Risk Scoring Detailed Overview
Part 1 - Introduction to Risk Scoring
Risk Scoring is a crucial tool for evaluating the safety of blockchain addresses, smart contracts, and individual transactions. Our system provides an initial score and adapts over time, reflecting new transaction data to keep you informed of potential risks in real-time.
The following guide explains how our Risk Scoring system works, detailing the components, processes, and categories of risk it detects, as well as practical examples and integration guidance to help you maximize its effectiveness.
Part 2 - Key Components and Processes of Risk Scoring
AI-Driven Analysis:
Smart Contracts: Anytime a smart contract is deployed or upgraded, we instantly perform detailed byte and op code analysis to identify the presence of malicious intent. This analysis is further enhanced through pattern recognition and intent prediction, leveraging both blockchain and web2 data.
EOAs (Externally Owned Accounts): For EOAs, we apply advanced pattern recognition and intent prediction techniques, supported by both blockchain and web2 data.
Heuristic Evaluation:
We complement the AI-driven analysis with heuristic rules and historical data, refining the initial risk score to ensure it aligns with known risk patterns. This adds an additional layer of reliability to the AI-generated scores.
Risk Score:
CUBE3.AI employs a combination of AI-driven analysis and heuristic evaluation to determine risk levels. Addresses and transactions are evaluated on a 1-100 scale, with the Risk Score representing a cumulative analysis of intent, patterns, and historical signals derived by our AI and ruleset — it’s not merely a confidence score. Use the following ranges to interpret risk scores:
0-30: Safe. Very unlikely to be malicious or have an illicit history.
31-69: Warning. Signals of malicious intent or history. Should be engaged with caution but not declined by default.
70-100: Unsafe. High likelihood of malicious intent or illicit history. Engage with utmost caution and we recommend declining by default.
Dynamic Scoring Evolution:
The risk score for EOAs and smart contracts evolves over time based on their transaction history, behavior, and any new signals received. This dynamic scoring ensures that the risk assessment remains accurate and reflects ongoing activities:
Healing: Positive transaction history can improve the score over time.
Sudden Increases: Suspicious or malicious transactions can cause a sudden rise in the risk score.
Transaction Risk Scoring:
Every transaction is assigned a risk score, synthesizing insights from EOAs and smart contracts interaction. If a transaction is proposed, we evaluate the involved parties, their respective risk scores, and the intent behind the transaction. If the transaction has been completed and is determined to be malicious, we provide not only the risk score but also the outcome details, such as identifying the victim, attacker, and flow of funds. This holistic approach ensures that each transaction is assessed with a full understanding of the associated risks.
Part 3 - Use Cases and Examples
To illustrate how our Risk Scoring system is applied, here are a few real-world examples across different scenarios:
3.1 - Preventing Fraud in Crypto Transactions
Scenario: An exchange customer initiates a transaction to purchase cryptocurrency from an unfamiliar address.
Application: Our system instantly assesses the risk scores of the involved EOAs and the transaction itself. The unfamiliar address is identified as an EOA associated with a phishing scam, with a high-risk score of over 90, indicating a strong likelihood of fraudulent activity.
Outcome: The transaction risk score is categorized as Unsafe, signaling a very high likelihood in malicious intent. The system alerts the exchange and ultimately advises the end-user to decline the transaction, preventing potential financial loss and protecting the user from fraud.
3.2 - Avoiding Transaction with Malicious Smart Contracts
Scenario: A new smart contract is deployed on an L1 blockchain. This contract is designed to exploit another smart contract but has no transaction history, and the deployer also has virtually no history.
Application: Upon deployment, our system immediately performs a detailed code analysis on the new contract. Despite the lack of history, the system identifies a pattern of malicious intent based on known exploit techniques and patterns.
Outcome: The contract is assigned a high-risk score, categorizing it as Unsafe. The system flags the contract, warning users and platforms to avoid interacting with it, thereby preventing potential exploitation and loss.
3.3. Ongoing Monitoring of EOAs
Scenario: A user’s wallet has been interacting with multiple addresses, some of which have questionable histories.
Application: Our system continuously monitors the wallet’s transaction history. Initially, the wallet had a low-risk score (20), but after interacting with a flagged address, its score suddenly increases to 65.
Outcome: The risk score is updated to Warning, signaling that the wallet should be monitored closely.
3.4 - Post-Transaction Analysis
Scenario: A transaction between two parties is completed, and our system detects it as fraudulent.
Application: Our system instantly analyzes the transaction, identifying the flow of funds, the victim, and the attacker. The risk score is updated to Unsafe due to the confirmed malicious intent. Additionally, our system automatically labels the attacker EOAs, any deployer EOAs of fraudulent contracts (if they existed), and the recipient EOAs of the stolen funds.
Outcome: The detailed outcome analysis, including the identification and labeling of involved parties.
3.5 - Identifying and Preventing Romance Scams
Scenario: A user is about to transfer cryptocurrency to an address associated with a new online acquaintance. This address had previously been identified by our system as likely belonging to a romance scammer.
Application: Using both web2 and blockchain data, our system had earlier flagged this address due to patterns consistent with romance scams. The address had been assigned a probabilistic risk score resulting in a Warning signal, indicating a significant risk and advising caution.
Outcome: When the user attempts to engage with this address, the system issues a warning based on the pre-existing risk score, advising the user to proceed carefully. This intervention potentially prevents the user from falling victim to the scam by flagging the suspicious activity before the transaction occurs.
Part 4 - Detection Categories
Our Risk Scoring system is designed to detect a broad range of threats across three primary categories:
Fraud: Detects deceptive activities, including romance scams, phishing fraud, token scams, ponzi schemes, and other fraudulent behaviors that aim to deceive users and misappropriate funds.
Cyber: Monitors EOAs and smart contracts for involvement in exploits or malicious intent.
Compliance: Identifies non-compliant EOAs and smart contracts, ensuring that interactions align with regulatory standards and avoiding entities that might pose a legal or reputational risk.
Your CUBE3.AI contact will provide a detailed list of threat categories detected by our models. I, please contact us to discuss how quickly we can add support.
Part 5 - Conclusion and Integration
CUBE3.AI’s Risk Scoring system provides a comprehensive and dynamic approach to identifying and mitigating risks associated with EOAs, smart contracts, and transactions. By leveraging AI-driven analysis, heuristic evaluation, and continuous monitoring, our platform ensures that users are equipped with the insights needed to make informed decisions and protect their assets.
To integrate our Risk Scoring system into your workflows, please refer to our Testing Guide. The guide provides detailed instructions on how to test and implement our solutions within your environment, ensuring a smooth and effective integration.
For additional support or specific inquiries, please contact your CUBE3.AI representative.
Part 6 - FAQs
1. What is Risk Scoring, and why is it important?
Risk Scoring is the process of evaluating the potential risk associated with EOAs, smart contracts, and transactions. It’s important because it helps users identify and avoid interactions that may involve fraud, exploits, or compliance risks, protecting their assets and maintaining the integrity of their platforms.
2. How does CUBE3.AI calculate the Risk Score?
CUBE3.AI calculates the Risk Score using AI-driven analysis, heuristic rules, and historical data. The AI analyzes smart contracts for malicious intent and EOAs for patterns of risky behavior, leveraging both blockchain and web2 data to produce a risk score ranging from 0 to 100.
3. What do Risk Scores mean?
Addresses and transactions are evaluated on a 1-100 scale, with the Risk Score representing a cumulative analysis of intent, patterns, and historical signals derived by our AI and ruleset — it’s not merely a confidence score.
Risk scores reflect the likelihood that an entity is safe or potentially harmful:
0-30: Safe. Very unlikely to be malicious or have an illicit history.
31-69: Warning. Signals of malicious intent or history. Should be engaged with caution but not declined by default.
70-100: Unsafe. High likelihood of malicious intent or illicit history. Engage with utmost caution and we recommend declining by default.
4. How does the Risk Score evolve over time?
The Risk Score evolves based on the entity’s transaction history, behavior, and any new signals received. Positive transactions can heal the score, improving it over time, while suspicious or malicious activities can cause a sudden increase in risk.
5. What happens if a transaction is flagged as malicious after completion?
If a transaction is flagged as malicious after completion, CUBE3.AI provides a transaction outcome overview, including the identification of the attacker, victim, and flow of funds. We also automatically label attacker EOAs, deployer EOAs of fraudulent contracts, and stolen money recipient EOAs.
6. What are the primary categories of risk detected by CUBE3.AI?
CUBE3.AI’s Risk Scoring system detects risks across three primary categories:
• Fraud: Detects deceptive activities like token scams and Ponzi schemes.
• Cyber: Monitors EOAs and smart contracts for potential exploits.
• Compliance: Identifies non-compliant EOAs and smart contracts.
7. Can I integrate CUBE3.AI’s Risk Scoring into my existing systems?
Yes, CUBE3.AI’s Risk Scoring can be integrated into your existing workflows. Please refer to our Testing Guide for detailed instructions on testing and implementation.
8. How quickly can new risk types be supported by CUBE3.AI?
If you have specific risk types that are not currently covered by our system, contact your CUBE3.AI representative. We can discuss how quickly we can add support for those categories based on your needs.
9. What should I do if I encounter a high-risk score?
If you encounter a high-risk score (Red), it is strongly recommended to avoid engaging with the associated EOA, smart contract, or transaction. High-risk scores indicate a very high likelihood of malicious intent or illicit history.
10. How does CUBE3.AI handle false positives?
While false positives are minimal, CUBE3.AI helps clients fine-tune the sensitivity of the Risk Scoring system to balance risk tolerance according to their specific business needs. This customization allows for an optimal balance between security and operational efficiency, reducing the likelihood of unnecessary alerts.
11. Who can I contact for more information or support?
For more information or support, you can contact your CUBE3.AI representative. They can provide detailed insights, help with integration, and offer guidance on how to best utilize our Risk Scoring system.
Last updated