Whiteboard Tutorial: Image Technology
Bad actors are adept at avoiding traditional detection techniques used by Brand Protection teams. You need to deploy image capabilities to discover the hidden listings and posts that are undermining your brand and putting consumers at risk.
In this Whiteboard Tutorial, we:
- Detail why keyword searching, despite still being essential, is not sufficient alone in tackling brand misuse
- Define the image capabilities that you will need to discover hidden infringements: Logo Detection, Similarity Matching, and Optical Character Recognition (OCR)
- Explore four distinct examples of image capabilities being used to discover infringing products
Audio recording
An edited version of the Whiteboard transcript can be found below.
Text and keyword searching are necessary but not sufficient alone
- A Brand Protection program will be using text and keyword searches to build a base profile of the threats out there, but this is no longer enough as bad actors have evolved their behaviors.
- Images need to be used as a data source in their own right, to aid the discovery and prioritization of infringements.
- Typically, we now see a lot bad actors not using text in their listings and posts but instead images – this is especially the case for advanced bad actors. For example, in a recent account over half of the high-risk items identified came from the imaging capabilities.
The Power of Three
In addition to image matching/searching, a comprehensive image-based program will include three main areas: Logo Detection, Similarity Matching, and OCR.
- Image matching/searching look for exact copies of, or closely similar images to those that belong to the brand.
- This capability ensures that infringements that do not contain brand related keywords are still collected and prioritized by Talisman, Incopro’s Brand Protection platform – for instance, use of copyright images.
Logo Detection
- Logo detection goes one step further, and uses machine learning based image classification, to identify the use of logos within an image.
- Can find logo in an image even if the logo is purposely distorted or adjusted to avoid detection.
Similarity Matching
- Our similar matching capability identifies lookalike products based on shape, texture, and color, regardless of how the item is described in text, – and even if the trademark is not mentioned.
- This capability can be used hand in hand with design rights.
Optical Character Recognition (OCR)
- As noted, text searches are always used in Brand Protection programs, but we also supplement them with our Optical Character Recognition (OCR), to discover and collect text that has been embedded in images to evade detection.
- If the text is “within” the image – it will not be captured using standard keyword searching – we use OCR technology to pull out and process the text within the image and integrate with our prioritization tools.
Logo Detection Example
This technology uses machine learning to identify the use of the brand’s logo within an image, even where the seller only uses generic words to describe the item that is being sold.
This face mask listing is an example, where the product is generically described as ‘designer inspired face masks’, but Logo Detection is able to identify the use of the brand’s logo on the product itself.
The red flag: the seller is being evasive in not mentioning the brand name by text, as that would have been easily detected, which points to their activity being illegitimate.
Similarity Matching Example
Similarity Matching is crucial for identifying copycat items that imitate the design of a product but do not mention the trademark in text, or use the exact same logo as the brand.
The example above shows a listing that copies the design of a well-known brand of sandals. Even though the listing does not mention the brand name – and has altered the logo on the product – similarity matching prioritized the listing based on images of the official product, with similarities in the shape.
OCR Example 1
Bad actors will embed text within images to avoid keyword detection. One of the ways in which bad actors try to evade enforcement, is by disguising the price in a listing to make it look legitimate.
In the example listing of replica luxury watches above, the seller uses prices in the standard layout that on the surface look legitimate. However, the true price is stated in the text embedded within the image – which is much lower – together with contact details for buyers to get in touch and carry out the transaction outside the platform.
A very low price is often a key signal for a counterfeit good – in this example that price signal was hidden with the image but detected using OCR.
OCR Example 2
This example of a listing for a ‘branded diamond ring’ is even more evasive and disguises the true price even further.
Rather than mentioning the true price, the seller states in Chinese at the top of the image that the buyer is to disregard the first 2 digits of the listing price. By showing a falsely high price, the seller is attempting to avoid tools that prioritize risk based on the listing price. They then also provide their WeChat ID at the bottom for buyers to enquire prices and browse the full product range.
Talisman’s Advanced Image Capabilities enable these infringements to be discovered and prioritized automatically. Similarity matching can detect the promotion of these items, regardless of how the seller describes it, and OCR automatically extracts the text from the images, regardless of what language it is written in.
Summary
In this Whiteboard Tutorial, we’ve detailed why image capabilities are critical for discovering high-risk infringements; explored the ‘Power of Three’ – Logo Detection, Similarity Matching, & OCR; and set out real-world examples of image capabilities being used to discover hidden bad actors.
Talisman, Incopro’s Brand Protection platform, is equipped with Advanced Image capabilities to identify high-risk threats impacting your brands that can’t be detected through traditional approaches.
Talk to one of our experts to find out more about Talisman’s Advanced Image Capabilities and how we can help you prioritize threats most visible to your consumers.
Whiteboard Tutorial: Network Analysis
Not all infringers pose the same risk. The biggest threats to your brands are often network-level operators that are structured like agile multinational businesses.
Network Analysis is a core module of Incopro’s Brand Protection platform Talisman that powers a strategic approach to tackling networks. Access our Whiteboard Tutorial to learn why it is critical to discover and act against bad actor networks and how Network Analysis technology is used to connect entities together – online and offline.
