Patent Application
20210182628
Phishing attacks generally involve an attacker attempting to steal sensitive information using fraudulent webpages that impersonate the legitimate brand webpages. As such, those fraudulent webpages mimic the legitimate ones, including one or more logos of the impersonated brands that are present in the legitimate pages. As the phishers's work product increases in quality, it has become increasingly difficult to detect fraudulent webpages from the legitimate ones.
Embodiments are drawn to computer-implemented methods and systems for combining object detection models, and more specifically the combination of logo detection models, to output a better combined detection given an input image. These computer-implemented methods and systems may be used to detect brand logos in images and help with the detection and characterization of phishing attacks where an attacker attempts to steal sensitive information using fraudulent webpages impersonating the legitimate brand webpages. Those fraudulent webpages mimic the legitimate ones, including one or more logos of the impersonated brands that are present in the legitimate pages. By applying logo detection to images representing webpages, emails or any other kind of electronic documents, a better characterization of the phishing attempt may be derived and the phished brand may be detected with a greater degree of certainty. The term “logo”, as used herein, includes within its scope any graphic mark, emblem, or symbol used to aid and promote public identification and recognition. It may be of an abstract or figurative design or include the text of the name it represents as in a wordmark. The term “logo” also includes most any graphical representation of most anything (and slight variations thereof), as the object detectors referred to herein may be trained with most any kind of annotated training images.
Object detection is a machine learning task for which an estimator (referred here as an object detector) learns, given annotated images, to detect objects on new images, such that each object detected on an image is associated to an object class (typically the object type), a confidence score (typically a float in the [0,1] range), and its position in the image (e.g., the coordinates of the bounding box in the pixel's space).
A number of research efforts have been made to improve the accuracy of single estimators for a specific image processing task. Similarly, research has been carried out to determine the manner in which multiple estimators may be combined to improve performance, resulting in new “ensemble” models, such as a “Random Forest”, which is a combination of decision trees. Embodiments, therefore, are drawn to computer-implemented methods and systems for combining several estimators' predictions in the context of object detection.
Estimator Combination and Logo Detection
Logo detection is a particular case of object detection. Object detection in computer vision is both a classification and regression problem. Indeed, given an input image, the objective is to output detections; that is, to predict locations of bounding boxes containing target objects and their corresponding classes. Detections are based on constraints induced by the annotation strategy and the algorithm's loss function. A detection's bounding box may be rectangular, may contain only one object and may be of similar size as the object it contains. The input and output of an object detection algorithm is detailed in the table below:
Similarly, to other computer vision tasks, object detection task generally relies on Convolutional Neural Networks (CNNs). CNNs are a class of deep neural networks, most commonly applied to analyzing visual imagery. For example, CNNs may include the SSD version of the ResNet-50 and VGG-16 algorithms. According to one embodiment, a CNN may be used to detect logos. Pixm, Inc., for example, includes in their product pipeline a CNN to detect logos and icons in order to flag suspicious websites or emails. Research in the field is also active and multiple methods based on CNNs have been proposed recently to improve object detection performance.
A well-known approach in machine learning to improve performance on a given task is to combine different estimators (e.g. SVM, CNN). Indeed, combining estimators allows to reduce generalization error. Empirically, estimators ensembles tend to yield better results when there is a significant diversity among estimators (i.e. when estimators errors are not correlated). Diversity between estimators can be increased using various levers such as the training data (e.g. data augmentation, bagging), the estimator algorithm (e.g. SVM, Logistic Regression) or the architecture and training parameters with neural networks. Proposals have been made, for example, to create a set of diverse CNNs in order to combine them and accurately classify objects.
Besides estimators diversity, the combination method also impacts performance of the estimators ensemble. Different methods have been proposed to combine estimators such as voting, the Demptser-Shafer theory, and other machine learning algorithms. Other methods such as boosting tackles both estimators diversity and combination.
In the context of object detection where each estimator (object detector) can make several candidate detections, each of them with their own location (see Table 1—Definition of inputs and outputs of an object detector), specific combining methods have been proposed, which make use of the overlap between detections from different object detectors. For instance, detections using a machine learning algorithm may be combined to rank candidate detections on each image. The ranking algorithm's features includes information regarding the degree to which each detection overlaps with others and the object-object relations likelihood. Overlapping low-rank detections are discarded.
Other methods to cluster detections are based on overlap, with a score being computed for each cluster. In order to compute the cluster's score, such methods combine scores given by detections within clusters using, for instance, the Demptser-Shaffer theory. Once each cluster has been assigned a score, they may be filtered and redundant detections may be removed according to some criteria (e.g. non-max suppression).
One embodiment is configured to combine detections from multiple object detectors through successive filtering operations in order to output an optimal set of combined detections. The resulting set of combined detections performs better than any set outputted by object detectors taken individually. Indeed, one embodiment of the present computer-implemented method may comprise two filtering steps, such that the optimal set of detections is generated at the end of the second step. These steps include a first step (Step 1 herein) of prior performance-based filtering and a second step (Step 2 herein) of score fusion filtering.
The following data are defined:
The embodiments described and shown herein may comprise combining the detections outputted by several object detectors on one image such that: LIm≤Σi=1nlP
The following two phases are defined:
Prior Performance-Based Filtering
According to one embodiment, Step 1, shown at 2006 in
We now describe bounding box filtering through the overlapping criteria as well as giving further details on how the prior knowledge database 3010 can be built.
Overlapping Based Rules
The objective of the first stage of detections filtering is to discard redundant or incorrect detections based on their overlap with detections that have been added to the kept detection set 2010, i.e. detections expected to be correct (once again, this set is initially empty).
A detection D1 defined by (cls1, s1, b1) is redundant if its bounding box overlaps significantly with the bounding box of a detection D2 defined by (cls2,s2, b2) expected to be correct (present in the kept detection set 2010) such that ƒoverlap(b1,b2)>overlap* and if the two detections predict the same object class, i.e. cls1=cls2. Under such conditions, D1 and D2 are likely to detect the same object.
A detection D1 defined as (cls1, s1, b1) is incorrect if its bounding box overlaps significantly with the bounding box of a detection D2 defined by (cls2, s2, b2) expected to be correct (present in the kept detection set 2010) such that ƒoverlap(b1, b2)>overlap* and if the two detections recognize different object classes, i.e. cls1≠cls2. Indeed, in this case, D1 and D2 have detected an object at the same spatial position on the image but differ in their prediction of the object class. As D2 is expected to be correct (already present in the kept detection set 2010), D1 must be discarded; e.g., added to the discarded detections store 2008. The overlapping metric ƒoverlap (e.g. IoU) and the overlapping threshold overlap* (e.g., IoU=0.5) may be determined by an expert. In particular, once ƒoverlap has been chosen, the overlapping threshold overlap* may be determined using a trivial iteration process on the value of overlap* for example.
Prior Knowledge Construction
In order to build the prior knowledge database 3010 in
We define DP
The performance related to the score sα and detector Pi may then be computed using the performance metric ƒperf. ƒperf takes as input the detection subset DP
For each detector, this operation may be repeated several times with different sα values. For instance, as class confidences scores are comprised in [0,1], we may choose sα values from 0 to 1 with a step of 0.05. We can then plot performance graphs for each detector with sα values as abscissa and the corresponding performances as ordinate.
Based on these performance graphs, we can determine the set of confidence thresholds {sP
For instance, we define ƒperf to be the precision in information retrieval settings. In these settings, precision may be defined as follow:
is the number of True Positives, i.e. the number of correct detections; FP the number of False Positives, i.e. the number of incorrect detections. We may choose perf*=0.99 as the precision on V above which an object detector Pi is expected to output correct detections. Indeed, if a detector Pi has its precision above 0.99 on V, we may expect that detections from Pi are correct at prediction time.
We also introduce l* to mitigate the risk of generalizing the performance observed on a small subset of detections. Indeed, lP
With perf* and l* fixed, the prior knowledge database 3010 may be built as follow at parameters setting time: for each performance graph Graphi we search for the smallest sα among the values chosen such that the performance of the detection subset given by sα is strictly greater than perf* and the number of detections in the subset is strictly greater than l*, i.e. ƒperf(DP
Score Fusion Filtering
The second step of filtering (i.e., Step 2 2012 in
We define the following notations:
Part 1: Cluster Detections
In order to filter detections 4002 that have not been rejected or kept after Step 1 filtering, detections may be clustered, as shown at 4004 in
The functions ƒsimilarity, ƒcluster, ƒcleaning should be defined at parameters settings time. First, the similarity matrix MIm may be computed using the similarity metric ƒsimilarity on the set of unfiltered detections resulting from Step 1 on an image Im. Then the chosen clustering algorithm ƒcluster may be applied to output the set of clusters C′Im. Finally, after applying ƒcleaning on each cluster of the set C′Im, a new set CIm may be output such that each cluster from CIm 4006 contains at most one detection from each object detector.
Part 2: Score Clusters
Once the clusters are created, they may be scored based on the detections present in the clusters as well as the performance of the object detectors that made those detections, as suggested in
ƒaggregate should be defined at parameters setting time. We apply ƒaggregate on each cluster of the set CIm and associate each cluster to its score before Filter clusters operation.
Part 3: Filter Clusters
After applying f cleaning, each cluster contains at most one detection from each object detector. Clusters may then be filtered, as shown at 4012 and
As shown in
as shown at 4020 with ƒconf(CxIm) the configuration of CxIm, as shown at 4018. Clusters whose scores are strictly greater than their respective configuration threshold may be kept at 4024 and further processed, while the other clusters may be discarded at 2008 with all their associated detections.
The cluster configuration thresholds 4022 should be determined at parameters setting time. For instance, with all other elements fixed, the combination algorithm on V may be iterated several times with different values of cluster configuration thresholds. The set of values that give the best detection combination on V according to a defined performance metric would be kept at prediction time. According to one embodiment, each threshold may be determined using a hyperparameter optimization method on an annotated object detection dataset. In one embodiment, the hyperparameter optimization may comprise a random search method. Random search is a method to perform hyper-parameter optimization, that is, a method to find close-to-optimal combination of hyper-parameters, for a given model, performance metric and test dataset.
An example of cluster filtering at prediction time is shown below, with reference to
In order to distinguish detection from the object detectors, the following marking conventions may be adopted:
In
The table below summarizes the clusters shown in
Based on the table above, the following clusters may be kept {C1Im
Part 4: Select Detections
Finally, for each cluster kept 4024 after the previous filtering operation, a single detection may be outputted that represents the object predicted by the detections from that cluster. To do so, the following function 4026 may be defined:
The function ƒselect 4026 is defined at parameters setting time. ƒselect is applied on each kept cluster returned by Filter cluster operation. After applying ƒselect all detections are filtered, i.e. kept 2010 or discarded 2008, there are no unfiltered detections left. The kept detections 2010 are returned and form the optimal set of detections 2014.
Exemplary Use Case
In this use case, logo detection is used to detect a phishing URL in the context of a Universal Resource Locator (URL) scanning service. See
1. A spambot 6001 generates a phishing email containing the phishing URL: http://phishingdomain.com/index.php, as shown at 6002. The recipient of the phishing email is john.doe@domain.com, as shown at 6020.
2. In order to send the phishing email to john.doe@domain.com 6020, the spambot 6001 looks up the Domain Naming Server Mail exchanger (DNS MX) record associated to domain.com. The DNS MX record specifies the Mail Transfer Agent (MTA) 6004 responsible for accepting messages on behalf of the domain name.
3. Using Simple Mail Transfer Protocol (SMTP), the spambot 6001 connects to the MTA 6004 specified in the looked-up DNS MX and then sends the content of the phishing email.
4. When the MTA 6004 receives the email, it first applies a spam filter 6006 to detect and block unsolicited emails such as spam, phishing, etc. A large proportion of the unsolicited email traffic is usually detected and blocked, but there are also many unsolicited emails that are not detected and blocked: we consider that the phishing email mentioned in Step 1 is not detected and blocked. For the emails that have not been blocked, the MTA 6004 may then apply a URL rewriting mechanism 6008 to protect the end user at the time of click: the URLs in the phishing email are rewritten so that they point to a URL scanning service 6010 that will analyze the original URL when the end user clicks on a rewritten URL. In this example, http://urlscanningservice.com designates the URL scanning service and http://urlscanningservice.com/url/aHR0cDovL3BoaXNoaW5nZG9tYWluLmNvbS9pbmRleC5waHA= is the rewriting of http://phishingdomain.com/index.php URL where aHROcDovL3BoaXNoaW5nZG9tYWluLmNvbS9pbmRleC5waHA= is the encoding of http://phishingdomain.com/index.php in Base64.
5. Using SMTP, the MTA then sends the emails to the Mail Delivery Agent (MDA) 6012.
6. The MDA 6012 stores the email in the Mail Store 6014.
7. The end user john.doe@domain.com 6020 launches his mail client software, also known as Mail User Agent (MUA) 6016. The MUA 6016 fetches new emails from the mail store 6014, typically with POP3 or IMAP protocol. The MDA 6012 usually acts as the POP3 and/or IMAP server. The MUA 6016 fetches the phishing email containing the rewritten phishing URL.
8. The end user opens the phishing email and clicks on http://urlscanningservice.com/url/aHROcDovL3BoaXNoaW5nZG9tYWluLmNvbS9pbmRleC5waHA=.
9. The URL scanning service 6010 decodes the aHROcDovL3BoaXNoaW5nZG9tYWluLmNvbS9pbmRleC5waHA= Base64 encoded value. The URL scanning service 6010 then analyzes the http://phishingdomain.com/index.php URL. For this purpose, it extracts features from the URL and associated webpage such as: URL domain DNS information, URL domain WHOIS information, HTML content of the webpage, graphical rendering of the webpage, etc. The URL scanning services then applies one or several algorithms on the features to determine if the URL is a phishing URL. Examples of such algorithms are fingerprints algorithms, decision tree, supervised learning algorithm (SVM, Random Forest and the like), among other detection technologies. In this use case, we consider that the URL scanning service extracts one or several logo(s) from the graphical rendering of the webpage associated to the analyzed URL (
10. Logo Detection API 6022 is a REST API that exposes the Logo Detection function 6024. The Logo Detection function 6024 analyzes the graphical rendering of the webpage and then extracts one or several brand logo(s) using an embodiment of the present computer-implemented method, as shown and described herein. The result is returned to the URL scanning service 6010.
11. The URL scanning service 6010 has extracted all the features from the URL and associated webpage, including the fact that the graphical rendering of the webpage contains one or several known brand logo(s) which is indicative of a potential phishing. The URL scanning service 6010 then applies one or several algorithms on the features and as such determine that the URL is indeed a phishing URL.
12. Consequently, the URL scanning service redirects the end user to a safe webpage indicating that the URL is a phishing URL.
Physical Hardware
As shown, the storage device 807 may include direct access data storage devices such as magnetic disks 830, non-volatile semiconductor memories (EEPROM, Flash, etc.) 832, a hybrid data storage device comprising both magnetic disks and non-volatile semiconductor memories, as suggested at 831. References 804, 806 and 807 are examples of tangible, non-transitory computer-readable media having data stored thereon representing sequences of instructions which, when executed by one or more computing devices, implement aspects of the embodiments described and shown herein. Some of these instructions may be stored locally in a client computing device, while others of these instructions may be stored (and/or executed) remotely and communicated to the client computing over the network 826. In other embodiments, all of these instructions may be stored locally in the client or other standalone computing device, while in still other embodiments, all of these instructions are stored and executed remotely (e.g., in one or more remote servers) and the results communicated to the client computing device. In yet another embodiment, the instructions (processing logic) may be stored on another form of a tangible, non-transitory computer readable medium, such as shown at 828. For example, reference 828 may be implemented as an optical (or some other storage technology) disk, which may constitute a suitable data carrier to load the instructions stored thereon onto one or more computing devices, thereby re-configuring the computing device(s) to one or more of the embodiments described and shown herein. In other implementations, reference 828 may be embodied as an encrypted solid-state drive. Other implementations are possible.
Embodiments of the present invention are related to the use of computing devices for combining detection models, as shown and described herein. According to one embodiment, the methods, devices and systems described herein may be provided by one or more computing devices in response to processor(s) 802 executing sequences of instructions, embodying aspects of the computer-implemented methods shown and described herein, contained in memory 804. Such instructions may be read into memory 804 from another computer-readable medium, such as data storage device 807 or another (optical, magnetic, etc.) data carrier, such as shown at 828. Execution of the sequences of instructions contained in memory 804 causes processor(s) 802 to perform the steps and have the functionality described herein. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions to implement the described embodiments. Thus, embodiments are not limited to any specific combination of hardware circuitry and software. Indeed, it should be understood by those skilled in the art that any suitable computer system may implement the functionality described herein. The computing devices may include one or a plurality of microprocessors working to perform the desired functions. In one embodiment, the instructions executed by the microprocessor or microprocessors are operable to cause the microprocessor(s) to perform the steps described herein. The instructions may be stored in any computer-readable medium. In one embodiment, they may be stored on a non-volatile semiconductor memory external to the microprocessor, or integrated with the microprocessor. In another embodiment, the instructions may be stored on a disk and read into a volatile semiconductor memory before execution by the microprocessor.
Accordingly, one embodiment is a computer-implemented method of detecting logos in a graphical rendering, comprising detecting, using a first trained object detector, logos in the graphical rendering and outputting a first list of detections; detecting, using a second trained object detector, logos in the graphical rendering and outputting a second list of detections; filtering, using a first and a second prior performance-based filter, the received first and second lists of detections into a first group of kept detections, a second group of discarded detections and a third group of detections; clustering detections in the third group of detections, if any, in at least one cluster comprising detections that are of a same class and that are generally co-located within the electronic image; assigning a cluster score to each cluster; and outputting a set of detections of logos in the graphical rendering, the set comprising the detections in the first group and a detection from each of the clusters whose assigned cluster score is greater than a respective threshold. Each threshold may be specific to a set of one or more of the first and second trained object detectors.
According to further embodiments, the first and/or second trained object detectors may comprise a convolution neural network (CNN)-based detector. The CNN-based detector may comprise one of SSD ResNet-50 and SSD VGG-16, for example. Each detection in the first to third groups of detections may comprise a tuple comprising a predicted class, a class confidence score and coordinates of a bounding box of a detected logo in the graphical rendering. According to one embodiment, filtering further may comprise: generating the first prior performance-based filter by testing the first trained object detector on a first annotated object detection dataset; and generating the second prior performance-based filter by testing the second trained object detector on a second annotated object detection dataset.
The first annotated object detection dataset and the second annotated object detection dataset may be the same. In one embodiment, filtering may be based on: a first prior knowledge value specific to the first trained object detector and a first confidence score associated with each detection in the first list of detections; and a second prior knowledge value specific to the second trained object detector and a second confidence score associated with each detection in the second list of detections. The first group of kept detections may comprise detections that are to be included in the outputted set of detections of logos, the second group of discarded detections may comprise detections that are discarded and not included in the outputted set of detections of logos; and the third group may comprise detections that require further processing to determine whether they are to be discarded into the second group or included in the first group. In one embodiment, clustering detections in the third group of detections that are generally co-located within the electronic image may comprise clustering detections that have overlapping bounding boxes within the electronic image. Clustering detections that have overlapping bounding boxes within the electronic image may comprise clustering detections that have bounding boxes whose Intersection Over Union (IoU) is greater than an overlapping threshold. According to one embodiment, assigning a cluster score to each cluster may comprise calculating the cluster score based on the confidence scores of the detections in the cluster for which the cluster score is being calculated. Calculating the cluster score may comprise using an aggregation function. For each cluster, the cluster score may comprise an average of the confidence scores of the detections in the cluster.
In one embodiment, the computer-implemented method may further comprise determining each threshold using a hyperparameter optimization method on an annotated object detection dataset. The hyperparameter optimization method may comprise, for example, a random search method. The computer-implemented method may further comprise denoting each cluster having a cluster score that is greater than a predetermined cluster threshold as a pertinent cluster that is associated to a single detection that represents the cluster. In one embodiment, the single detection representing the cluster is one of the detections contained in the cluster. The computer-implemented method may further comprise adding pertinent clusters to the first group of kept detections.
Another embodiment is a computing device that may comprise at least one processor; at least one data storage device coupled to the at least one processor; a network interface coupled to the at least one processor and to a computer network and a plurality of processes spawned by the at least one processor to detect logos in a graphical rendering. The processes may include processing logic for: detecting, using a first trained object detector, logos in the graphical rendering and outputting a first list of detections; detecting, using a second trained object detector, logos in the graphical rendering and outputting a second list of detections; filtering, using a first and a second prior performance-based filter, the received first and second lists of detections into a first group of kept detections, a second group of discarded detections and a third group of detections; clustering detections in the third group of detections, if any, in at least one cluster comprising detections that are of a same class and that are generally co-located within the electronic image; assigning a cluster score to each cluster; and outputting a set of detections of logos in the graphical rendering, the set comprising the detections in the first group and a detection from each of the clusters whose assigned cluster score is greater than a respective threshold.
According to one embodiment, at least one of the first and second trained object detectors may comprise a convolution neural network (CNN)-based detector. The CNN-based detector may comprise one of SSD ResNet-50 and SSD VGG-16, for example. Each detection in the first to third groups of detections may comprise a tuple comprising a predicted class, a class confidence score and coordinates of a bounding box of a detected logo in the graphical rendering. The processing logic for filtering further may comprise processing logic for: generating the first prior performance-based filter by testing the first trained object detector on a first annotated object detection dataset; and generating the second prior performance-based filter by testing the second trained object detector on a second annotated object detection dataset. In one embodiment, the first annotated object detection dataset and the second annotated object detection dataset are the same.
According to one embodiment, filtering may be based on: a first prior knowledge value specific to the first trained object detector and a first confidence score associated with each detection in the first list of detections; and a second prior knowledge value specific to the second trained object detector and a second confidence score associated with each detection in the second list of detections. The first group of kept detections may comprise detections that are to be included in the outputted set of detections of logos; the second group of discarded detections may comprise detections that are discarded and not included in the outputted set of detections of logos; and the third group may comprise detections that require further processing to determine whether they are to be discarded into the second group or included in the first group. The processing logic for clustering detections in the third group of detections that are generally co-located within the electronic image may comprise processing logic for clustering detections that have overlapping bounding boxes within the electronic image. The processing logic for clustering detections that have overlapping bounding boxes within the electronic image may comprise processing logic for clustering detections that have bounding boxes whose Intersection Over Union (IoU) may be greater than an overlapping threshold.
In one embodiment, the processing logic for assigning a cluster score to each cluster may comprise processing logic for calculating the cluster score based on the confidence scores of the detections in the cluster for which the cluster score may be being calculated. The processing logic for calculating the cluster score may comprise processing logic for using an aggregation function. For each cluster, the cluster score may comprise an average of the confidence scores of the detections in the cluster. Each threshold may be specific to a set of one or more of the first and second trained object detectors.
The computing device, according to one embodiment, may further comprise processing logic for determining each threshold using a hyperparameter optimization method on an annotated object detection dataset. The hyperparameter optimization method may comprise a random search method. Processing logic may also be provided for denoting each cluster having a cluster score that is greater than a predetermined cluster threshold as a pertinent cluster that is associated to a single detection that represents the cluster. The single detection representing the cluster may be one of the detections contained in the cluster. Processing logic may be provided for adding pertinent clusters to the first group of kept detections.
Portions of the detailed description above describe processes and symbolic representations of operations by computing devices that may include computer components, including a local processing unit, memory storage devices for the local processing unit, display devices, and input devices. Furthermore, such processes and operations may utilize computer components in a heterogeneous distributed computing environment including, for example, remote file servers, computer servers, and memory storage devices. These distributed computing components may be accessible to the local processing unit by a communication network.
The processes and operations performed by the computer include the manipulation of data bits by a local processing unit and/or remote server and the maintenance of these bits within data structures resident in one or more of the local or remote memory storage devices. These data structures impose a physical organization upon the collection of data bits stored within a memory storage device and represent electromagnetic spectrum elements. Moreover, the computer-implemented methods disclosed herein improve the functioning of computers by enabling the migration of filesystems from a donor filesystem to a beneficiary filesystem while commands are issued and executed to change the metadata and data thereof. Such computer-implemented methods are not capable of being effectively carried out by the mental processes of humans.
A process, such as the computer-implemented methods described and shown herein, may generally be defined as being a sequence of computer-executed steps leading to a desired result. These steps generally require physical manipulations of physical quantities. Usually, though not necessarily, these quantities may take the form of electrical, magnetic, or optical signals capable of being stored, transferred, combined, compared, or otherwise manipulated. It is conventional for those skilled in the art to refer to these signals as bits or bytes (when they have binary logic levels), pixel values, works, values, elements, symbols, characters, terms, numbers, points, records, objects, images, files, directories, subdirectories, or the like. It should be kept in mind, however, that these and similar terms should be associated with appropriate physical quantities for computer operations, and that these terms are merely conventional labels applied to physical quantities that exist within and during operation of the computer.
It should also be understood that manipulations within the computer are often referred to in terms such as adding, comparing, moving, positioning, placing, illuminating, removing, altering and the like. The operations described herein are machine operations performed in conjunction with various input provided by a human or artificial intelligence agent operator or user that interacts with the computer. The machines used for performing the operations described herein include local or remote general-purpose digital computers or other similar computing devices.
In addition, it should be understood that the programs, processes, methods, etc. described herein are not related or limited to any particular computer or apparatus nor are they related or limited to any particular communication network architecture. Rather, various types of general-purpose hardware machines may be used with program modules constructed in accordance with the teachings described herein. Similarly, it may prove advantageous to construct a specialized apparatus to perform the method steps described herein by way of dedicated computer systems in a specific network architecture with hard-wired logic or programs stored in nonvolatile memory, such as read only memory.
While certain example embodiments have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the embodiments disclosed herein. Thus, nothing in the foregoing description is intended to imply that any particular feature, characteristic, step, module, or block is necessary or indispensable. Indeed, the novel methods and systems described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of the methods and systems described herein may be made without departing from the spirit of the embodiments disclosed herein.
While certain embodiments of the disclosure have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the disclosure. Indeed, the novel methods, devices and systems described herein may be embodied in a variety of other forms. Furthermore, various omissions, substitutions and changes in the form of the methods and systems described herein may be made without departing from the spirit of the disclosure. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the disclosure. For example, those skilled in the art will appreciate that in various embodiments, the actual physical and logical structures may differ from those shown in the figures. Depending on the embodiment, certain steps described in the example above may be removed, others may be added. Also, the features and attributes of the specific embodiments disclosed above may be combined in different ways to form additional embodiments, all of which fall within the scope of the present disclosure. Although the present disclosure provides certain preferred embodiments and applications, other embodiments that are apparent to those of ordinary skill in the art, including embodiments which do not provide all of the features and advantages set forth herein, are also within the scope of this disclosure. Accordingly, the scope of the present disclosure is intended to be defined only by reference to the appended claims.
