Solution

How to combat the abuse of deepfake


AI vs. AI:The race to Generate, share&detect Deepfake

Most people cannot easily identify the authenticity of the deepfake videos, so we must rely on the technology, which is the algorithm as well.

The first method is using eyes reflection detector, which can detect whether the videos are synthesized or not. Computer scientists at the University of Buffalo created the deepfake-spotting algorithm that analyzes the reflections in the eyes of portraits to determine their authenticity. The researchers claim that the algorithm is 94 percent effective at identifying the false images or videos.The human being's cornea is very smooth, so there should be exactly the same reflected light in both eyes. However, the deepfake AI could not create consistent reflections in it.

(a. show the principle behind the detected algorithm.    b. the difference between real photo and synthesized photo created by deepfake AI.)


blockchain solution to our deepfake problem

what is blockchain?

Blockchain's ability to provide decentralized validation of authenticity makes it potentially effective as a tool to track and verify not just financial resources like Bitcoin , but all sorts of forms of content.


To understand the potential of blockchain in combatting deepfakes, understanding the basics of blochain’s features such as hashing algorithms, cryptographic signatures and blockchain timestamping (a secure way of tracking the creation and modification time of a document) is important.

A simple analogy for understanding blockchain technology is a Google Doc. When we create a document and share it with a group of people, the document is distributed instead of copied or transferred. This creates a decentralized distribution chain that gives everyone access to the document at the same time. No one is locked out awaiting changes from another party, while all modifications to the doc are being recorded in real-time, making changes completely transparent.

 how does blockchain work?

Data, hash, hash of previous block are the main characters in each block and it is the last one that makes blockchain so secure.

As you can see, each block has a hash and the hash of the previous block. So block number 3 point to block number 2 and number 2 point to number 1. Let's say we tamper with the second block. This cause the hash of the block to change as well. In turn that will make block 3 and all following blocks invalid because they no longer store a valid hash of the previous block, so changing a single block will make all following blocks invalid. 

But this is not enough, nowadays computer can compute thousands of codes in a second, so to mitigate this, blockchains have something called proof-of-work. It's a mechanism that slows down the creation of new blocks. So if you tamper one block, you'll need to recalculate the proof-of -work for all the following blocks.

how blockchain can help?

Think of a cryptographic hash as a unique fingerprint for the specific content or file. For example, the image below illustrates how information passing through a cryptographic hashing algorithm takes a text input of any size and creates an output of fixed length. This output is called a hash. Anyone else with similar data can use that algorithm to generate the same hash and comparing hashes can tell you that you share the same information. This can be useful in proving that a piece of text, file, or content has not been altered over time.

Add to this the immutability feature that blockchain brings - the ability for a record on blockchain to stay permanent and irreversible - and the hash or "fingerprint" cannot be modified and becomes a tamper-proof reference of the digital content at a specific point in time. If we apply this to deepfake identification, it means that on a blockchain network, hashes can be useful in proving that images, videos, or content in general have not been altered over time. Blockchain technology, by enabling provenance and traceability of digital content, hence can help to create an audit trail for digital content.

Together with conventional technologies such as digital signatures and standard encryption to ensure nonrepudiation, timestamping on the blockchain can serve to inform on items useful in identifying deepfakes, such as to confirm the date of the item's origin or to show that the content has been in someone's possession at a particular time. Such timestamping is already in use in specific applications today across a variety of industries, including for provenance in supply chain goods, notaries verifying signatures in contracts and others.

Although there are two feasible solutions we mentioned about, but it is still hard to eradicate entirely. However, With the right mix of education, policy, and technology, the leaders of today will be poised to build a future in which all of us can trust.


By ABChester
Webnode 提供技術支援
免費建立您的網站! 此網站是在 Webnode 上建立的。今天開始免費建立您的個人網站 立即開始