Ex Parte LinDownload PDFPatent Trial and Appeal BoardMay 13, 201412113947 (P.T.A.B. May. 13, 2014) Copy Citation UNITED STATES PATENT AND TRADEMARK OFFICE 1 ___________ 2 3 BEFORE THE PATENT TRIAL AND APPEAL BOARD 4 ___________ 5 6 Ex parte KUNHUA LIN 7 ___________ 8 9 Appeal 2011-009415 10 Application 12/113,947 11 Technology Center 2400 12 ___________ 13 14 15 Before MURRIEL E. CRAWFORD, HUBERT C. LORIN, and 16 ANTON W. FETTING, Administrative Patent Judges. 17 18 FETTING, Administrative Patent Judge. 19 20 21 DECISION ON APPEAL 22 Appeal 2011-009415 Application 12/113,947 2 STATEMENT OF THE CASE1 1 1 Our decision will make reference to the Appellant’s Appeal Brief (“App. Br.,” filed November 23, 2010) and the Examiner’s Answer (“Ans.,” mailed February 22, 2011). Kunhua Lin (Appellant) seeks review under 35 U.S.C. § 134 of a final 2 rejection of claims 1 and 3-8, the only claims pending in the application on 3 appeal. We have jurisdiction over the appeal pursuant to 35 U.S.C. § 6(b). 4 The Appellant invented to a way to integrate global intelligence 5 regarding email messages and senders into the email delivery network to 6 allow more accurate local spam identification to be performed (Specification 7 para. [0003]). 8 An understanding of the invention can be derived from a reading of 9 exemplary claim 1, which is reproduced below [bracketed matter and some 10 paragraphing added]. 11 1. A spam detection system comprising: 12 [1] a global intelligence network including one or more global 13 intelligence servers coupled to a network and configured to 14 directly or indirectly 15 (i) gather intelligence from a plurality of distributed anti-16 spam engines, 17 (ii) maintain and update electronic mail (email) message 18 signatures and associated spam score information, 19 (iii) maintain and update sender Internet Protocol (IP) 20 reputation information 21 and 22 (iv) readjust spam detection characteristics of the 23 plurality of distributed anti-spam engines 24 based on observations and analysis performed by 25 the global intelligence network, 26 wherein the spam detection characteristics are readjusted 27 at least in part 28 Appeal 2011-009415 Application 12/113,947 3 based on heuristic rules provided to the plurality of 1 distributed anti-spam engines by a heuristic rule 2 update server of the global intelligence network 3 and 4 wherein the heuristic rules are generated 5 responsive to spam trends observed by the global 6 intelligence network 7 based on 8 direct or indirect observations 9 and 10 analysis of 11 query volume 12 or 13 query patterns 14 regarding e-mail message signatures 15 received from the plurality of anti-16 spam engines; 17 and 18 [2] a network device coupled to the network and through which 19 e-mail messages pass, 20 the network device including an anti-spam engine of the 21 plurality of distributed anti-spam engines, 22 the anti-spam engine configured to 23 (i) consider spam score information and sender 24 Internet Protocol (IP) reputation information of the 25 e-mail messages, 26 including 27 querying the global intelligence network for 28 the spam score information associated with 29 the e-mail messages 30 and 31 querying the global intelligence network for 32 the sender IP reputation information 33 associated with the e-mail messages, 34 (ii) perform content analysis on the e-mail 35 messages using the heuristic rules 36 and 37 (iii) provide the global intelligence network with 38 an opportunity 39 Appeal 2011-009415 Application 12/113,947 4 to gather further information 1 to make the content analysis more accurate 2 by queuing e-mail messages for which a 3 satisfactory spam or clean categorization 4 cannot be made in real-time 5 for 6 subsequent rechecking of 7 the spam score information 8 and 9 the sender IP reputation 10 information 11 and 12 reapplication of the content analysis. 13 14 The Examiner relies upon the following prior art: 15 Goodman US 2004/0215977 A1 Oct. 28, 2004 Rajan US 2006/0149821 A1 Jul. 6, 2006 16 Claims 1 and 4-8 stand rejected under 35 U.S.C. § 102(b) as anticipated 17 by Goodman. 18 Claim 3 stands rejected under 35 U.S.C. § 103(a) as unpatentable over 19 Goodman and Rajan. 20 ISSUES 21 The issues of anticipation and obviousness turn primarily on whether 22 the recited “global” modifier patentably distinguishes the claims from 23 Goodman. 24 FACTS PERTINENT TO THE ISSUES 25 The following enumerated Findings of Fact (FF) are believed to be 26 supported by a preponderance of the evidence. 27 28 Appeal 2011-009415 Application 12/113,947 5 Facts Related to Claim Construction 1 01. The disclosure contains no lexicographic definition of “global.” 2 02. The ordinary definition of global may be either worldwide or 3 comprehensive. In computer science, the term global generally 4 means of or relating to an entire program, document, or file.2 5 Facts Related to the Prior Art 6 Goodman 7 03. Goodman is directed to identifying both legitimate (e.g., good 8 mail) and undesired information (e.g., junk mail), and more 9 particularly to classifying messages for spam prevention in part by 10 at least delaying delivery of suspicious messages until further 11 information can be gathered about the messages to facilitate 12 classification of such messages. Goodman para. [0002]. 13 04. Goodman describes an intelligent quarantining system that 14 facilitates classifying items in connection with spam prevention, 15 and classifies or flags messages as suspicious and/or temporarily 16 delays their classification (as either spam or good). A delay or 17 quarantine period can be set by the filter and/or by the system 18 which provides a suitable amount of time to learn more 19 information about the message(s) and/or about the sender. 20 Goodman para. [0011]. 21 05. System components can monitor activities and/or behavior such 22 as message volume (e.g., message volume per sender). Message 23 content can be analyzed to determine whether it substantially 24 2 https://education.yahoo.com/reference/dictionary/entry/global (Houghton Mifflin) Appeal 2011-009415 Application 12/113,947 6 resembles a message found in a honeypot (a honeypot refers to a 1 known spam target to identify incoming messages as spam and/or 2 to track specific merchant message address processing). For 3 instance, the e-mail address can be disclosed on a website in a 4 restrictive manner not likely to be found by people. Hence, any 5 messages sent to this address can be considered spam. 6 Alternatively, the e-mail address may have only been disclosed to 7 a merchant from whom legitimate messages are expected to be 8 received. Thus, messages received from the merchant are 9 legitimate, but all other messages received can safely be 10 considered spam. Spam data derived from honeypots and/or other 11 sources (e.g., users) can be integrated into the feedback loop 12 system, but because of the substantial increase in spam 13 classification with honeypots, such data can be down weighted to 14 mitigate obtaining biased feedback results. Goodman para. 15 [0012]-[0013]. 16 06. Message content can be analyzed to determine whether it 17 substantially resembles messages that have received feedback 18 through other methods, including: being marked junk or not junk; 19 being categorized by a Feedback Loop user; being categorized by 20 a deployment of the Feedback Loop technology in some other 21 setting; or by comparing it to other spam repositories. 22 Quarantining can be combined with hash-based techniques. 23 Goodman para. [0014]-[0015]. 24 07. Quarantined messages can be stored in a special folder that may 25 be either visible or invisible to the user. Messages sent to the 26 Appeal 2011-009415 Application 12/113,947 7 quarantine folder may be selected for the Feedback Loop, whether 1 or not the quarantined messages are normally visible to the user. 2 That is, just like messages that are deleted, put in the junk folder, 3 or put in the inbox, messages sent to the quarantine folder may be 4 selected for user classification. The Feedback Loop is a polling 5 mechanism that involves asking users to classify at least a subset 6 of messages as spam or good to facilitate detecting spam and 7 building more robust spam filters. Goodman para. [0016]. 8 08. As an alternative, or in addition to the Feedback Loop, the 9 quarantine folder can be visible to message recipients to provide 10 them an opportunity to classify at least a subset of the messages 11 held in the special folder. They may be able to report such 12 messages as good or as junk. Thus, user data, either through the 13 Feedback Loop or junk/good reporting methods, with respect to at 14 least a limited selection of quarantined messages can facilitate 15 determining whether a particular quarantined message is spam. 16 Goodman para. [0017]. 17 09. Goodman delays classification (as spam or otherwise) and 18 allows some users to provide their opinions about particular 19 messages to facilitate subsequent classification. Moreover, user 20 complaints such as those submitted by feedback loop participants 21 and/or unsolicited message recipients can be utilized to facilitate 22 determining whether at least some of the messages under 23 quarantine are spam. The lack of complaints from users can also 24 be noted and employed to assist in determining whether particular 25 messages are spam. Goodman para. [0018]. 26 Appeal 2011-009415 Application 12/113,947 8 10. Machine learning systems (e.g., neural networks, Support 1 Vector Machines (SVMs), Bayesian Belief Networks) facilitate 2 creating improved and/or updated spam filters that are trained to 3 recognize both legitimate and spam messages and further, to 4 distinguish between them. Once a new or updated spam filter has 5 been trained in accordance with the invention, it can be distributed 6 to mail servers and client e-mail software programs. Furthermore, 7 the new or updated spam filter can be trained with respect to 8 classifications and/or other information provided by a particular 9 user(s) to improve performance of a personalized filter(s). As 10 additional training data sets are built, the spam filter can undergo 11 further training via machine learning to optimize its performance 12 and accuracy. User feedback by way of message classification 13 can also be utilized to generate lists for spam filters and parental 14 controls, to test spam filter performance, and/or to identify spam 15 origination. Goodman para. [0020]. 16 11. Information gathered from user and/or system feedback can be 17 employed to update the one or more filters already in use. As a 18 result, the delayed messages can be processed or sent through the 19 filter(s) again for classification. In addition, new filters can be 20 trained for application to subsequent incoming messages subjected 21 to quarantine. Goodman para. [0023]. 22 12. One or more components may reside within a process and/or 23 thread of execution and a component may be localized on one 24 computer and/or distributed between two or more computers. 25 Goodman can incorporate various inference schemes and/or 26 Appeal 2011-009415 Application 12/113,947 9 techniques in connection with generating training data for 1 machine learned spam filtering. As used by Goodman, the term 2 "inference" refers generally to the process of reasoning about or 3 inferring states of the system, environment, and/or user from a set 4 of observations as captured via events and/or data. Inference can 5 be employed to identify a specific context or action, or can 6 generate a probability distribution over states, for example. The 7 inference can be probabilistic--that is, the computation of a 8 probability distribution over states of interest based on a 9 consideration of data and events. Inference can also refer to 10 techniques employed for composing higher-level events from a set 11 of events and/or data. Such inference results in the construction of 12 new events or actions from a set of observed events and/or stored 13 event data, whether or not the events are correlated in close 14 temporal proximity, and whether the events and data come from 15 one or several event and data sources. Goodman para. [0038]-16 [0039]. 17 13. Referring to FIG. 1, there is illustrated a general block diagram 18 of a quarantining system that implements a feedback loop system. 19 A message receipt component receives and delivers incoming 20 messages (denoted as IM) to intended recipients. The message 21 receipt component can include or can operate together with at 22 least one filter (e.g., first classification component), as is 23 customary with many message receipt components to mitigate 24 delivery of undesirable messages (e.g., spam). The message 25 receipt component, in connection with the filter, processes the 26 Appeal 2011-009415 Application 12/113,947 10 messages (IM) and provides a filtered subset of the messages 1 (denoted as FILTERED IM) to the intended recipients. The 2 filter(s) may have been trained using a feedback loop system. In 3 particular, the filter(s) are previously trained to identify not only 4 spam, but also to distinguish between spam and good mail based 5 at least in part upon trusted user feedback. Machine learning 6 systems facilitate the training of such filters by utilizing training 7 data comprising user feedback regarding both good and spam 8 messages. Goodman para. [0043]-[0044]. 9 14. Unlike conventional spam prevention systems, messages that 10 lack information for classification can be held back or quarantined 11 (flagged for further analysis)--while more information is collected 12 about them. The quarantined messages can be moved to a delayed 13 message store for a period of time (e.g., delay or quarantine 14 period) until the filters can be updated with any information 15 collected during the quarantine period. The delayed message store 16 may be the same as some other store, e.g., the junk folder, or 17 queues on a server. Quarantined messages in this store may be 18 specially marked, or all messages in this folder or queue may be 19 periodically rescored as if they were quarantined. There are 20 several types of information that can be obtained. One type is a 21 trickle out component. User feedback on quarantined messages 22 may involve employing a trickle out mechanism in which at least 23 a subset of quarantined messages is allowed to "trickle out" of 24 quarantine or bypass the filter classification process for delivery to 25 their intended recipients. Messages which are trickled out may be 26 Appeal 2011-009415 Application 12/113,947 11 selected based in part on the fact that the intended recipient (e.g., 1 random or selected user) is a participant in the feedback loop 2 system for training spam filters. Alternatively, or in addition, the 3 trickled out messages can be randomly selected. Goodman para. 4 [0047]-[0048]. 5 15. In addition to user feedback, information can also be gathered 6 by a message analysis component that is operatively connected to 7 the delayed message store. The message analysis component can 8 monitor quarantined messages with respect to volume per sender 9 and similarities among quarantined messages and can analyze 10 them as well for their content and/or origination information. For 11 instance, messages sent in low volume are less likely to be spam 12 than messages sent in high volume, which is more representative 13 of spammer behavior. Thus, information that a particular sender 14 is sending a low volume of messages can be a feature learned 15 about the sender and used to update the filters so that in the future, 16 the sender's messages may not deemed to be suspicious, but rather 17 may be classified as good. Additionally, a hash function can be 18 computed with respect to at least a subset of quarantined messages 19 to determine similarity among the messages such as per sender. 20 For instance, messages in quarantine can be compared to other 21 recent messages based on content or based on sender. If other 22 recent messages with the same or a similar hash or from the same 23 or a similar user were reported as junk by users, classified as junk 24 in the feedback loop, or arrived in honeypots, the message can be 25 classified as spam. If similar messages were marked as good, or 26 Appeal 2011-009415 Application 12/113,947 12 rescued from a quarantine or junk folder, or classified as good in 1 the feedback loop, the message can be classified as good. If many 2 similar messages reached the message receipt component, then the 3 volume of such messages can be an indicator that the messages are 4 spam. If many similar messages were delivered to user’s inboxes 5 (e.g. through trickle out), and none or few were marked as junk by 6 users, this can be taken as an indicator that the messages were 7 good. If no similar messages arrived in honeypots, this can be 8 taken as an indicator that the message is good. Goodman para. 9 [0053]-[0054]. 10 16. System feedback on the quarantined messages can also be 11 collected. This can include data collected from monitoring at least 12 a subset of messages in the quarantine folder for characteristics 13 such as volume (low or high volume of message), similarity of 14 message to other quarantined messages, and/or similarity of 15 message to honeypot message. This information together with 16 any available user feedback can be utilized by a filter update 17 component as respective features (or training data) to train and 18 update the filter(s). Following therefrom, updated filters can be 19 generated. The quarantined messages can be, in substantial part, 20 run through the updated filters to resume the classification 21 process. Hence, once classified, designated spam can be 22 permanently deleted from the quarantine folder or sent to a trash 23 bin for deletion. A first delivery component can release the 24 “good” messages from quarantine for delivery to their intended 25 recipients. If the quarantine folder is not also the junk folder, 26 Appeal 2011-009415 Application 12/113,947 13 quarantined messages can be placed in the junk folder by way of a 1 second delivery component. If the quarantine folder is the junk 2 folder, quarantined messages may have a special “quarantined” 3 flag removed. FIG. 3 schematically illustrates one particular 4 course a new incoming message may take. The message is 5 initially processed through a filter. The filter has been trained to 6 distinguish between good and spam messages by computing 7 probability scores, for example. However, some messages may 8 fall on the edge of being decisively classified as spam or good. 9 These can be suspicious messages. One reason for this is that the 10 filter may lack information about the message merely because the 11 message includes aspects or features it has not seen before or does 12 not recognize. For these particular messages, classification as 13 spam or good is deferred for a time period. This time period 14 allows the filter to learn more information about the message 15 before committing to a spam or good classification. As a result, 16 classification error rates can be reduced and user satisfaction can 17 be increased since never-seen-before “good” messages are not 18 arbitrarily classified as “spam” simply due to an ignorant filter or 19 a filter lacking the appropriate information. Goodman para. 20 [0060]-[0061]. 21 17. When the IP address is previously unseen, it typically can be 22 quarantined--unless the message is obviously spam or good based 23 on other features of the message. If a message having an IP 24 address that has not been seen before is received, there are three 25 possibilities: 26 Appeal 2011-009415 Application 12/113,947 14 it is a low volume IP address (e.g., perhaps the server for a 1 small business or an individual and it is not spam--or at the 2 least, it is very targeted spam); 3 it is a new IP address perhaps for a large legitimate company as 4 they add more servers (e.g., The New York Times)--again not 5 spam; or 6 it is a spammer's IP address. 7 By waiting even a few hours, the filter can probably distinguish 8 between these three possibilities and obtain very valuable 9 information. For an unknown IP address, it may be desirable to 10 delay the message (quarantine temporarily) even if the message 11 falls in a pretty wide range. The sender's domain can be handled in 12 a similar manner. As anti-spoofing technology becomes more 13 prevalent, messages can be quarantined as well to ascertain true 14 sender's of certain questionable messages. In addition, if there is 15 no reverse IP address entry for the sender's IP address and/or a 16 forward lookup on the sender's domain does not at least 17 approximately match the sender's IP address, the message can be 18 quarantined. Another substantial indicator of spam is the presence 19 of embedded domains, especially in the links. If a message 20 contains a domain name that is never or rarely seen before, it can 21 be deemed suspicious. Just as with IP addresses, delaying delivery 22 of such messages can be helpful to properly classify them as spam 23 or good. Certain types of attached files are particularly suspicious 24 (e.g., typical of viruses) and messages containing such extensions 25 Appeal 2011-009415 Application 12/113,947 15 (e.g., executable files or document files with embedded macros) 1 can be quarantined. Goodman para. [0067]-[0072]. 2 18. Quarantining may also be able to detect attempts to use holes in 3 keyword-based filtering. For instance, if a spammer discovers 4 many good words and adds these good words to his message, but 5 a few bad words are still detectable, the message can be viewed as 6 suspicious (even though it has a good overall score). The message 7 can be held back from classification for a few hours, for example, 8 and through the feedback loop system, many messages of this 9 kind can be discovered. After which, the filter can learn that a 10 message of this type is actually bad. To combat this type of 11 spammer tactic, words that are previously thought to be good can 12 be down weighted and the filter can learn that the origin of the 13 message is bad, etc. In other words, when a message appears to 14 be of a type that is difficult for a machine learning filter because it 15 includes conflicting evidence, it can be quarantined. Additionally, 16 any message that appears to be difficult for any kind of filter 17 because it includes HTML, which can be difficult to parse, or 18 includes primarily an image can be quarantined. Overall, an 19 assessment of a plurality of features can be performed before it 20 can be determined whether to quarantine a message. Goodman 21 para. [0073]. 22 19. In general, individual users tend to receive messages from a 23 relatively small number of locations and in a small number of 24 languages. With respect to personal filters, when a user receives a 25 message from a location they do not typically receive good 26 Appeal 2011-009415 Application 12/113,947 16 messages from or in a language they do not typically receive good 1 messages in, the message can be quarantined as well. The 2 location can be determined based in part on geographic location, 3 IP address, IP address allocation information, country code in 4 FROM domain name, and the like. Goodman para. [0074]. 5 20. FIG. 7 illustrates an exemplary process that facilitates delaying 6 classification of suspicious or questionable messages. The 7 process involves receiving incoming messages for classification as 8 spam or good. Good messages can be delivered and spam 9 messages can be discarded. Messages from senders on safe-lists 10 are not quarantined, though they may otherwise appear suspicious. 11 Users can add senders to their respective safe lists if they notice 12 that particular messages are consistently under quarantine (per 13 sender). Goodman para. [0081]. 14 21. FIG. 7 illustrates an exemplary process that facilitates delaying 15 classification of suspicious or questionable messages. The 16 process involves receiving incoming messages for classification as 17 spam or good. Good messages can be delivered and spam 18 messages can be discarded. Messages from senders on safe-lists 19 are not quarantined though they may otherwise appear suspicious. 20 Users can add senders to their respective safe lists if they notice 21 that particular messages are consistently under quarantine (per 22 sender). Goodman para. [0081]. 23 22. Quarantined messages can be analyzed for their similarity. In 24 particular, hash values can be computed for the messages to 25 determine which messages are similar to each other among 26 Appeal 2011-009415 Application 12/113,947 17 different senders or per sender. High volumes of similar messages 1 can indicate spam and this information can be used to update the 2 filter. In addition, quarantined messages can be compared to 3 recently quarantined messages that have been classified as spam 4 or good. When similar messages are found, they can be removed 5 from quarantine and classified as their earlier counterparts were. 6 In addition, messages can be compared based on sender analysis 7 (e.g., sender IP address). As some messages come in, special 8 queries are sent to the recipients specifically asking them to 9 categorize the messages as good or spam. FIG. 9 depicts how 10 information collected during a quarantine period can be utilized to 11 improve classification of messages. The information can be 12 employed as training data in conjunction with machine learning 13 techniques to effectively update a spam filter, for example. By 14 updating the spam filter, classification of messages as spam or 15 good can be improved to mitigate false-good or false-spam 16 classifications. Alternatively, or in addition, at least a subset of 17 the information obtained can be employed to build or train a new 18 spam filter (sub-filter) for recently quarantined messages. 19 Goodman para. [0087]-[0089]. 20 ANALYSIS 21 Claim 1 recites a structural claim, viz. a system, with a network and a 22 network device. The network has servers that gather intelligence from plural 23 anti-spam engines, maintain email message signatures and spam scores and 24 sender IP reputation information, and readjust the spam detection 25 Appeal 2011-009415 Application 12/113,947 18 characteristics. The network also has a heuristic rule update server. The 1 network device has an anti-spam engine that considers the spam score and IP 2 reputation information, performs email content analysis using heuristic rules, 3 and provides some form of opportunity to gather more information. 4 The Examiner found that Goodman anticipates such a structure and 5 the contention does not reach the above recited limitations. 6 The principal argument is that Goodman has no global intelligence 7 network. We are not persuaded by the Appellant’s argument that “it should 8 be clear that ‘global’ is used in its traditional sense to mean intelligence is 9 gathered from around the world – the network is not limited or 10 provincial in scope.” Br. 11. The intelligence gathered by the recited 11 system is not a structural component and is not afforded patentable weight. 12 “[E]xpressions relating the apparatus to contents thereof during an 13 intended operation are of no significance in determining patentability of the 14 apparatus claim.” Ex parte Thibault, 164 USPQ 666, 667 (Bd. App. 1969). 15 Furthermore, “inclusion of material or article worked upon by a structure 16 being claimed does not impart patentability to the claims.” In re Otto, 312 17 F.2d 937, 940 (CCPA 1963). 18 In any event, the word global in the phrase “global intelligence 19 network” is an adjective modifying the noun “network.” The internet is a 20 global network. Were “global” construed to modify “intelligence” instead, 21 the phrase “global intelligence” would impose no structural limitation on the 22 recited network, but would merely be an aspirational field of use. 23 For similar reasons, we are not persuaded by the Appellant’s 24 argument that Goodman does not teach or suggest a participating anti-spam 25 engine of a plurality of distributed anti-spam engines should query a global 26 Appeal 2011-009415 Application 12/113,947 19 intelligence network for spam score information or for sender IP reputation 1 information. Br. 13. Goodman describes querying for spam scores and IP 2 reputation information. The scope of the query is given no patentable 3 weight, as no structural components affecting such scope are recited. 4 We are not persuaded by the Appellant’s argument that 5 Goodman's indication that its “message analysis component 190 6 can monitor quarantined messages with respect to volume per 7 sender” (emphasis added) has nothing to do with generating 8 heuristic rules “responsive to ... analysis of query volume or 9 query patterns regarding e-mail message signatures received 10 from the plurality of anti-spam engines." 11 . . .Goodman's suggestion that the limited scope 12 “intelligent quarantining system” can learn based on certain 13 characteristics of quarantined messages is inapposite as the 14 analysis of quarantined messages is not comparable to analysis 15 of queries 16 Br. 15. 17 Goodman applies machine learning to train its filters using email volume. 18 Such training is a heuristics generation. The contention is that Goodman’s 19 subject of the analysis differs from the recited limitations. The limitation at 20 issue is 21 the heuristic rules are generated responsive to spam trends 22 observed by the global intelligence network based on direct or 23 indirect observations and analysis of query volume or query 24 patterns regarding e-mail message signatures received from the 25 plurality of anti-spam engines. 26 Claim 1. 27 The claims do not recite who or what generates the heuristic rules, or 28 when this is done. Generating such rules offline and loading them into 29 Goodman is within the scope of the claim. As this limitation describes 30 neither a structural limitation, nor a function that the structure necessarily 31 Appeal 2011-009415 Application 12/113,947 20 performs, the limitation is deserving of no patentable weight. “[A]pparatus 1 claims cover what a device is, not what a device does.” Hewlett-Packard 2 Co. v. Bausch & Lomb Inc., 909 F.2d 1464, 1468 (Fed. Cir. 1990). 3 Even were patentable weight afforded, the nature of the analysis and 4 the manner in which the generation of heuristics is based on such analysis is 5 not further narrowed or specified. The nature or manner of such queries is 6 also not further narrowed or specified. Goodman quarantines the emails that 7 are most likely problems, and runs statistical checks on them. Goodman 8 queries users based on feedback loops regarding these quarantined 9 messages. Thus, the data Goodman derives from such feedback are based 10 on query volume and patterns. 11 CONCLUSIONS OF LAW 12 The rejection of claims 1 and 4-8 under 35 U.S.C. § 102(b) as 13 anticipated by Goodman is proper. 14 The rejection of claim 3 under 35 U.S.C. § 103(a) as unpatentable 15 over Goodman and Rajan is proper. 16 DECISION 17 The rejection of claims 1 and 3-8 is affirmed. 18 No time period for taking any subsequent action in connection with 19 this appeal may be extended under 37 C.F.R. § 1.136(a). See 37 C.F.R. 20 § 1.136(a)(1)(iv) (2011). 21 22 AFFIRMED 23 Klh 24 Copy with citationCopy as parenthetical citation