Learning Library

← Back to Library

Deepfake Audio Threats Explained

Key Points

  • Jeff demonstrates a voice deepfake created by an AI tool that can mimic his speech after only a short audio sample.
  • Modern deepfake technology can generate realistic audio and video from as little as three seconds of input, making convincing fakes increasingly easy to produce.
  • These fakes pose significant financial risks, enabling scams such as “grandparent” fraud and large corporate frauds that have resulted in multi‑million‑dollar losses.
  • Awareness and verification (e.g., confirming identities through independent channels) are essential defenses against deepfake‑enabled deception.

Full Transcript

# Deepfake Audio Threats Explained **Source:** [https://www.youtube.com/watch?v=cVvJgdm19Ak](https://www.youtube.com/watch?v=cVvJgdm19Ak) **Duration:** 00:14:33 ## Summary - Jeff demonstrates a voice deepfake created by an AI tool that can mimic his speech after only a short audio sample. - Modern deepfake technology can generate realistic audio and video from as little as three seconds of input, making convincing fakes increasingly easy to produce. - These fakes pose significant financial risks, enabling scams such as “grandparent” fraud and large corporate frauds that have resulted in multi‑million‑dollar losses. - Awareness and verification (e.g., confirming identities through independent channels) are essential defenses against deepfake‑enabled deception. ## Sections - [00:00:00](https://www.youtube.com/watch?v=cVvJgdm19Ak&t=0s) **Untitled Section** - - [00:03:04](https://www.youtube.com/watch?v=cVvJgdm19Ak&t=184s) **Deepfake Threats: Fraud & Disinformation** - The passage explains how convincing deepfake videos and audio can facilitate financial scams and spread false political messages, creating severe economic, electoral, and geopolitical consequences. - [00:06:06](https://www.youtube.com/watch?v=cVvJgdm19Ak&t=366s) **Deepfakes Undermine Courtroom Evidence** - The speaker warns that deepfake videos could cause wrongful convictions, current detection technologies are unreliable, and the legal system is unprepared to address these challenges. - [00:09:11](https://www.youtube.com/watch?v=cVvJgdm19Ak&t=551s) **Challenges of Universal Deepfake Verification** - The speaker argues that requiring every audio/video app to embed a deep‑fake verification label is technically daunting and ultimately ineffective because compliance is limited to good actors while bad actors will simply ignore the rules. - [00:12:21](https://www.youtube.com/watch?v=cVvJgdm19Ak&t=741s) **Out‑of‑Band Verification Strategies** - The speaker explains using alternative communication channels, third‑party confirmation, and pre‑shared secret code words to authenticate high‑risk interactions and guard against deepfake fraud. ## Full Transcript
0:00Hi, this is Jeff and you are listening to a deepfake of my voice. 0:04This is not a recording. 0:06I never actually said these words. 0:08This was all generated by an AI tool trained on audio samples of my voice. 0:12Many of you have the same technology in your pocket right now and don't even know it. 0:17In fact, it's included in a popular mobile phone operating system 0:21that you may use every day. 0:22In this video, we are going to talk about what deepfakes are, 0:25what risks they pose, and how we can defend against them. 0:29So that was deepfake Jeff. 0:30This is the real Jeff or is it really? 0:33Maybe this is a deepfake who just played a deepfake for you. 0:36I'll let you chase that recursion as long as you like and work on that on your own. 0:40Okay, let's see how these deepfakes actually work, what they are, how you build one? 0:46Well, you start off with an actual human being. 0:49So in the case of the deepfake that you heard me generate, 0:52I started off with me talking into my phone, 0:55speaking a set of of words that it told me I needed to say. 1:00It then listens to all of that and builds a model of my speech 1:05so it can do that after I've read all of that sample text. 1:10But there are some models that can do this with as little as three seconds of audio sample from an individual, 1:15so it's not all that hard to make very convincing deepfakes these days. 1:19Then, once you've built the model, what you do is you type in to the system whatever you want it to say. 1:26So if I type in, say this and I enter that text into the deepfake generator, 1:32then it will generate a sound that sounds just like that person or very similar to them. 1:37And we can do this with audio. We can do this with video. 1:41The video will actually show the mannerisms of the person and what they look like as well. 1:46These things can be very convincing and this technology is only getting better. 1:51Okay, let's take a look at some of the risks now that you know how a deepfake can be generated. 1:57How could someone use this to do bad things? 2:00Well, one type of risk 2:01classification of these would be a financial risk of some sort, a fraud. 2:06One of these things is also often referred to as a grandparent scam, 2:11because they are frequently the targets of these, 2:14although it really could happen to any family member or anyone that you know. 2:18In fact, the way these things work is 2:20you get a deepfake of someone's voice, let's say a grandchild, 2:24and then you have them call, the deepfake makes the call 2:27and talks to the grandparent and tells them, help, I'm in trouble. I wrecked my car, 2:33I got robbed, I've been arrested. Something like that. 2:37And I need you to send money. Please help. 2:39And what grandparent isn't going to help their grandchild? 2:42So they send money. 2:43But of course, they're not sending money to who they think, 2:46they're sending it to the bad guy. 2:48Another case, even very sophisticated organizations can fall for these kinds of scams and that's corporations. 2:55There was one organization that wired $35 million 3:01to a scammer based upon a deep fake phone call. 3:04Another organization did 25 million based upon a deep fake video call, 3:10where the person on the video was claiming to be the chief financial officer of the company, 3:16and it was convincing enough to make someone send the money, in following those instructions. 3:21So this can happen to a lot of folks, and, bad stuff happens when it does. 3:27What's another risk that can happen here? Well, how about disinformation? 3:31In the case of disinformation, this could have a lot of national and political side effects as well. 3:37There was one case recently in, the US presidential election, the lead up, 3:42where a robocall was calling people in a particular state, 3:47telling them they didn't need to go out and vote, 3:49that, in fact, they could just save their vote for the general election because this was just a primary. 3:54In fact, it was a robocall, and the robocall was in the voice of the president of the United States, which was a recognizable voice. 4:03And people thought they were hearing the voice of the president, and they weren't. 4:07Imagine if someone took that technology further still and used it to create some really damaging fake news of some sort. 4:14Maybe you have a head of state who appears to be on video 4:18declaring war on another country, 4:21or the head of a company, then saying, you know, the drugs that we manufacture, 4:27they kill half the people that take them. 4:29Even though it's not true, and even though the CEO never said that, 4:33it's going to cause the stock price to plummet. 4:35And if someone knows that when that's going to be released, 4:38they would know to buy shorts. 4:40That is a bet that a stock will go down and they can profit from that. 4:44So disinformation campaigns would be very damaging in a lot of cases. 4:48And then one other case, and there is a lot more than the ones that I'm just mentioning here. 4:53But one other possibility is an extortion attack where someone is trying to extort money from you. 4:59They say I've got compromising photos. 5:02I've got an audio of you saying something that you never said, or video of you doing something you never did. 5:09But it's not something that you would want anyone to know about, because it's a damage to your reputation. 5:15And this type of reputational, a threat could be enough to cause someone to pay real money just in order to keep this away, 5:24because it will be very difficult to detect whether it's real or not. 5:28So think about it this way. We have a threat because of the mere existence of deepfakes. 5:33These things create a lot of uncertainty. 5:36So we will have if we consider in the world of possibilities, we have false negatives and false positives. 5:43So a false positive is if we identified a deepfake and it really wasn't a deepfake. 5:49So imagine if, if you were a juror in a trial 5:54and someone shows you, the prosecution shows you a video 5:57of someone going into a bank with a gun and then walking out with money, 6:02and they show you that and they say, "will you convict?" 6:05You'll say, "yes, I just saw the video". 6:07But what if it wasn't a video? 6:08What if it was a deepfake? 6:10Now you might convict someone who was actually innocent on the other side. 6:15It could be that it wasn't actual video, but that the prosecutor showed you. 6:21But the defense just has to argue. No, we think that was a deepfake. 6:25So the mere presence of a deepfake causes doubt. 6:28And doubt is, of course, something that will hang a jury. 6:31So this is a technology that we're going to have to struggle with, 6:34and we're not really fully prepared yet, I think, to understand all of those implications. 6:41Okay, now, we've talked about what deepfakes are and how you can generate them. 6:44We've talked about what the risks are. What are the downsides that can happen here. 6:48Now what are you supposed to do about it? 6:50What kind of defense can we have? 6:52There are some things that I think work and some things that I think really don't work. 6:57And the number one in that category of things that don't work 7:00is the one that most people, especially technology oriented people, jump to. 7:04And that is, let's use technology that created this problem to solve the problem for us. 7:10Let's have some software that's able to detect 7:13the difference between a deepfake and the real thing. 7:17Well, that sounds like a good idea, but in practice it hasn't worked out so well. 7:22In fact, NPR did an investigative report 7:26where they looked at some of these deepfake detection tools. 7:29One of those tools actually did no better than 50/50, 7:3350% accuracy with its deepfake detection. 7:36Well, you know, I don't need to buy one of those tools. 7:39I got a deepfake detector that will give me 50/50 accuracy with this. 7:43All I have to do is that, and yeah, deepfake. 7:47So obviously that's not going to be really very accurate. 7:50That's not something we can really count on. 7:52The reason for this is that if I start looking at the technology itself, 7:58the deepfake detectors have continued to get better, 8:01and I suspect they will continue to get better. 8:04What's the problem? 8:05The deepfakes themselves are getting way better and much faster. 8:10So that means that we quickly reach an inflection point 8:13where the detection technology just isn't keeping up. 8:17And you can see, at least in the case of that, we're already there. 8:20So that I think, is going to be, a losing battle, 8:24because if you think about how good the deepfakes will get, 8:27at one point, they will be indistinguishable from an authentic video. 8:32So what's something else? 8:33This is the other area a lot of technology people look to, 8:37and that is some sort of authentication scheme 8:40where I'm going to be able to tell because at the time that the video is recorded, 8:45it will include some sort of label, some sort of marking. 8:49So maybe a digital watermark that you don't see, 8:51but that software that plays it will be able to look for the presence of, 8:56and then tell you, "this is a deepfake", "this is not a deepfake". 8:59Sounds like a good idea. 9:00But first of all, there's no standard for doing this, 9:03no industry standard, no common way that everyone agrees this is how we're going to do it. 9:07So we have to create that first and it doesn't exist. 9:11Secondly, I would need to have some sort of verification capability 9:17that would be part of the standard, 9:19and it would need to go into every single piece of software 9:22that ever does rendering of audio and video. 9:25That would be a lot of work. 9:26Think about all the apps on your phone, 9:28all of the different websites and things like that that might do audio or do video. 9:33All of those would have to be written to look for this deepfake label 9:38and be able to render that and tell you. 9:41But I'll tell you, even if we got all of that part solved 9:44and everyone who did a recording was able to label it as deepfake or not, 9:49which again, is a monumental "if" to begin with. 9:53The other issue is one of compliance. 9:57That is, think about we set up a system of rules. 10:01Who follows rules and who doesn't? 10:03Who's not going to label their deep fakes when they are? 10:06Who's not going to follow the rules? 10:08It's the bad guys. 10:09So in other words, we'll create a system that pretty much is followed by the good guys 10:14who are not really a threat to begin with, 10:16and the bad guys won't follow it. 10:18And every time we get a video that's not marked, you will not know. 10:22And that's where we are right now. 10:24So it really it's a lot of work to put us kind of almost back to where we start. 10:29Now, maybe there will be some other technological advances that I can't foresee, 10:32but that's my take on it. 10:34Now, what does work? 10:35Okay, I think number one, 10:37I was asked at a security conference one time, 10:40"if you were the head of the FTC", the Federal Trade Commission in the US that looks over fraud and things like that. 10:46"What would you do if you had only one thing you can do?" 10:49And I told them this, and I still believe it. 10:52It's education. 10:54I would want to run some sort of campaign 10:57to let people know what these deep fakes are. 11:00What is the art of the possible? 11:02What are the risks that go along with this? 11:05So that they wouldn't understand and be on the lookout for these things. 11:08Because I can tell you, most people have no idea how good this technology is already. 11:14Now, you heard the deepfake of my voice and it sounded a little deep fakey. 11:18It sounded a little depressed, a little, you know, lacking in emotion and things like that. 11:22So you might have detected that that was not really my voice. 11:26However, these technologies are a lot better than that. 11:29That's just one of the common ones that everyone has access to. 11:34And the one thing I'm sure of, again, the deepfake technology will keep improving. 11:39What's the other thing I'm trying to do? 11:41I'm trying to create a certain level of healthy skepticism. 11:44Now I'm a security guy so we can find the dark cloud in every silver lining. 11:49So we're skeptics by nature. 11:52You don't want to be overly negative and overly skeptical, 11:55but there's a certain level of healthy skepticism that is going to be necessary. 12:00We all need to be skeptics to one degree or another. 12:03If you weren't in the room when you heard it or saw it, 12:06maybe you didn't hear it and see it. 12:08Maybe what you heard and saw was a deepfake. 12:11If you're watching a show for entertainment, doesn't really matter. 12:14But if you're about to wire $25 million or even your life savings, then it matters. 12:21So in those cases where the stakes are high, 12:24then we should be relying on other mechanisms like out-of-band communications. 12:29Out-of-band means if I got a phone call and I hear your voice on it, 12:33then I'm going to hang up and then call you back 12:36at another number that I know you're supposed to answer at. 12:39Maybe even I call a family member or a friend of yours 12:43to verify that the story holds up. 12:45You know, are they really in that other country that they claim they're in? 12:48Because I didn't know they were supposed to be there. 12:50So that's one way. 12:52Also, using other means. 12:53If I got a voice call, maybe I send an email. 12:57Maybe I do it from a different device even. 13:00We can talk a little bit more about what some of these options are 13:03in another video that I'll point you to at the end. 13:05And another thing that that helps in this case, 13:09a lot of people have have really tried to get smart about this. 13:13And they'll say, well, in that grandparent scheme you mentioned, 13:16what if in advance we agreed on a code word? 13:20Is that going to work? 13:21So in other words, I tell all my family members, if I ever call you asking for money, 13:26ask me what the code word is, and if I don't know it, then it's probably a deepfake. 13:31And because the deepfake generator can generate my voice, 13:35but it doesn't know all the things that I know. 13:37So this sort of secret knowledge pre-shared in advance 13:41would be the way to tell if I trust it or not. 13:44I put a question mark on this for a reason though. 13:47There is a special type of attack, 13:48I'll make reference to where you can find out more about it, that defeats even that. 13:53Deepfakes represent an escalation in the cyber arms race. 13:57The bad guys keep getting more and better tools. 14:00That means we're going to have to keep getting smarter in order to defend against it. 14:04So that was the purpose of this video to make you aware of what deepfakes are, 14:09what some of the risks are, 14:10and warn you in terms of what kinds of capabilities you might use 14:14in order to defeat these things and detect them when they happen. 14:18Take a look also at the video I did on audio jacking, 14:21so that you'll understand why code words may not be the panacea that you had hoped they'd be. 14:27They say that for warned is forearmed. 14:30Now you should consider yourself to be both.