because their developers learn security from a YouTube tutorial that was not this channel. 06:11.300 --> 06:16.820 Password managers like KeyPass and Bitwarden do actually implement this correctly 06:17.240 --> 06:19.400 for things like their master password. 06:19.640 --> 06:25.400 BearCrypt is great in that it lets you configure millions of iterations for disk 06:25.400 --> 06:25.820 encryption. 06:25.820 --> 06:32.600 Even some consumer routers actually support stretched passwords if you dig through their 06:32.600 --> 06:32.980 settings. 06:33.540 --> 06:34.480 The tools exist. 06:34.680 --> 06:38.100 Companies just really refuse to actually use them properly. 06:38.420 --> 06:42.520 And here's the kind of dark comedy of our surveillance state. 06:42.640 --> 06:46.760 The FBI complains about going dark when criminals use encryption. 06:47.060 --> 06:54.200 But what that actually means is that properly stretched keys cost too much to actually 06:54.200 --> 06:54.520 crack. 06:54.520 --> 06:58.220 Actual security interferes with their dragnet surveillance program. 06:58.420 --> 07:02.980 And every additional iteration of key stretching represents friction against 07:03.480 --> 07:07.220 the unconstitutional spying that's constantly going on. 07:07.360 --> 07:13.440 The same government that's spending billions on quantum computers to break encryption can 07:13.440 --> 07:18.800 be defeated by properly configured password stretching that any competent developer could 07:18.800 --> 07:20.500 implement in an afternoon. 07:20.500 --> 07:25.940 They're building death stars while we're winning with computational guerrilla warfare. 07:26.200 --> 07:31.420 Your operational security depends on understanding these kind of basic dynamics that exist. 07:31.760 --> 07:37.660 Strong passwords alone will not protect you when they're stored completely incorrectly. 07:38.180 --> 07:43.800 But even mediocre passwords become nearly unbreakable when properly stretched. 07:44.400 --> 07:47.840 The transformation happens through pure mathematics. 07:48.260 --> 07:51.640 And that means like there's no back doors or weaknesses. 07:52.060 --> 07:59.900 It's just computational expertise that scales with attack or resources. 08:00.740 --> 08:05.840 Like individual users can take control immediately when choosing services. 08:06.100 --> 08:08.080 Ask about their key stretching parameters. 08:08.580 --> 08:12.060 Use tools that let you control these settings. 08:12.600 --> 08:15.920 Make your secrets too expensive for anyone to try and steal. 08:16.510 --> 08:19.540 Stop trusting companies that have proven repeatedly 08:20.110 --> 08:23.560 that they choose profit over your protection. 08:24.100 --> 08:29.000 The economics become pretty brutal for attackers when we get to scale, right? 08:29.460 --> 08:32.620 Like cryptocurrency miners, for example, perfectly understand this. 08:33.000 --> 08:36.180 They calculate attrition costs versus potential profits. 08:36.600 --> 08:40.460 Password cracking follows identical economics. 08:40.860 --> 08:44.820 Make the cost exceed the value and the attack stops. 08:44.820 --> 08:48.380 And not because attackers can't actually succeed, so what I'm saying. 08:48.820 --> 08:51.740 But because they just can't afford to succeed. 08:52.260 --> 08:57.040 Like surveillance capitalism depends on weak security to remain profitable. 08:57.320 --> 09:00.900 Data breaches need to stay cheap if every password required 09:00.900 --> 09:04.640 serious computational investments to actually crack 09:04.640 --> 09:08.900 the entire ecosystem of buying and selling stolen credentials 09:08.900 --> 09:11.080 to some degree would collapse. 09:11.080 --> 09:16.080 And that's exactly why proper implementation faces such strong resistance. 09:16.640 --> 09:22.680 Every properly protected password increases the global cost of surveillance. 09:22.980 --> 09:26.180 Every company implementing real key stretching 09:26.470 --> 09:30.920 makes mass data breaches less and less valuable, right? 09:31.340 --> 09:34.760 Like the aggregate effect could absolutely cripple 09:34.760 --> 09:38.280 the surveillance economy if deployed widely enough. 09:38.280 --> 09:42.520 And this isn't really, this has nothing to do with personal privacy, right? 09:43.060 --> 09:48.040 I'm talking more about systemic resistance to authoritarian overreach. 09:48.400 --> 09:54.480 For example, key stretching transforms authentication from a speed bump 09:54.480 --> 09:59.180 into kind of an economic battlefield where defenders set the terms 09:59.180 --> 10:02.800 and you get to decide the computational requirements that will be needed. 10:03.200 --> 10:05.060 You control the memory demands. 10:05.060 --> 10:10.940 You force attackers to fight where their advantage is essentially evaporate 10:10.940 --> 10:13.880 and their budgets screaming pain. 10:14.580 --> 10:17.200 The solution is not that complex. 10:17.480 --> 10:23.340 New systems should use Argon to with at least 64 megabytes of memory 10:23.340 --> 10:28.000 with multiple iterations legacy systems need S-crypt or B-crypt 10:28.000 --> 10:33.380 with aggressive parameters that never accept PBKDF2 with shot one 10:33.380 --> 10:37.580 despite what outdated compliance frameworks actually suggest. 10:38.260 --> 10:40.700 These definitely are not recommendations. 10:40.840 --> 10:47.320 They're kind of minimal viable security in this day and age, right? 10:47.460 --> 10:52.060 With like $8 per hour criminals renting massive GPU clusters. 10:52.780 --> 10:57.480 What password managers or encryption tools are you using 10:57.480 --> 10:59.780 that actually implement proper key stretching? 10:59.780 --> 11:05.440 Security theater to most companies they don't bother actually doing this 11:05.440 --> 11:10.320 which is why I saw making this video as a worthwhile project 11:10.320 --> 11:12.540 for us to sit down and discuss key stretching. 11:12.800 --> 11:15.900 If you haven't heard of this, I'd love to know about it in the comments 11:16.320 --> 11:18.940 and if you did hear about it, I'd like to know what's about it 11:18.940 --> 11:22.240 and if you did know about it, I'm sorry that you already knew about it 11:22.240 --> 11:22.940 and saw nothing new. 11:23.080 --> 11:24.160 I get comments all the time. 11:24.280 --> 11:25.520 I was like, I already knew about this. 11:25.660 --> 11:27.680 Why didn't you talk about something more detailed 11:27.680 --> 11:33.220 and in depth, sorry, I get a poll of what everyone knows 11:33.220 --> 11:35.360 when I make these videos and I guess I digress. 11:36.260 --> 11:38.920 Thank you for watching to the end and I'll see you in the next video.