Contemplating the Impact of the USPTO’s AI-Focused Patent Eligibility Guidance on Networking Applications

For years, artificial intelligence (AI) has been deployed in the networking industry to make evaluations and predictions about computer networks for the purpose of improving overall efficiency, performance, and security.  For example, network assurance systems utilize AI to predict the impact of configuration changes, analyze network health, and diagnose potential threats and vulnerabilities. As another example, traffic routing schemes use machine learning (ML) models to anticipate congested communication paths and predict optimal routes for network traffic in view of a vast range of key performance indicators.  Unlike human administrators who take action in response to events that occur in the network, AI allows for predictive network management that is proactive, eliminating performance degradations and security breaches before they ever occur.

As companies compete to develop new AI solutions that offer customers incremental improvements in network operability, their need for patent protection continues to grow.  But patent applicants in this technology area often endure a thorn in their side in the form of 35 U.S.C. § 101, which establishes subject matter eligibility requirements that can render even the most innovative AI-powered networking tools unpatentable—if not claimed correctly.

Recently, the United States Patent and Trademark Office (USPTO) issued an important update to its guidance on patent subject matter eligibility under 35 U.S.C. § 101, focusing on AI and other software-related emerging technologies.  We previously summarized the USPTO’s update and provided our key takeaways here.  In this article, we contemplate the implications for networking applications, specifically, and the patent eligibility of claims directed to use of AI within the networking field.

The most instructive bit of guidance concerning networking and AI may be found in the USPTO’s new examples, particularly Example 47, which we summarize below.

Example 47: Anomaly Detection

Example 47 describes a hypothetical invention involving the use of an artificial neural network (ANN), which is a type of ML model consisting of an interconnected group of nodes, or “neurons,” generally configured to produce an output by applying a transformation function to an input, often received from another neuron.  In this invention, the ANN is specially trained to detect anomalies in a computer network, an improvement over traditional methods of detecting anomalies in both accuracy and efficiency.  The ANN may be implemented by an application-specific integrated circuit (ASIC), which itself may be customized for a specific AI application and provide superior computing capabilities compared to traditional computer processing units (CPUs).

The Example also describes hypothetical embodiments disclosed in the patent application.  For instance, the ANN may detect anomalies indicating potential network intrusions or malicious attacks.  If the detected anomaly is associated with a malicious packet, the ANN may cause a network device to drop the malicious packet.  Furthermore, the system may identify the source of the packet and take various remedial actions, including alerting a network administrator or blocking future traffic from the source address, thereby enhancing network security by allowing for automatic, proactive remediation of network attacks.

The invention is claimed differently in three example independent claims, one of which is deemed ineligible and the other two eligible.  A comparison of each provides insight on the patent eligibility of claims directed to networking applications that utilize AI. 

Claim 1 – ELIGIBLE

1.         An application specific integrated circuit (ASIC) for an artificial neural network (ANN), the ASIC comprising:

a plurality of neurons organized in an array, wherein each neuron comprises a register, a microprocessor, and at least one input; and

a plurality of synaptic circuits, each synaptic circuit including a memory for storing a synaptic weight, wherein each neuron is connected to at least one other neuron via one of the plurality of synaptic circuits.

Eligibility Summary

The guidance update indicates that claim 1 is patent eligible since there is no judicial exception recited in the claim.  Unlike claims 2 and 3 which recite an abstract idea, as discussed below, claim 1 does not set forth any abstract ideas (i.e., mental process, mathematical concept, or method of organizing human activity). MPEP 2106.04(a)(2).  Even though ANNs may be trained using mathematic processes, as the guidance notes, claim 1 fundamentally recites hardware components in the form of neurons, comprising a register and a microprocessor, and a plurality of synaptic circuits, which together form an ANN.  Because the claim is directed to hardware and does not recite a judicial exception (Step 2A, Prong One), the claim is eligible.

Claim 2 – INELIGIBLE

2.         A method of using an artificial neural network (ANN) comprising:

(a) receiving, at a computer, continuous training data;

(b) discretizing, by the computer, the continuous training data to generate input data;

(c) training, by the computer, the ANN based on the input data and a selected training algorithm to generate a trained ANN, wherein the selected training algorithm includes a backpropagation algorithm and a gradient descent algorithm;

(d) detecting one or more anomalies in a data set using the trained ANN;

(e) analyzing the one or more detected anomalies using the trained ANN to generate

anomaly data; and 

(f) outputting the anomaly data from the trained ANN.

Eligibility Summary

The guidance update indicates that claim 2 is patent ineligible since the claim is directed to a judicial exception without integrating the exception into a practical application.  Under the broadest reasonable interpretation, steps (b) through (e) recite judicial exceptions in the form of abstract ideas (Step 2A, Prong One).  Specifically, steps (b), (d), and step (e) fall within the “mental process” category of abstract ideas since they describe concepts performed in the human mind, while steps (b) and (c) fall within the “mathematical concepts” category of abstract ideas since they describe processes (discretization) or algorithms (backpropagation and gradient descent algorithms) that are fundamentally rooted in mathematics. MPEP 2106.04(a)(2). Considering whether there are any additional elements in the claim beyond these abstract ideas and whether the claim as a whole integrates the abstract ideas into a practical application (Step 2A, Prong Two), the guidance notes that steps (a), (b), and (c) are recited as being performed by a computer.  But the computer is recited at such a high level of generality that it amounts to mere instructions to apply the exception using a generic computer. MPEP 2106.05(f). Similarly, steps (d) and (e), which each recite “using the trained ANN,” provide nothing more than mere instructions to implement an abstract idea on a generic computer.  Although some may argue that “using a trained ANN” would integrate the judicial exception into a practical application, the limitation merely confines the exception to a certain field of use or technological environment (neural networks). MPEP 2106.05(h).  The guidance finally indicates that steps (a) and (f) represent insignificant extra-solution activity as they recite operations that are well understood, routine and conventional in the networking field. MPEP 2106.05(g).  Therefore, even when considered in combination, these claimed elements recite insignificant extra-solution activity and mere instructions to implement an abstract idea on a generic computer, which fail to provide an inventive concept (Step 2B), rendering the claim ineligible.

Claim 3 – ELIGIBLE

3.         A method of using an artificial neural network (ANN) to detect malicious network packets comprising:

(a) training, by a computer, the ANN based on input data and a selected training algorithm to generate a trained ANN, wherein the selected training algorithm includes a backpropagation algorithm and a gradient descent algorithm;

(b) detecting one or more anomalies in network traffic using the trained ANN;

(c) determining at least one detected anomaly is associated with one or more malicious network packets; 

(d) detecting a source address associated with the one or more malicious network packets in real time; 

(e) dropping the one or more malicious network packets in real time; and

(f) blocking future traffic from the source address. 

Eligibility Summary

The guidance update indicates that claim 3 is patent eligible since, despite reciting a judicial exception, the claim recites additional elements that integrate the exception into a practical application by improving the functioning of a computer or technical field.  The key distinction between claims 2 and 3 rests in the inclusion of steps (d) through (f) in claim 3.  While the abstract idea analysis (Step 2A, Prong One) for steps (a) and (b) is consistent with claim 2, the practical application analysis (Step 2A, Prong Two) for claim 3 diverges, since one way to establish integration into a practical application is when the claimed invention improves the functioning of a computer or another technology or technical field. MPEP 2106.04(d)(1). Focusing on steps (d) through (f), the guidance states that the steps provide for improved network security using the detected anomalies to enhance security by taking proactive measures to remediate the potential threat by detecting the source address associated with the malicious packets, and then dropping such packets and blocking future traffic from the source address.  Importantly, the guidance focuses on the specification since it must set forth an

improvement in technology, and the claim itself must reflect the disclosed improvement, in order to establish an improvement to the functioning of a computer or a technical field. MPEP 2106.04(d)(1) and 2106.05(a). Because the claim as a whole integrates the recited exception into a practical application by reciting an improvement to a technical field (network intrusion detection), the claim is eligible. 

Key Takeaways for Drafting Claims Reciting Usage of AI in the Networking Field

Any claim directed to networking applications, including applications incorporating AI technology, should be drafted with patent eligibility in mind. As shown above, the examples provided in the subject matter eligibility guidance represent insights we can use to develop claims that are less likely to encounter a subject matter eligibility rejection under 35 U.S.C. § 101. 

1. Avoid relying on generic limitations like “by a computer” or “using a trained ANN” for patent eligibility.  The guidance update makes clear that such limitations recited at a high-level of generality amount to mere instructions to apply an exception using a generic computer or merely confine the exception to a certain field of use or technological environment.  Take care to understand what is well understood, routine and conventional in the networking field so as to identify what claimed elements might be considered insignificant extra-solution activity.
2. Rely on “improvements to the functioning of a computer or to any other technology or technical field” as a basis for establishing that elements in the claim integrate an exception into a practical application.  Crucially, ensure that the specification describes the improvement in technology—e.g., network intrusion detection, traffic routing, network health monitoring, etc.—in detail, including how the invention improves upon conventional approaches, and that the claim itself reflects the disclosed improvement.
3. Consider whether the invention can be claimed as a combination of hardware components. While some practitioners may argue that method claims offer the broadest possible coverage for most networking-related innovations (network anomaly detection, for example), claiming the invention in hardware terms as an ASIC or other apparatus represents a viable route to eligibility. 

Subscribe To Viewpoints