AI tools claim to be 97% effective in preventing "Address Poisoning" attacks.

Source: Cointelegraph Original text: "AI tools claim to be 97% effective in preventing 'address poisoning' attacks"

The cryptocurrency cybersecurity company Trugard has developed an AI-based system in collaboration with the on-chain trust protocol Webacy to detect cryptocurrency wallet address poisoning attacks.

According to an announcement shared with Cointelegraph on May 21, this new tool is part of the Webacy crypto decision-making tool, "utilizing supervised machine learning models, combined with real-time trading data, on-chain analysis, feature engineering, and behavioral context for training."

It is claimed that this new tool has a success rate of up to 97% in known attack cases. Maika Isogawa, co-founder of Webacy, stated: "Address poisoning is one of the underreported but highly damaging scams in the cryptocurrency space, exploiting the simplest assumption: what you see is what you get."

Cryptocurrency address poisoning is a scam where attackers send a small amount of cryptocurrency from a wallet address that is very similar to the target's real address, usually with the same starting and ending characters. The goal is to trick users into inadvertently copying and using the attacker's address in future transactions, leading to financial loss.

This technique takes advantage of the habit of users who often rely on partial address matching or clipboard history when sending cryptocurrency. A January 2025 study found that between July 1, 2022 and June 30, 2024, there were more than 270 million address poisoning attempts on BNB Chain and Ethereum. Of those, 6,000 attempts were successful, resulting in losses of more than $83 million.

Jeremiah O’Connor, the Chief Technology Officer of Trugard, told Cointelegraph that their team brings deep cybersecurity expertise from the Web2 world and has "applied it to Web3 data since the early days of cryptocurrency." The team applies their experience in algorithm feature engineering from traditional systems to Web3. He added that:

"Most existing Web3 attack detection systems rely on static rules or basic transaction filtering. These methods often struggle to keep up with the ever-evolving strategies, techniques, and procedures of attackers."

The newly developed system uses machine learning to create a system capable of learning and adapting to address poisoning attacks. O’Connor emphasized that the uniqueness of their system lies in its "focus on context and pattern recognition." Isogawa explained, "AI can detect patterns that often exceed the analytical capabilities of humans."

O’Connor stated that Trugard generated synthetic training data for AI to simulate various attack patterns. The model was then trained using supervised learning, which is a type of machine learning that involves training the model on labeled data, including input variables and the correct output.

In this setup, the goal is to have the model learn the relationship between inputs and outputs in order to predict the correct output for new, unseen inputs. Common examples include spam detection, image classification, and price prediction.

O'Connor stated that with the emergence of new strategies, the model will also be updated by training on new data. He said, "The most important thing is that we have built a synthetic data generation layer that allows us to continuously test the model's performance against simulated toxic scenarios. This is very effective in helping the model generalize and maintain robustness over the long term."

Related news: Bold Technologies and My Aion launch a $2.5 billion smart city AI platform.