Scammers now leverage Artificial Intelligence (AI), synthesized voice (so-called "deep fakes"), and caller ID spoofing, creating fraud of over $22 billion annually. In Phase I, we leveraged Decentralized Ledger Technology with blockchain encryption, real-time parsing of records, and real-time machine algorithms to block robocalls and reduce connection delays.

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase II project is to preserve the integrity of communication channels for business enterprises, financial institutions, and consumers. Attackers operate like businesses by investing in campaigns and generate returns by using stolen credentials to gain network access, inject ransomware, or commit direct fraud.

The innovation aims to prevent $4 B a year in financial scams through robust and enduring countermeasures. 90% of large enterprise customer breaches start from email or p2p/SMS messages that trick customers and employees into revealing sensitive information.

An average phishing attack spans 21 hours between the first and last victim and the detection of each attack occurs an average 9 hours after the first victim. This gives attackers a window of opportunity during which most of the damage is done. This project seeks to identify and protect against such attacks through a real-time machine learning system.

Phase I focused on filtering cyber attacks at the device level; initially applied to prevent robocalls, but relevant for other secure applications. The innovation also leveraged blockchain's shared storage and memory, ability to operate in a "trustless environment" (due to lack of cross-telecom network collaboration on centralized robocall lists), as well as advances in blockchain encryption, artificial intelligence and machine learning, and real-time parsing of records and machine algorithms to block robocalls and reduce connection delays.

In Phase I, we digitally transformed Federal Communication Commission and Federal Trade Commission compliant filing processes using Robotic Process Automation (RPA)/Intelligent Automation. We developed an Enterprise Knowledge Graph, Real-Time Machine Learner, and data auto-tagging using semantic metadata. The technology won the Federal Computer Week/Fed100 Award for Digital Transformation, and also the Service to the Citizens Award. See also US Patent# 11,689,660 B2.

Forward Edge-AI developed tools to automate data analytics. Leveraging our tools we analyze raw data to identify novel scams and predict events. The figure below shows the predictive analytics capabilities of the system. A snapshot of the data on January 5th showed a spike in data classified as "disinformation." The data returned to normal after January 14th, a week after the January 6th insurrection at the capital. Many of the techniques and processes of data analytics have been automated into algorithms that work over raw data for human consumption

​Our real-time machine learner was trained using 10,000+ state and federal court cases, 151 state and local statutes, and 7 federal laws (such as the Telecommunications Consumer Protection Act, and the Fair Debt Collection Act). As a result, the software is able use Natural Language Processing (NLP), and semantic reasoning to recognize messages that violate state and federal laws and regulations. Detection occurs in seconds versus the 30 - 90 days required by competitor solutions.

In Phase II, we developed a method for storing data in a blockchain in a high-performing manner, to ensure quality training of ML/AI models, training data (data governance) and production level inference. The innovation addressed the challenge of data drift and adversarial AI attacks where the provenance of the data is unknown. The blockchain also facilitates electronic discovery throughout the complaint filing and litigation process.

Phase II leveraged a Large Language Model (LLM/Generative AI (GAI)), Enterprise Knowledge Graph, and semantic reasoning to counter scams, misinformation, and disinformation. The detection and counter malign influence technology works in 28 different languages. A honeypot was developed that interacts autonomously with bots and malign influence actors. As a result, Forward Edge-AI won multiple Phase III SBIRs including:

The Phase II project leveraged swarm intelligence as a new Blockchain consensus mechanism specifically designed for more democratic online group-to-group collaboration in a distributed system. The project's innovations expand the understanding of telecommunications, the economics of smishing and vishing, and 5G security. Forward Edge-AI also is currently integrating ChatGPT into the software application to automatically respond to scam messages, and counter malign influence operations.

The technology uses NLP to transcribe spoken conversations and apply our Machine Learning models to determine which laws are being violated, and which calls are simply nuisance (e.g., not illegal, but can be classified as SPAM. The solution automatically predicts the fines and fees that can be levied based on the jurisdiction (state and federal).

The solution also leverages sentiment analysis to assess if the call is at risk for violation of the Fair Debt Collection Act (i.e., prohibitions against abusive language and pressure tactics).

Finally, this SBIR Phase II project develops a method for storing data in a blockchain in a high-performing manner, to ensure quality training of ML/AI models, and production level inference. The innovation addresses the challenge of data drift and adversarial AI attacks where the provenance of the data is unknown. The Phase II project also leverages swarm intelligence as a new consensus mechanism specifically designed for more democratic online group-to-group collaboration in a distributed system. The project's innovations will expand the understanding of telecommunications, the economics of smishing and vishing, and 5G security.

Most of the deep learning algorithms are heavily dependent on large-scale sample datasets, which has become a major factor that limits the application of these methods in some areas. Transfer learning is most effective when the source network has been trained with data that is similar to the target network.

Supervised learning, including incomplete supervision, inexact supervision, inaccurate supervision, or even unsupervised learning, may be an effective way to solve the problem of expensive data acquisition.

Achieving highly reliable object detection would be a significant success in and of itself. However, to fully leverage the power of Gabriel and provide the context, meaning, and relationships that can be then leveraged by AI applications, we combined the semantic data model, or ontology, with the data in an Enterprise graph database as shown below. Applying the semantic data model produced a true semantically enriched enterprise knowledge graph (Patent #11,689,660 B2 - An Enterprise Knowledge Graph, Blockchain, and Robotic Process Automation to Combat Scams, Misinformation and Disinformation).

Our approach demonstrated the additional benefit of enabling the incorporation of Intelligent Automation, to eliminate manual and error prone data entry and paperwork processes. For example, the Forward Edge-AI knowledge graph helps the machine understand that a pincer is also a plier, but used for different tasks, and that a pincer can have specific attributes that other pliers do not possess.

Our Knowledge graph allows the machine to make inferences about the data, which are the basis of a relevant recommendation engine, predictive analytics, and other advanced AI applications. Thereby enabling us to teach the machine to understand what is a scam.

Final Report Here

A Federal Agency may enter into a Phase III SBIR/STTR agreement at any time with a Phase I OR II Awardee. A subcontract to a Federally funded prime contract may be a Phase III award.

Awarding Agency: National Science Foundation

Grant Numbers: AN 1938135, AN 2028451

Contractor Name: Forward Edge-AI, Inc.

Contractor Address: 10108 Carter Canyon, San Antonio, TX 78255

Expiration of SBIR Data: 01 December 2041

Protection Period: 20 years from award of contract on 01 December 2021

The Government's rights to use, modify, reproduce, release, perform, display, or disclose technical data or computer software marked with this legend are restricted during the period shown as provided in paragraph (b)(5) of the Rights In Other Than Commercial Technical Data and Computer Software–Small Business Innovation Research (SBIR) Program clause contained in the above identified contract. After the expiration date shown above, the Government has perpetual government purpose rights as provided in paragraph (b)(5) of that clause. Any reproduction of technical data, computer software, or portions thereof marked with this legend must also reproduce the markings.

© 2024 Forward Edge-AI, Inc. All rights reserved.