Granted Patent B2

US 12,657,234 · App. 18/381,168 · Granted Jun 16, 2026

Inventors: Michael Fallihee (Helena, MT); Sridhar Chandrashekar (Sammamish, WA); Mohan Thimmappa (Redmond, WA)

G06F16/383

Abstract

Claims (25)

1 . A computerized method for heuristic-based classification of software discovered connected devices, comprising:

implementing a discovery process on a set of devices and obtaining a set of attributes of each device, wherein each device comprise an industrial internet of things (IIOT) device;

identify a set of heuristics, wherein the set of heuristics combines an attribute from each discovered asset of the set of devices with an associated weight;

comparing the attribute from each asset and the device catalog and using the weight to determine a partial match score;

using the set of heuristics to reduce a search space used to identify the candidates for a friendly name of each device; and

calculating a dot product of the weight of each of the heuristics and using the dot product to identify a set of candidate devices in the device catalog,

wherein the set of attributes of each device is stored in a data table,

wherein the set of attributes of each device comprises a friendly name of each device that does not follow a standard taxonomy as provided in a device catalog,

wherein the friendly name is unstructured data and has a specific context that is esoteric to all but an entity that programmed the friendly name,

wherein the friendly name comprises a key identifier used by an operational team to identify a device,

wherein the friendly name comprises an informal taxonomy that is specific to a facility or a context in which the device with the friendly name is used, and

wherein a Levenshtein distance is used to compare each attribute between the device catalog and the discovered asset,

calculating a confidence level of each of the set of candidate devices and expressing the confidence level in the form of a probability value, and wherein each device candidate has a confidence level that is expressed in the form of a probability of match, and

wherein at least one heuristic comprises a conditional logic expression combining multiple attributes of the discovered asset, including a rule of the form “IF manufacturer=X AND model number=Y THEN apply a description-based comparison,

wherein the set of heuristics is experimentally derived and iteratively refined based on validation results from previously classified devices, and

wherein at least one heuristic assigns a disproportionately high weight to an attribute selected from manufacturer, model number, or device category to reduce the search space.

2 . The computerized method of claim 1 , wherein the heuristics comprises a logic combining multiple attributes.

3 . The computerized method of claim 2 , wherein the device catalog comprises plurality of devices and does not include a friendly name for each device.

4 . The computerized method of claim 3 , wherein the heuristics reduces the search space by giving a numerical score to potential matches.

5 . The computerized method of claim 4 , further comprising:

providing a specific weight to each heuristic such that each heuristic has a specific weight associated with the heuristic.

6 . The computerized method of claim 5 , further comprising:

providing the specific weight to each heuristic such that weight is used identify the importance that a specific heuristic has for the overall determination of the candidates.

7 . The computerized method of claim 6 , wherein a plurality of heuristics are provided and wherein a specific weight is assigned to each heuristic, and wherein each heurist is adjusted based on a plurality of endogenous parameters.

8 . The computerized method of claim 7 , plurality of endogenous parameters comprises a device category, an industry in which a device is used, and a specific customer who is using the device.

Assignments (2)

Continuity (3)

References Cited (26)