Message profiling systems and methods

Description of equivalent WO 2006052736 A2

MESSAGE PROFILING SYSTEMS AND METHODS

0002BACKGROUND AND SUMMARY This document relates generally to systems and methods for processing communications and more particularly to systems and methods for filtering communications.

0003In the anti-spam industry, spammers use various creative means for evading detection by spam filters. Available anti-spam systems include fail-open systems in which all incoming messages are filtered for spam. However, these systems can be inefficient and inaccurate in properly classifying messages as legitimate or spam. In accordance with the teachings disclosed herein, methods and systems are provided for operation upon one or more data processors that assign a reputation to a messaging entity. For example, a method and system can include receiving data that identifies one or more characteristics related to a messaging entity's communication; determining a reputation score based upon the received identification data; wherein the determined reputation score is indicative of reputation of the messaging entity; wherein the determined reputation score is used in deciding what action is to be taken with respect to a communication associated with the messaging entity. As another example, systems and methods of performing transmission filtering utilizing reputation scores of transmission sender are provided. A system and method can include: identifying at least one characteristic about a transmission from a sender; performing a real-time query to the reputation system that includes the transmission characteristic; receiving a score representing reputation related to the transmission; and, performing an action on the transmission from the sender corresponding to the score range of the sender's reputation.

0004As another example, systems and methods for performing filtering of groups of transmissions utilizing reputation scores of senders of transmissions are provided.

0005For example, a system and method can include: grouping multiple transmissions together based on content similarities or similarities in transmission sender behavior; identifying at least one characteristic about each transmission in the groupings; performing a query to the reputation system and receiving a score representing reputation of each sender; and, classifying groups of transmissions based on the percentage of reputable and non-reputable senders in the group. As another example, systems and methods for performing tuning and training of filtering systems utilizing reputation scores of senders of transmissions in sets of trainable transmissions are provided. For example, the method can include: identifying at least one characteristic about transmissions from senders; performing queries to a reputation system and receiving scores representing reputations of the senders; classifying transmissions into multiple categories based on a range a sender's reputation score falls into; and, passing on transmissions and their classification categories to a trainer of another filtering system to be used for optimization of the filtering system.

0006As another example, systems and methods for operation upon one or more data processors to classify communications from messaging entities are provided. For example, a system and method can include: receiving a communication from a messaging entity; using a plurality of message classification techniques to classify the communication; and, combining the message classification outputs in order to generate a message profile score; wherein the message profile score is used in deciding what action is to be taken with respect to the communication associated with the messaging entity.

0007As another example, such a system and method can include: identifying at least one characteristic about transmissions from senders; performing queries to a reputation system and receiving scores representing reputations of the senders; classifying transmissions into multiple categories based on a range a sender's reputation score falls into; and, passing on transmissions and their classification categories to a trainer of another filtering system to be used for optimization of the filtering system.

0008As another example, such a system and method can include: receiving a communication from a messaging entity; using a plurality of message classification techniques to classify the communication; and, combining the message classification outputs in order to generate a message profile score; wherein the message profile score is used in deciding what action is to be taken with respect to the communication associated with the messaging entity. In accordance with the teachings disclosed herein, methods and systems are provided for operation upon one or more data processors that classify communications from messaging entities. For example, a system and method can include a plurality of message classification techniques, wherein the techniques are configured to classify a communication received from a messaging entity. A system and method can further include message profiling logic configured to combine the message classification outputs in order to generate a message profile score, wherein the message profile score is used in deciding what action is to be taken with respect to the communication associated with the messaging entity.

0009As another example, a method and system can include receiving a communication that was sent from a messaging entity. A plurality of message classification techniques is used to classify the communication. A message classification technique is associated with a confidence value which is used in generating a message classification output from the message classification technique. The message classification outputs are combined in order to generate a message profile score. The message profile score is used in deciding what action is to be taken with respect to the communication associated with the messaging entity.

0010As another example, a system and method can utilize a plurality of message classification techniques, wherein the plurality of message classification techniques are configured to classify a communication received from a messaging entity. Message profiling logic can be configured to combine the message classification outputs in order to generate a message profile score. The message profile score is used in deciding what action is to be taken with respect to the communication associated with the messaging entity.

0011As another example, a system and method can be used for tuning message classification parameters for use by one or more message classification techniques. A plurality of input data is received (such as through input logic or processing instructions) that is or is representative of a plurality of communications. A tuner program is used to tune the message classification parameters associated with the message classification techniques. A communication is received from a messaging entity. The tuned message classification parameters are used by the plurality of message classification techniques to classify the communication. Message classification outputs from the plurality of message classification techniques are combined in order to generate a message profile score. The message profile score is ^' used in deciding what action is to be taken with respect to the communication associated with the messaging entity.

BRIEF DESCRIPTION OF THE DRAWINGS

0013FIG. 1 is a block diagram depicting a system for handling transmissions received over a network. FIG. 2 is a block diagram depicting a reputation system that has been configured for determining reputation scores.

0014FIG. 3 is a table depicting reputation scores at various calculated probability values.

0015FIG. 4 is a graph depicting reputation scores at various calculated probability values.

0016FIG. 5 is a flowchart depicting an operational scenario for generating reputation scores.

0017FIG. 6 is a block diagram depicting use of non-reputable criteria and reputable criteria for determining reputation scores. FIG. 7 is a block diagram depicting a reputation system configured to respond with a return value that includes the reputation score of a sender.

0018FIG. 8 is a block diagram depicting a system for handling transmissions received over a network.

0019FIG. 9 is a block diagram depicting a filtering system having a message profiler program. FIG. 10 is a block diagram depicting a message classification tuner program. FIG. 11 is a block diagram depicting use of a genetic algorithm as a message classification tuner program.

0020FIG. 12 is a flowchart depicting an operational scenario wherein a message profiler is used.

0021FIG. 13 is a block diagram depicting a message profiler that has been adapted to operate with an adaptive message blocking and whitelisting.

0022FIG. 14 is a block diagram depicting a reputation system for handling transmissions received over a network. FIG. 15 is a block diagram depicting a reputation system that has been configured for determining reputation scores.

0023FIG. 16 is a table depicting reputation scores at various calculated probability values.

0024FIG. 17 is a block diagram depicting a server access architecture.

DETAILED DESCRIPTION

0026FIG. 1 depicts at 30 a system for handling transmissions received over a network 40. The transmissions can be many different types of communications, such as electronic mail (e-mail) messages sent from one or more messaging entities 50. The system 30 assigns a classification to a messaging entity (e.g., messaging entity 52), and based upon the classification assigned to the messaging entity, an action is taken with respect to the messaging entity's communication.

0027The system 30 uses a filtering system 60 and a reputation system 70 to help process communications from the messaging entities 50. The filtering system 60 uses the reputation system 70 to help determine what filtering action (if any) should be taken upon the messaging entities' communications. For example, the communication may be determined to be from a reputable source and thus the communication should not be filtered.

0028The filtering system 60 identifies at 62 one or more message characteristics associated with a received communication and provides that identification information to the reputation system 70. The reputation system 70 evaluates the reputation by calculating probabilities that the identified message characteristic(s) exhibit certain qualities. An overall reputation score is determined based upon the calculated probabilities and is provided to the filtering system 60. The filtering system 60 examines at 64 the reputation score in order to determine what action should be taken for the sender's communication (such as whether the communication transmission should be delivered to the communication's designated recipient located within a message receiving system 80). The filtering system 60 could decide that a communication should be handled differently based in whole or in part upon the reputation scored that was provided by the reputation system 70. As an illustration, a communication may be determined to be from a non- reputable sender and thus the communication should be handled as Spam (e.g., deleted, quarantined, etc.).

0029Reputation systems may be configured in many different ways in order to assist a filtering system. For example, a reputation system 70 can be located externally or internally relative to the filtering system 60 depending upon the situation at hand. As another example, FIG. 2 depicts a reputation system 70 that has been configured to calculate reputation scores based upon such message characteristic identification information as sender identity as shown at 82. It should be understood that other message characteristics can be used instead of or in addition to sender identity. Moreover, transmissions may be from many different types of messaging entities, such as a domain name, IP address, phone number, or individual electronic address or username representing an organization, computer, or individual user that transmits electronic messages. For example, generated classifications of reputable and non-reputable can be based upon a tendency for an IP address to send unwanted transmissions or legitimate communication.

0030The system's configuration 90 could also, as shown in FIG. 2, be established by identifying a set of binary, testable criteria 92 which appear to be strong discriminators between good and bad senders. P (NR\Ci) can be defined as the probability that a sender is non-reputable, given that it conforms to quality/criterion C and P (R\Ci) can be defined as the probability that a sender is reputable, given that it conforms to quality/criterion Q.

0031For each quality/criterion C₁, periodic (e.g., daily, weekly, monthly, etc.) sampling exercises can be performed to recalculate P (NR\Ci). A sampling exercise may include selecting a random sample set S of N senders for which quality/criterion Ci is known to be true. The senders in the sample are then sorted into one of the following sets: reputable (R), non-reputable (NR) or unknown (U). N_R is the number of senders in the sample that are reputable senders, Nm is the number of senders that are non-reputable senders, etc. Then, P (NR\Cϊ) and P (R|G) are estimated using the formulas:

0032P(NR I G) = ^-

0033¹ N

0034P(R I G) = ^

0035N For this purpose, N = 30 was determined to be a large enough sample size to achieve an accurate estimate of P (NR|C,) and P (RlQ) for each quality/criterion C₁-.

0036After calculating P (NR[Ci) and P (R\Q) for all criteria, the computed probabilities are used to calculate an aggregate non-reputable probability 94, P_NR, and an aggregate reputable sender probability 96, P_R, for each sender hi the reputation space. These probabilities can be calculated using the formulas:

0037, (#o/ criteria that apply) a apopmliieess <img file="WO2006052736A2_D0001.tif" />

0038^PR <img file="WO2006052736A2_D0002.tif" /> In experimentation, the above formulas appeared to behave very well for a wide range of input criteria combinations, and in practice their behavior appears to be similar to the behavior of the formula for correctly computing naive joint conditional probabilities of "non-reputable" and "reputable" behavior for the input criteria. After calculating Pm and PR for each sender, a reputation score is calculated for that sender using the following reputation function:

0039<img file="WO2006052736A2_D0003.tif" />^{+ C}5^NR

0040+ c₅P_R³ + c₆P_mP* + c₆P^P_R )((P_NR - P_R )³ + C₇ (P_NR - P_R ))

0041where

0042C₁ = 86.50 C₂ = -193.45 C₃ = -35.19 c₄ = 581.09 C₅ = 234.81 C₆ = -233.18 C₇ = 0.51

0043It should be understood that different functions can act as a reputation score determinator 98 and can be expressed in many different forms in addition to a functional expression. As an illustration, FIG. 3 depicts at 100 a tabular form for determining reputation scores. The table shows reputation scores produced by the above function, based on values of P^_R and PR as they each vary between 0.0 and 1.0. For example as shown at 110, a reputation score of 53 is obtained for the combination of P_m = 0.9 and P_R = 0.2. This reputation score is a relatively high indicator that the sender should not be considered reputable. A reputation score of 0 is obtained if P^ and Pi_? are the same (e.g., the reputation score is 0 if P_NR = 0.7 and P_R = 0.7 as shown at 120). A reputation score can have a negative value to indicate that a sender is relatively reputable as determined when Pi? is greater than P_NR- For example, if Pm = 0.5 and P_R = 0.8 as shown at 130, then the reputation score is -12. Reputation scores can be shown graphically as depicted in FIG. 4 at 150. Graph 150 was produced by the above function, based on values of Pm and PR. FIG. 4 illustrates reputation score determinations in the context of Spam in that the terms Pi_VΛ and P_R are used respectively as probability of hamminess and probability of spamminess as the probabilities each vary between 0.0 and 1.0.

0044As shown in these examples, reputation scores can be numeric reputations that are assigned to messaging entities based on characteristics of a communication (e.g., messaging entity characteristic(s)) and/or a messaging entity's behavior. Numeric reputations can fluctuate between a continuous spectrum of reputable and non- reputable classifications. However, reputations may be non-numeric, such as by having textual, or multiple level textual categories.

0045FIG. 5 depicts an operational scenario wherein a reputation system is used by a filtering system to generate reputation scores, hi this operational scenario, a reputation score is computed for a particular sender (e.g., IP address, domain name, phone number, address, name, etc), from a set of input data. With reference to FIG. 5, data is gathered at step 200 that is needed to calculate non-reputable and reputable probabilities for a sender. The data is then aggregated at step 210 and used in probability calculations at step 220. This includes determining, for a sender, non- reputable probabilities and reputable probabilities for various selected criteria. An aggregate non-reputable probability and an aggregate reputable probability are then calculated for each sender.

0046After calculating an aggregate non-reputable probability and an aggregate reputable probability for each sender, a reputation score is calculated at 230 for that sender using a reputation function. At step 240, the sender's reputation score is distributed locally and/or to one or more systems to evaluate a communication associated with the sender. As an illustration, reputation scores can be distributed to a filtering system. With the reputation score, the filtering system can choose to take an action on the transmission based on the range the sender reputation score falls into. For unreputable senders, a filtering system can choose to drop the transmission (e.g., silently), save it in a quarantine area, or flag the transmission as suspicious, hi addition, a filter system can choose to apply such actions to all future transmissions from this sender for a specified period of time, without requiring new lookup queries to be made to the reputation system. For reputable senders, a filtering system can similarly apply actions to the transmissions to allow them to bypass all or certain filtering techniques that cause significant processing, network, or storage overhead for the filtering system.

0047It should be understood that similar to the other processing flows described herein, the processing and the order of the processing may be altered, modified and/or augmented and still achieve the desired outcome. For example, an optional addition to the step of extracting unique identifying information about the sender of the transmission would be to use sender authentication techniques to authenticate certain parts of the transmission, such as the purported sending domain name in the header of the message, to unforgeable information about the sender, such as the IP address the transmission originated from. This process can allow the filtering system to perform lookups on the reputation system by querying for information that can potentially be forged, had it not been authenticated, such as a domain name or email address. If such domain or address has a positive reputation, the transmission can be delivered directly to the recipient system bypassing all or some filtering techniques. If it has a negative reputation, the filtering system can choose to drop the transmission, save it in a quarantine area, or flag it as suspicious.

0048Many different types of sender authentication techniques can be used, such as the Sender Policy Framework (SPF) technique. SPF is a protocol by which domain owners publish DNS records that indicate which IP addresses are allowed to send mail on behalf of a given domain. As other non-limiting examples, SenderID or DomainKeys can be used as sender authentication techniques.

0049As another example, many different types of criteria may be used in processing a sender's communication. FIG. 6 depicts the use of non-reputable criteria 300 and reputable criteria 310 for use in determining reputation scores.

0050The non-reputable criteria 300 and reputable criteria 310 help to distinguish non-reputable senders and reputable senders. A set of criteria can change often without significantly affecting the reputation scores produced using this scoring technique. As an illustration within the context of SPAM identification, the following is a list of spamminess criteria that could be used in the reputation scoring of a message sender. The list is not intended to be exhaustive, and can be adapted to include other criteria or remove criteria based upon observed behavior.

00511. Mean Spam Score: A sender is declared "non-reputable" if a mean spam profiler score of transmissions that it sends exceeds some threshold, W.

00522. RDNS Lookup Failure: A sender is declared "non-reputable" if reverse domain name system (RDNS) queries for its IP addresses fail. 3. RBL Membership: A sender is declared "non-reputable" if it is included in a real-time blackhole list (RBL). (Note: multiple RBLs may be used. Each RBL can constitute a separate testing criterion.)

00534. Mail Volume: A sender is declared "non-reputable" if its average (mean or median) transmission volume exceeds a threshold, X, where X is measured in transmissions over a period of time (such as, e.g., a day, week, or month). (Note: multiple average volumes over multiple time periods may be used, and each average volume can constitute a separate testing criterion.)

00545. Mail Burstiness/Sending History: A sender is declared "non-reputable" if its average (mean or median) transmission traffic pattern burstiness

0055(defined by the number of active sending sub-periods within a larger time period, e.g., number of active sending hours in a day or number of active sending days in a month) is less than some threshold, Y, where Y is measured in sub-periods per period. (Note: multiple average burstiness measures over multiple time periods may be used, and each average burstiness measure can constitute a separate testing criterion.)

00566. Mail Breadth: A sender is declared "non-reputable" if its average (mean or median) transmission traffic breadth (as defined by the percentage of systems that receive transmissions from the same sender during a period of time (such as, e.g., a day, week, or month)) exceeds some threshold, Z. (Note: multiple average breadths over multiple time periods may be used, and each average breadth measure can constitute a separate testing criterion.)

00577. Malware Activity: A sender is declared "non-reputable" if it is known to have delivered one or more malware codes (such as, e.g., viruses, spyware, intrusion code, etc) during a measurement period (e.g., a day, week, or month).

00588. Type of Address: A sender is declared "non-reputable" if it is known to be dynamically assigned to dial-up or broadband dynamic host control protocol (DHCP) clients by an internet service provider (ISP).

00599. CIDR Block Spamminess: A sender is declared "non-reputable" if its IP addresses are known to exist within classless inter-domain routing (CIDR) blocks that contain predominantly "non-reputable" IP addresses.

006010. Human Feedback: A sender is declared "non-reputable" if it is reported to have sent undesirable transmissions by people analyzing the content and other characteristics of those transmissions.

006111. SpamTrap Feedback: A sender is declared "non-reputable" if it is sending transmissions to accounts that have been declared as spamtraps and as such are not supposed to receive any legitimate transmissions. 12. Bounceback Feedback: A sender is declared "non-reputable" if it is sending bounceback transmissions or transmissions to accounts that do not exist on the destination system.

006213. Legislation/Standards Conformance: A sender is declared "non- reputable" if it is not conforming to laws, regulations, and well-established standards of transmission behavior in the countries of operation of either the sender and/or the recipient of the transmissions.

006314. Continuity of Operation: A sender is declared "non-reputable" if it has not operated at that sending location longer than some threshold Z.

006415. Responsiveness to Recipient Demands: A sender is declared "non- reputable" if it is not responding in a reasonable timeframe to legitimate demands of the recipients to terminate their relationship with the sender to not receive any more transmissions from them.

0065The following is a list of "reputable" criteria that could be used in determining the "reputability" of a sender. The list is not intended to be exhaustive, and can be adapted to include other criteria or remove criteria based upon observed behavior.

00661. Mean Spam Score: A sender is declared "reputable" if the mean spam profiler score of transmissions that it sends falls below some threshold, W.

00672. Human Feedback: A sender is declared "reputable" if it is reported to have sent only legitimate transmissions by people analyzing transmission flows from that sender, in conjunction with the reputation of the organization that owns those sending stations.

0068After computing a reputation grade for each sender in the universe of senders, a reputation classification can be made available via a communication protocol that can be interpreted by the queriers that make use of the reputation system (e.g., DNS, HTTP, etc). As shown in FIG. 7, when a query 350 is issued for a sender, the reputation system can respond with a return value 360 that includes the reputation score of that sender, as well as any other relevant additional information that can be used by the querier to make the final judgment on the acceptability of the sender' s transmission (e.g., age of the reputation score, input data that determined the score, etc).

0069An example of a communication protocol that can be used is a domain name system (DNS) server which can respond with a return value in the form of an IP address: Y12.x.y.z. The IP address can be encoded using the formula:

0070IP = 172. ^{rep ~}\^rep\ I (jrep|dzv256)(|rep|mod256)

0071The reputation of the queried sender can be deciphered from the return value as follows: rep = (-ϊ)²-^χ x(256y + z)

0072Therefore, when x = 0, the returned reputation is a positive number, and when x = l, the returned reputation is a negative number. The absolute value of the reputation is determined by the values of y and z. This encoding scheme enables the server to return via the DNS protocol reputation values within the range [-65535, 65535]. It also leaves seven (7) unused bits, namely the seven high-order bits of x. These bits can be reserved for extensions to the reputation system. (For example, the age of a reputation score may be communicated back to the querier.) FIG. 8 depicts at 430 a system for handling transmissions received over a network 440. The transmissions can be many different types of communications, such as electronic mail (e-mail) messages sent from one or more messaging entities 450. The system 430 uses a filtering system 460 to help process the communications from the messaging entities 450. The filtering system 460 examines characteristics associated with the communications from the messaging entities 450, and based upon the examination, an action is taken with respect to the communications. For example, a communication may be determined to be legitimate and thus the communication should not be filtered by the filtering system 460 and instead provided to a receiving system 70 for delivery to the intended recipient. To increase the accuracy of classifying messages properly (e.g., as spam or legitimate), a filtering system 460 can be configured with a message profiler program 500 as shown in FIG. 9. A message profiler 500 uses multiple message classification techniques or filters 510 to classify messages as shown in FIG. 9. Example message classification techniques or filters 510 that a message profiler 500 can use include: • Reverse DNS (RDNS) - a classification technique that performs a reverse domain name service (DNS) lookup, based on the message sender's IP address, to check (1) whether a domain exists in the DNS system for that IP address, and (2) if such a domain exists, whether the domain matches the domain from which the sender claims to be sending the message. • Real-time Black-hole Lists (RBLs) - a classification technique that performs one or more real-time black-hole list (RBL) queries, based on the message sender's IP address, to check whether the IP address has been identified by any RBLs as an IP address that is likely to send unwanted messages. • Reputation Server - a classification technique that performs one or more reputation server queries, based on the message sender's IP address and/or their domain name and other message sender characteristics, to receive a score that describes the sender's reputation.

0073• Signatur el fingerpήnting-based Analysis (e.g., Statistical Lookup Service (SLS)) - a classification technique that computes a set of message hashes and queries a centralized statistical lookup service (SLS) to determine how frequently the computed message hashes have been seen in recent mail flow.

0074• Message Header Analysis Classification Technique - as examples, this technique can include System Defined Header analysis (SDHA), User Defined Header Analysis (UDHA), etc.

0075• System Defined Header Analysis (SDHA) - a set of classification techniques that examine a message and identify whether the message's headers exhibit certain system-defined characteristics that tend to identify likely unwanted message senders. • User Defined Header Analysis (UDHA) - a set of classification techniques that examine a message and identify whether the message's headers exhibit certain user-defined characteristics that tend to identify likely unwanted message senders.

0076• Sender Authentication - a set of classification techniques that perform lookups to determine (1) whether the sender's claimed domain has published a record of mail servers that are authorized to send mail for that domain, and (2) if such a record has been published, whether the record authorizes the sender's IP address to send mail on behalf of the claimed domain. Examples of commonly used Sender Authentication techniques include Sender Policy Framework (SPF) and Sender ID.

0077• Bayesian Filtering - a statistical classification technique that computes an estimate of the joint conditional probability that a message falls into a specific category, based on the set of textual tokens (words) in the message.

0078• Content Filtering - a classification technique that searches the contents of a message for words that have been associated with certain message categories.

0079• Clustering Classification - a classification technique that based upon measuring similarity among features, communications are clustered into such groups as desired, undesired (e.g., spam), etc. The clustering is performed such that intra-group similarities are high and inter-group similarities are low.

0080The list is not intended to be exhaustive, and can be adapted to include other techniques as they are discovered. Some of the descriptions in the list constitute a single technique, while others constitute a combined set of many similar or closely related techniques, hi cases where multiple techniques are described collectively, the message profiler 500 permits each technique to have its own confidence value.

0081A message profiler 500 classifies messages using a threshold-based technique. Each of the classification techniques 510 used by the message profiler 500 has an associated confidence value 520. When a message arrives for profiling, the message profiler 500 iterates through the classification techniques and allows each technique to attempt to classify the message. The result of each classification is a decimal value in the range [0, 1]. After iterating through each classification technique, the message profiler 500 computes a score for the message using the following formula:

0082N ScOTe = Y₄SViXd

00831=1 where SVi is the confidence value associated with classification technique i, and Q is the classification value in [0, 1] produced by classification technique i. In using classification techniques with non-linear scoring functions, the following formula can be used:

N

0085Score = ]T (SVu X Ct + SVn X C₁² )

0086where SVu and SV₂; are the confidence value associated with classification technique i, and C, is the classification value in [0, 1] produced by classification technique /.

0087If the message score exceeds some specified threshold T as determined at 520, then the message is declared to belong to the first defined category. If the message score is below the threshold, it is declared to belong to the opposite category. The system can then take an appropriate action based on the threshold reached by the message score, such as quarantining the message, dropping the message (i.e., deleting the message without delivery as shown at 530), rewriting the subject of the message to contain some specific string (e.g., "SUSPECTED SPAM"), passing on a message to the encryption engine for secure delivery, etc. The system can also allow for specifying multiple thresholds and applying a different action or actions at each threshold, which would signify the increased confidence of the message profiler 500 in the result of the classification.

0088The effectiveness and accuracy of a message profiler 500 is dependent on several factors, such as on the set of SV; or SV₂/SV₂; confidence values 520 associated with the classification techniques 510. A tunable message classification configuration can be used to generate an optimized set of values along with an associated set of thresholds and actions and that can be generated periodically to keep a message profiler 500 updated with the latest protection against the frequent changes in the score distributions of classification techniques operating on the constantly changing message flow patterns. As such, a message profiler configuration includes a vector <img file="WO2006052736A2_D0004.tif" />

0089(which represents the confidence values of all N classification techniques). As shown in FIG. 10, a message classification tuner program 600 can be configured to tune a message profiler 500 by performing a probabilistic search through the vector space of all possible vectors and identifying a vector that maximizes the filtering accuracy of the profiler for a pre-selected threshold. The tuner 600 can use different approaches to do this, such as by using a heuristic approach 610.

0090FIG. 11 illustrates the tuner using a heuristic approach known as a genetic algorithm 650 in order to perform a vector space search. The concepts underpinning a genetic algorithm come from the theory of evolution, in which genotypes (expressed via chromosomes) compete with each other via their phenotypes (expressed as biological organisms). Over time, biological evolution produces highly adapted and complex organisms that are capable of surviving in the environment for which they evolved. Similarly, a genetic algorithm searches through a vector space consisting of candidate solutions to a problem, where each candidate solution is expressed as a vector. Over many simulated generations of candidate solutions, the genetic algorithm gradually evolves towards increasingly well-adapted solutions to the problem.

0091The genetic algorithm's ability to evolve good solutions to a problem over time depends upon the existence of an accurate mechanism for evaluating the relative fitness level of a candidate solution compared to other candidate solutions. Thus, the genetic algorithm 650 is designed with a fitness function 660 that accurately models the fitness of candidate solutions in the actual problem domain.

0092Below is a fitness function 660 that could be used for optimization of a message profiler 500:

0093y \SCAT1 _ MISTAKES, ~ T Y" \ScAT2_ MIKSTAKESj — T + ll

0094Fitness = ^ + Cx ^ ^]-

0095NCATl NCΛT2

0096The definitions of terms in the function are as follows:

0097Nc_ATi = number of message vectors from the overall data set that belong to the first category

0098N_CA_T2 = number of message vectors from the overall data set that belong to the second category C = constant multiplier for misclassified messages from the second category ScATij_ti_KSTAKEi = message profiler score of message vector i from the first message category that has been misclassified to belong in the other category ScAT2jaisτAKEi = message profiler score of message vector i from the second message category that has been misclassified to belong in the other category

0099T = message profiler numeric threshold above which a message is considered to belong to the first category

0100The function expresses the cost associated with the mistakes that a configuration made in attempting to correctly classify message vectors in the set of pre-classified data. Accordingly, a lower fitness value is considered better for the genetic algorithm's purposes. The first term in the function expresses the cost associated with messages from the first category that have been misclassified to belong in the second (e.g., unwanted messages that have been classified as legitimate, also known as false negatives), and the second term expresses the cost associated with messages from the second category that have been misclassified to belong in the first (e.g., legitimate messages that have been classified as unwanted, also known as false positives). The summations represent the total number of points by which a configuration was wrong when attempting to classify message vectors. Intuitively, each term is essentially an expression of both the average frequency of classification errors and the average magnitude of classification errors. Note that the second term is to be multiplied by a constant, C. This constant (which can be set to a value of 20) represents the relative cost of a misclassification of a message from one category in relation to misclassification of a message from the opposing category. By setting C to 20, this indicates that classification mistakes on messages from the second category are 20 times more costly than mistakes from the second category. For example, if a message profiler 500 is used for classification of wanted and unwanted mail, the first category would represent the unwanted mail (e.g., spam) and the second category would represent legitimate messages. Then the above function would deem misclassifications of legitimate messages (false positives) to be 20 times as costly as misclassification of unwanted messages (false negatives). This reflects the real-world view in the anti-spam community that false positives carry much higher risk than false negatives. If a message profiler 500 is used for policy compliance-related classification, a false positive is a message that contains sensitive information but is not labeled as such by the message profiler 500 and is, therefore, allowed to evade the policies that an organization may have chosen to apply to that particular category. FIG. 12 depicts an operational scenario wherein a message profiler can be used. With reference to FIG. 12, the operational scenario includes receiving at step 710 a communication that was sent over a network from a messaging entity. A plurality of message classification techniques is then used at 710 to classify the communication. Each message classification technique is associated with a confidence value which is used in generating a message classification output from the message classification technique. The output of each classification can be numeric values, textual values, or categorical values. The message classification outputs are combined at step 720 in order to generate a message profile score at step 730. The message profile score is used at step 740 to decide what action is to be taken with respect to the communication associated with the messaging entity.

0101It should be understood that similar to the other processing flows described herein, the processing and the order of the processing may be altered, modified and/or augmented and still achieve the desired outcome. For example, a message profiler may be configured for an operational scenario that recognizes that there is a single technique is not capable of adequately classifying a message into two distinct categories, such as distinguishing between wanted (legitimate) and unwanted (spam, phishing, viruses, etc) message communications or determining whether a message complies with a specific organization policy, law, or regulation. In this operational scenario, such a configured message profiler can be designed to:

01021. Provide a framework for combining the results of many message classification techniques into an aggregate classification (such as "unwanted" or "legitimate", "HIPPA compliant", "GLBA violation", "HR policy violation", etc), without specifying a priori which classification technique(s) will be used.

01032. Decouple each classification technique's importance (expressed via its contribution towards the aggregate classification) from its classification logic, so that a technique's level of importance may be adjusted to reflect changes in its accuracy over time.

01043. Provide a mechanism through which to describe the relative importance of each classification technique within the framework and the correlation of their individual accuracy, so that the framework can be adjusted to use this information to achieve very accurate rates in aggregate classification.

01054. Provide a mechanism through which to discover the relative importance of each classification technique within the framework, so that the framework can be "tuned" for maximum classification accuracy in a given environment. Still further, a message profiler may be configured to operate in other operational scenarios. For example, FIG. 13 depicts a message profiler that has been adapted to operate with adaptive message blocking and whitelisting. With reference to FIG. 13, in addition to classification of individual messages, the aggregated results of a message profiler program 500 can also be used for classifying at 820 senders of messages based on the distributions of message profiler scores that their messages are receiving. If the average score of messages received from a particular sender (e.g., IP) during a specified timeframe (e.g., hour, day, week) exceeds a specified threshold 7V and the score distribution has a standard deviation smaller than STa, that sender can be classified 'un-reputable' (which information is stored in data store 840). Process 800 can then use the data from data store 840 to determine that all messages and connections originating from such a sender can be dropped at 810 without processing for the next X hours. Correspondingly, if the average score is below threshold T_L with a standard deviation smaller than ST_L, the sender can be considered legitimate (which information is stored in data store 830) and messages from that sender can be allowed by process 800 to bypass certain filtering techniques (e.g., the filtering of message profiler 500) that can cause significant processing, network, or storage overhead for the filtering system 460.

0106A message profiler may also be used in connection with adaptive training of endo and exo-filtering systems. Using the systems and methods of sender classifier described herein, a message profiler can be used for training of the various filtering techniques that are used within the profile, as well as others that lie completely outside of it. Such techniques may include Bayesian, Support Vector Machine (SVM) and other statistical content filtering techniques, as well as signature-based techniques such as Statistical Lookup Service (SLS) and message clustering-type techniques. The training strategies for such techniques may use sets of classified legitimate and unwanted messages, which can be provided by the message profiler based on sender reputations, assigned from the aggregate scores of messages from such senders. Messages from senders classified as un-reputable can be provided to the filtering system trainer as unwanted, and the wanted messages will be taken from stream sent by the legitimate senders.

0107As described above, a message profiler 500 may use a reputation-based approach as one classification technique. FIG. 14 depicts at 900 a reputation system that can be used by a filtering system 460 in handling transmissions received over a network 440 from messaging entities 450. More specifically, the filtering system 460 uses the reputation system 900 to help determine (at least in part) what filtering action (if any) should be taken upon the messaging entities' communications. For example, the communication may be determined to be from a reputable source and thus the communication should not be filtered.

0108The filtering system 460 identifies at 950 the sender of a received communication and provides that identification information to the reputation system 900. The reputation system 900 evaluates the reputation of the queried sender's identity by calculating probabilities that a messaging entity exhibits certain characteristics. An overall reputation score is determined based upon the calculated probabilities and is provided to the filtering system 460. A reputation score can be numeric, textual, or categorical in value. The filtering system 460 determines at 952 what action should be taken for the sender's communication. The filtering system 460 could use the reputation score from the reputation system 900 as a message classification filter which is to be multiplied by its respectively tuned confidence value and then aggregated with other message classification filter results.

0109Reputation systems may be configured in many different ways in order to assist a filtering system. For example, FIG. 15 depicts the reputation system 900 that has been configured to calculate reputation scores. The system's configuration 1000 can be established by identifying a set of binary, testable criteria 1002 which appear to be strong discriminators between good and bad senders. P (NR\Cϊ) can be defined as the probability that a sender is non-reputable, given that it conforms to quality/criterion Q, and P (R\Q) can be defined as the probability that a sender is reputable, given that it conforms to quality/criterion Q.

0110For each quality/criterion Q, periodic (e.g., daily, weekly, monthly, etc.) sampling exercises can be performed to recalculate P (NR\Q). A sampling exercise may include selecting a random sample set S of N senders for which quality/criterion Ci is known to be true. The senders in the sample are then sorted into one of the following sets: reputable (R), non-reputable (NR) or unknown (U). N_R is the number of senders in the sample that are reputable senders, NN_R is the number of senders that are non-reputable senders, etc. Then, P (NR\Cϊ) and P (R\Ci) are estimated using the formulas:

0111P(NR I G) = ^ ¹ _V

0112P(R \ Cι) = —

N

0114For this purpose, N = 30 was determined to be a large enough sample size to achieve an accurate estimate of P (NR\Ci) and P (R\Q) for each quality/criterion C₁. After calculating P (NR|C₍) and P (R]Ci) for all criteria, the computed probabilities are used to calculate an aggregate non-reputable probability 1004, P_MR, and an aggregate reputable sender probability 1006, P_R, for each sender in the reputation space. These probabilities can be calculated using the formulas:

0115/ w (Λ π_/ *τr>

0116_P1- I- P(NR

0117^Pm - ' ^i"-π ^{I X | 1}<img file="WO2006052736A2_D0005.tif" />

0118<img file="WO2006052736A2_D0006.tif" />

0119In experimentation, the above formulas appeared to behave very well for a wide range of input criteria combinations, and in practice their behavior appears to be similar to the behavior of the formula for correctly computing naive joint conditional probabilities of "non-reputable" and "reputable" behavior for the input criteria.

0120After calculating P_NR and P_R for each sender, a reputation score is calculated for that sender using the following reputation function:

0121/(*_ΝR , P_R ) = (C₁ + c₂P_m + c₂P_R + cA + C₃P^ + c₄P_NRP_R + c₅P_N³_R + c₅P_R³ + <img file="WO2006052736A2_D0007.tif" /> + c₆P^P_R X(P₁₀₁ - P_R Ϋ + C₁ (P_m - P_R ))

0122where

0123C₁ = 86.50 c₂ = -193.45 c₃ = -35.19 c₄ = 581.09 c₅ = 234.81 C₆ = -233.18 C₇ = 0.51 It should be understood that different functions can act as a reputation score determinator 1008 and can be expressed in many different forms in addition to a functional expression. As an illustration, FIG. 16 depicts at 1100 a tabular form for determining reputation scores. The table shows reputation scores produced by the above function, based on values of PNR and P_R as they each vary between 0.0 and 1.0. For example as shown at 1110, a reputation score of 53 is obtained for the combination of P_NR = 0.9 and P^ = 0.2. This reputation score is a relatively high indicator that the sender should not be considered reputable. A reputation score of 0 is obtained if P_NR and P_R are the same (e.g., the reputation score is 0 if P_NR = 0.7 and P_Λ = 0.7 as shown at 1120). A reputation score can have a negative value to indicate that a sender is relatively reputable as determined when P_R is greater than PNR- For example, if P_NR = 0.5 and P_R = 0.8 as shown at 1130, then the reputation score is -12.

0124Many different types of criteria may be used in a reputation system's processing of a sender's communication, such as using non-reputable criteria and reputable criteria to determine reputation scores. Examples of such criteria are disclosed in U.S. Provisional Application Ser. No. 60/625,507, entitled "CLASSIFICATION OF MESSAGING ENTITIES," filed on November 5, 2004.

0125The systems and methods disclosed herein are presented only by way of example and are not meant to limit the scope of the invention. Other variations of the systems and methods described above will be apparent to those skilled in the art and as such are considered to be within the scope of the invention. For example, a system and method can be configured to handle many different types of communications, such as legitimate messages or unwanted communications or communications violative of a pre-selected policy. As an illustration, an unwanted communication could include a spam or virus communication, and a pre-selected policy could include a corporate communication policy, a messaging policy, a legislation or regulatory policy, or an international communication policy.

0126As another example of the wide scope and variations of the systems and methods disclosed herein, the systems and methods may be implemented on various types of computer architectures, such as for example on different types of networked environments. As an illustration, FIG. 17 depicts a server access architecture within which the disclosed systems and methods may be used (e.g., as shown at 1330 in FIG. 17). The architecture in this example includes a corporation's local network 1290 and a variety of computer systems residing within the local network 1290. These systems can include application servers 1220 such as Web servers and e-mail servers, user workstations running local clients 1230 such as e-mail readers and Web browsers, and data storage devices 1210 such as databases and network connected disks. These systems communicate with each other via a local communication network such as Ethernet 1250. Firewall system 1240 resides between the local communication network and Internet 1260. Connected to the Internet 1260 are a host of external servers 1270 and external clients 1280. It should be understood that the present disclosure can any variety of network, including, but not limited to an intranet, wireless network, wide area networks, local area networks, and combinations thereof, in order to facilitate communication between components. Local clients 1230 can access application servers 1220 and shared data storage

01271210 via the local communication network. External clients 1280 can access external application servers 1270 via the Internet 1260. In instances where a local server 1220 or a local client 1230 requires access to an external server 1270 or where an external client 1280 or an external server 1270 requires access to a local server 1220, electronic communications in the appropriate protocol for a given application server flow through "always open" ports of firewall system 1240.

0128A system 1330 as disclosed herein may be located in a hardware device or on one or more servers connected to the local communication network such as Ethernet 1280 and logically interposed between the firewall system 1240 and the local servers 1220 and clients 1230. Application-related electronic communications attempting to enter or leave the local communications network through the firewall system 1240 are routed to the system 1330.

0129In the example of FIG. 17, system 1330 could be configured to store and process reputation data about many millions of senders as part of a threat management system. This would allow the threat management system to make better informed decisions about allowing or blocking electronic mail (e-mail).

0130System 1330 could be used to handle many different types of e-mail and its variety of protocols that are used for e-mail transmission, delivery and processing including SMTP and POP3. These protocols refer, respectively, to standards for communicating e-mail messages between servers and for server-client communication related to e-mail messages. These protocols are defined respectively in particular RFCs (Request for Comments) promulgated by the EETF (Internet Engineering Task Force). The SMTP protocol is defined in RFC 1221, and the POP3 protocol is defined in RFC 1939.

0131Since the inception of these standards, various needs have evolved in the field of e-mail leading to the development of further standards including enhancements or additional protocols. For instance, various enhancements have evolved to the SMTP standards leading to the evolution of extended SMTP. Examples of extensions may be seen in (1) RFC 1869 that defines a framework for extending the SMTP service by defining a means whereby a server SMTP can inform a client SMTP as to the service extensions it supports and in (2) RFC 1891 that defines an extension to the SMTP service, which allows an SMTP client to specify (a) that delivery status notifications (DSNs) should be generated under certain conditions, (b) whether such notifications should return the contents of the message, and (c) additional information, to be returned with a DSN, that allows the sender to identify both the recipient(s) for which the DSN was issued, and the transaction in which the original message was sent. In addition, the IMAP protocol has evolved as an alternative to POP3 that supports more advanced interactions between e-mail servers and clients. This protocol is described in RFC 2060.

0132Other communication mechanisms are also widely used over networks. These communication mechanisms include, but are not limited to, Voice Over D? (VoIP) and Instant Messaging. VoIP is used in IP telephony to provide a set of facilities for managing the delivery of voice information using the Internet Protocol (IP). Instant Messaging is a type of communication involving a client which hooks up to an instant messaging service that delivers communications (e.g., conversations) in realtime.

0133As the Internet has become more widely used, it has also created new troubles for users. In particular, the amount of spam received by individual users has increased dramatically in the recent past. Spam, as used in this specification, refers to any communication receipt of which is either unsolicited or not desired by its recipient. A system and method can be configured as disclosed herein to address these types of unsolicited or undesired communications. This can be helpful in that e-mail spamming consumes corporate resources and impacts productivity. It is further noted that the systems and methods disclosed herein may use data signals conveyed via networks (e.g., local area network, wide area network, internet, etc.), fiber optic medium, carrier waves, wireless networks, etc. for communication with one or more data processing devices. The data signals can carry any or all of the data disclosed herein that is provided to or from a device. Additionally, a methods and systems described herein may be implemented on many different types of processing devices by program code comprising program instructions that are executable by one or more processors. The software program instructions may include source code, object code, machine code, or any other stored data that is operable to cause a processing system to perform methods described herein.

0134The systems' and methods' data (e.g., associations, mappings, etc.) may be stored and implemented in one or more different types of computer-implemented ways, such as different types of storage devices and programming constructs (e.g., data stores, RAM, ROM, Flash memory, flat files, databases, programming data structures, programming variables, IF-THEN (or similar type) statement constructs, etc.). It is noted that data structures describe formats for use in organizing and storing data in databases, programs, memory, or other computer-readable media for use by a computer program.

0135The systems and methods may be provided on many different types of computer-readable media including computer storage mechanisms (e.g., CD-ROM, diskette, RAM, flash memory, computer's hard drive, etc.) that contain instructions for use in execution by a processor to perform the methods' operations and implement the systems described herein.

0136The computer components, software modules, functions and data structures described herein may be connected directly or indirectly to each other in order to allow the flow of data needed for their operations. It is also noted that software instructions or a module can be implemented for example as a subroutine unit of code, or as a software function unit of code, or as an object (as in an object-oriented paradigm), or as an applet, or in a computer script language, or as another type of computer code or firmware. The software components and/or functionality may be located on a single device or distributed across multiple devices depending upon the situation at hand.

0137It should be understood that as used in the description herein and throughout the claims that follow, the meaning of "a," "an," and "the" includes plural reference unless the context clearly dictates otherwise. Also, as used in the description herein and throughout the claims that follow, the meaning of "in" includes "in" and "on" unless the context clearly dictates otherwise. Finally, as used in the description herein and throughout the claims that follow, the meanings of "and" and "or" include both the conjunctive and disjunctive and may be used interchangeably unless the context clearly dictates otherwise; the phrase "exclusive or" may be used to indicate situation where only the disjunctive meaning may apply.