[Solved] Email Tracking BOT Detection

1 Email Tracking BOT Detection Data Exploration1.1 ProblemMore and more, email tracking data is originating from BOT activity as opposed to organically by the actual person the email is intended for. BOTs that look at emails are automated programs that open the email and/or click on links within the email. A BOTs intentions can be malicious, data gathering or beneficial. An example of a BOT could be a service that subscribes to various email lists and clicks on links to extract coupons.Through manual investigation, most of the BOTs found in this dataset are not malicious but come from protective services like Anti-Phishing devices that protects the email contact from dangerous emails.1.2 GoalTo find ways to identify which requests are from BOT activity and which are organic actively from real people. If we can reliably identify which requests are from BOT activity, we would then know IP Addresses or blocks of IPs that the BOTS are coming from. Knowing the IPs used by BOTs would allow us to take different actions between BOT activity and organic activity.But since manually review shows most of the BOTs are protective, we dont want to change the response of the request. In other words, we want the click to take you to the same URL if it is a BOT or organic. But we can do things like adding the activity to reporting.1.3 Why do we Care?It is critical for customers who send email to be able to judge the effectiveness of the email campaign. Without accurate open and click tracking, customers can not make modifications to improve their campaigns and cannot do things like A/B testing.1.4 What is Not looked at?Though it would be a big benefit to know if the BOT is malicious, data mining or protective, this is not looked at. To determine this, a lot of manually research along with subjective conclusions are required and not in the scope of this project.1.5 Email Tracking Data Overview:

1.6 Terminology Subscriber List A list of email addresses that have opted into receiving emails from the company. A list have have many emails. A company can have many subscriber lists. An email can belong to many lists and many companies. But a subscriber list can only belong to one company. Campaign An email message send out by a company. This email is sent to some or all the subscribers on the list. With the exception of some limited personalization, all subscribed receive the same content. A list can have many campaigns, but a campaign can only have one subscriber list. Inbox What the contact sees as a unique message in their inbox (Outlook, GMail, etc.). If a company sends a single campaign to 1 million contacts, there will be 1 million InboxIDs for that campaign. Request A tracking HTTP request from an action the contact takes on a message in their inbox. There are several types of requests, but they are primarily Opens which are pixel image reqeusts and Clicks. Link A clickable and trackable URL embedded in the email message. The tracking system identifies the click, saved the tracking information and returns a redirect URL so the contact gets to their intended URL destination. AS Number/Name Who owns the block of IP address. https://en.wikipedia.org/wiki/Autonomous_system_(Internet) CIDR A notation of shows us what range of IP address are in a block. https://en.wikipedia.org/wiki/Classless_Inter-Domain_Routing Network Address Trasnlation NAT Most computers and phones do not have a public IP Address. NAT is a way for many different computer to reuse the same public IP Address.https://en.wikipedia.org/wiki/Network_address_translation1.7 Columns#### Raw columns recieved with HTTP Request: InboxID Uniquely identifies the Campaign and the email it was sent to. Company Who owns the campaign RequestType Open The contact downloaded images for the message. A tracking pixel added to the HTML body of the message makes a request to the tracking system to identify who opened the message. Read The message was opened for a pre-defined period of time. Click The contact clicked on a URL in the message and the tracking system sees this request. Browser Link The contact requested seeing the message in a web browser. Mostly used if the message did not render well in the email client. RequestDate The date of the request (i.e. Open or Click). Link A hash of the URL requested. IPAddress The IP Address the requested was generated from. UserAgent A string submitted by email clients/browsers and Bots that attempts to identify the browser type. This can be made any string you want to it can be easily modified by Bots. #### Columns that can be looked up from the InboxID: SendDate The date and time (1 minute precision) of when the message was sent to that email. If a company sends a campaign to 5 million email, the send might take 10-20 mins. The SendDate is the date when that message was sent out and not the date the campaignstarted. List A unique ID of the subscription list. Campaign A unique ID of the campaign. UserID A unique ID of the email address the message was sent to. DomainID A unique ID of the domain of the email address the message was sent to. EmailDomain (redundant with DomainID)- A unique ID of the domain of the email address the message was sent to. EmailRootLevelDomain The top level domain of the email domain. i.e. [email protected] .com, [email protected] #### Columns that are derived from IP lookup data: AS Name Owner of the IP Range (see above). Can be Null if owner not found in ASN database. AS Number Owner of the IP Range (see above). Can be Null if owner not found in ASN database. CIDR Range The block the IPAddress belongs to (see above). Can be Null if owner not found in ASN database..edu Lat/Long Found from an IP Lookup service. Data available is not center of mass, but just the first Lat/Log in the polygon. OlsonName Time Zone of the location the request came from. Also found via the IP Lookup.Calculated columns after InboxID lookup is performed SendRequestSeconds Number of seconds between the time the message was sent to the request date.## Hierarchies in Data

Most computers and phones do not use a stand-along IP Address. NAT is used to give a common public IP address for out web requests to the internet.1.7.1 Because of this, an IP Address can be used my many people and many computers/phones.So a single IP might be 100 People Common places where NAT is used: Work Most companies will use NAT with a small number of public IP Address. If the company has 100 employees, all 100 employees will reuse the same few public IP Address. Internet Service Provider The subscribers to an ISP will share public IP Address. Bigger ISPs can have many subnets and therefore a lot of public IP Address. But there can still be hundreds of people reusing the same IP Address. Cell Phone Carriers NAT is used by you phone carriers. Carriers like AT&T have thousands of public IP Addresses, but millions of cell phones. Cloud Computer Platforms (AWS) Users of these platforms may or may not use NAT. So, if the IP Address is traced back to Amazon Web Services, this public IP may or may not belong to a single AWS account.1.7.2 The result, IP Address does not map to a single person or computer

1.7.3 BotDataSet ClassWe will use a class to make sure the loading, cleanup and sessionization is done consistantly for both the IP Reputation and the Campaing dataset. This class will also give us code reuse for the trainging and testing datasets.

0 {85A0B8A0-4712-1803-93FC-B8461995CFC9}1 {85A0B8A0-4712-1803-93FC-B8461995CFC9}2 {85A0B8A0-4712-1803-93FC-B8461995CFC9}3 {85A0B8A0-4712-1803-93FC-B8461995CFC9}4 {F4724933-4570-E4BE-9B17-3D0928E9EE76}Company {6C357BE7-9D8A-9D1E-FD3C-9D4C45077541}1 {6C357BE7-9D8A-9D1E-FD3C-9D4C45077541}2 {6C357BE7-9D8A-9D1E-FD3C-9D4C45077541}3 {6C357BE7-9D8A-9D1E-FD3C-9D4C45077541}4 {6C357BE7-9D8A-9D1E-FD3C-9D4C45077541}List {15D09884-3FDD-8390-8493-A428F11F308D}1 {15D09884-3FDD-8390-8493-A428F11F308D}2 {15D09884-3FDD-8390-8493-A428F11F308D}3 {15D09884-3FDD-8390-8493-A428F11F308D}4 {15D09884-3FDD-8390-8493-A428F11F308D}Campaign RequestType SendDate {C1D499D6-AA43-9ABA-4EFD-0D97D0A33BE8} Click 2020-01-17 09:03:001 {C1D499D6-AA43-9ABA-4EFD-0D97D0A33BE8} Click 2020-01-17 09:03:002 {C1D499D6-AA43-9ABA-4EFD-0D97D0A33BE8} Click 2020-01-17 09:03:003 {C1D499D6-AA43-9ABA-4EFD-0D97D0A33BE8} Click 2020-01-17 09:03:004 {C1D499D6-AA43-9ABA-4EFD-0D97D0A33BE8} Click 2020-01-17 09:03:00RequestDate Link 2020-01-17 09:03:00 {2091AC3A-12C7-8FBF-3AA8-8C8C5F874EE2}1 2020-01-17 09:03:00 {C7757DCC-7F71-8FAB-6836-0B061F8668F0}2 2020-01-17 09:03:00 {B68FB2B6-F5FF-B7DA-22DD-589DF7AE29F7}3 2020-01-17 09:03:00 {81E9F200-AB2B-54DE-B91D-506F173DA998}4 2020-01-17 09:03:00 {048D23DD-04F1-D783-DB55-4BA2891DAA60}IPAddress {9AD7D4CE-E247-DA9B-6E99-F5255E45F42F}1 {1508EC2E-DA30-4661-442A-74EAD9243FE2}2 {08E594F2-2009-3FB3-17F6-A37ABD24F6C9}3 {4A03B84C-1010-D5C7-3541-127FBB258929}4 {49F7F9A3-19F9-6E8A-5896-BC10581E5868}UserAgent AS Name AS Number {613F6166-E32A-4AE1-9FC4-7927349BDFE7} Unknown Unknown1 {613F6166-E32A-4AE1-9FC4-7927349BDFE7} Unknown Unknown2 {613F6166-E32A-4AE1-9FC4-7927349BDFE7} Unknown Unknown3 {613F6166-E32A-4AE1-9FC4-7927349BDFE7} Unknown Unknown4 {30634283-D150-237A-87FD-DA097EBBE5B9} Amazon.com, Inc. 16509CIDR Range Unknown1 Unknown2 Unknown3 Unknown4 {4ED21F0C-8573-052A-670B-E4D315133AF5}UserID {689E664D-1110-0BB1-7C6F-0DB872A731FA}1 {689E664D-1110-0BB1-7C6F-0DB872A731FA}2 {689E664D-1110-0BB1-7C6F-0DB872A731FA}3 {689E664D-1110-0BB1-7C6F-0DB872A731FA}4 {F89AA673-49FE-433F-7447-040277C7D897}DomainID {80050545-EC9D-D79E-A18E-1666F4C81AAB}1 {80050545-EC9D-D79E-A18E-1666F4C81AAB}2 {80050545-EC9D-D79E-A18E-1666F4C81AAB}3 {80050545-EC9D-D79E-A18E-1666F4C81AAB}4 {C595CAFD-E63F-58ED-E8C8-8DB70C548720}EmailDomain EmailRootLevelDomain CountryCode {685CD069-4D61-F6AD-A3BE-0D814C0BC40C} .org US1 {685CD069-4D61-F6AD-A3BE-0D814C0BC40C} .org US2 {685CD069-4D61-F6AD-A3BE-0D814C0BC40C} .org US3 {685CD069-4D61-F6AD-A3BE-0D814C0BC40C} .org US4 {C6A21CB1-169B-B0F6-A812-35899D02D459} .org USRegionCode PostalCode Lat Long OlsonName VA 22102 38.98191 -77.256903 America/New_York1 VA 22102 38.98191 -77.256903 America/New_York2 VA 22102 38.98191 -77.256903 America/New_York3 VA 22102 38.98191 -77.256903 America/New_York4 OR 97207 45.850021 -119.631154 America/Los_AngelesSendRequestSeconds SendRequestSeconds_ln0 0 1.01 0 1.02 0 1.03 0 1.04 0 1.0(24058, 25)RequestDate Range: 2020-01-17 09:03:00:2020-01-21 11:49:00[4]: InboxID 11566Company 1List 1CampaignRequestType 1 1

SendDate 13RequestDate 2187Link 37IPAddress 10339UserAgent 1181AS Name 21AS Number 22CIDR Range 354UserID 11566DomainID 952EmailDomain 952EmailRootLevelDomain 32CountryCode 77RegionCode 148PostalCode 2572Lat 1612Long 1659OlsonName 79SendRequestSeconds 2180SendRequestSeconds_ln 2180dtype: int642 Only One Metrizable Column in Raw Request Data2.1 But this metric is one of the major factors that we anticipate will help us identify BOTs.Mose of the click BOTs are safty devices that examin the risk of the email before it is released to the contacts inbox. So to prevent causing a long delay, the BOTs will want to process the links as soon as they are recieved. Through manual, most of the suspected BOTs click the message within 1 or 2 minutes after send went out.plt.xlabel(Seconds between Send and Click, fontsize=18)2.2 As you can see from the histogram, there is an overwelming number of clicks in the first few minusts after the send.To even out this distribution, the BotDataSet class created a calculated column called SendRequestSeconds_ln This was done by taking the natural log of the SendRequestSeconds. To avoid zero and negative, the number e was added before the the log was taking. This make the bottom of the scale = 1. To give a feel of the upper limit of the this value, 2 days (172800 seconds) would give a value of 14.77.2.3 $ SendRequestSeconds_ln = log_{e}(SendRequestSeconds + e) $

3 SessionizationBecause of NAT, an IP Address might be reused by many different people and many different computers/phones. So we cannot say click activity on the same IP Address for a 1,000 different emails must be a BOT. It is just as likely that the IP Address is used by a cell phone carrier.Sessionization help to mitigate this problem. Requests coming from the same IP Address in a short sequence of time has a much greater chance of being from the same person/computer. Combining IP and User Agent would tend to strengthen the correlation to a single person.If we sessionize based on InboxID, we have a very good chance of getting requests from a single person. But if the requests are from a BOT, they could be processing many different InboxIDs on several IP/UserAgents at the same time.

3.1 Extend Dataframe with Session ColumnsWe want to sessionize in 3 ways, the InboxID, P Address and the combination of IPAddress and User Agent3.1.1 Inbox Based InboxID This is straight forward. Any requests from the same inbox message will have the same session if there are not any gaps over 2 mins. This type of session will be used to judge behavior of the Contact.3.1.2 IP Based IP Address This type of sessionization will be used to judge the behavior of IP Address from where the requests come from. If I click links from 2 different messages, the 2 InboxIDs will come from the same IP Address. But it might be unusual to have requests from many different InboxIDs from the same IP. Especially if the inboxes are to different emails and different email domains. IP & UserAgent Historically, BOT detection has been done with a status list of IP/UserAgent combinations. Unsophisticated BOTs can use only 1 or 2 IP Address and always use the same User Agent string. These were easy to detect. Even though this approach is not adequate with the current times, the combination of IP/UserAgent is still very much a valid ay to sessionize requests. This holds up with manually scanning of data where we see the same UserAgent string used over and over again. Even if the UA they use is a common one, i.e. iPhone 10, the fact that the same IP Address uses the exact same User Agent over and over again for many different emails could be highly suspicious.Below, we use the BotDataSet class to append session columns to our dataframe.[8]: max_session_time_gap = timedelta(seconds=120)RequestDate InboxID 2020-01-17 09:03:00 {85A0B8A0-4712-1803-93FC-B8461995CFC9}71 2020-01-17 09:03:00 {834A4BE1-D5F7-2143-E451-B7D2189175AE}91 2020-01-17 09:03:00 {183046DF-1656-1D94-BBF1-63289C49F43C}92 2020-01-17 09:03:00 {A36C9EE8-B8CC-AC57-DC47-CEA3D1F99726} 93 2020-01-17 09:03:00 {30E5C814-9797-EA1B-ED40-4165172C81F8}IPAddress {9AD7D4CE-E247-DA9B-6E99-F5255E45F42F}71 {6AC33884-ADE3-2F3E-6DD7-1731EC5D2945}91 {0793D471-B890-C0C1-2C0B-DF4882B1D58C}92 {0793D471-B890-C0C1-2C0B-DF4882B1D58C}93 {0793D471-B890-C0C1-2C0B-DF4882B1D58C}CIDR Range AS Number InboxSessionID Unknown Unknown -570508913087568472971 {BF61FA52-8010-2731-2B4E-0C77A6C05E6A} 16509 -841045530237293781191 {7EF79BFD-BC3A-903D-C2BD-F5E295ACF4DE} 16509 -215553791378811038192 {7EF79BFD-BC3A-903D-C2BD-F5E295ACF4DE} 16509 419771867020121818993 {7EF79BFD-BC3A-903D-C2BD-F5E295ACF4DE}IPSessionID IPUASessionID0 -5152777450074098447 -462974343165331664871 263756345889371100 240273446379052377791 482115446712771047 623245686726965411792 482115446712771047 623245686726965411793 482115446712771047 6232456867269654117[9]: RequestDate 2187 InboxID 11566IPAddress 10339CIDR Range 354 AS Number 22InboxSessionID 12398 IPSessionID 11532 IPUASessionID 11638 dtype: int64 16509 -34676541194461105964 Data exploration of the 3 Session Types4.0.1 Approach (for each Session Type): Build Aggregations i.e. UniqueInboxIDs, UniqueEmails, SessionDuration. Reduce Aggregations Build another DF with just the aggregations we want to use in our unsupervised learning. We will use the other aggregations for visualization after we add labels. Rescale Features Use MinMaxScaler. Look at Covariance Ideally, we would like to see a lot of independence between the features. Explore Clustering Since the features are derived from aggregations, most of the data will be count data. Count data tends to follow an exponential decay PDF. Since our features are not Gaussian distributed, they will not be clustered around a mean and will be non-isotropic. A lot of the scatter plots will tend to show grouping around lines and curves. Because of this, DBSCAN looks to be a good fit for our initial clustering.4.1 Build AggregationsMost of the aggregations will be the same for all the session types. But there will be some differences. For example, the IP/UserAgent session does not make sense to use UniqueUserAgents since this type of session only has one User Agent string.4.1.1 Build Column Lists

4.2 Run groupby for each Session Type4.2.1 Inbox[11]: # Inbox df_Inbox_agg = df.groupby([InboxID]).agg(Inbox_agg) df_Inbox_agg.rename(columns=Inbox_agg_renames, inplace=True)# IP Onlydf_IP_agg = df.groupby([IPSessionID]).agg(IP_agg) df_IP_agg.rename(columns=IP_agg_renames, inplace=True)#IP and User Agentdf_IPUA_agg = df.groupby([IPUASessionID]).agg(IPUA_agg) df_IPUA_agg.rename(columns=IPUA_agg_renames, inplace=True)4.3 Reduce Aggregationsuse copy() to avoid views[12]: df_Inbox_agg_unsup = df_Inbox_agg[Inbox_unsup_columns].copy() df_IP_agg_unsup = df_IP_agg[IP_unsup_columns].copy() df_IPUA_agg_unsup = df_IPUA_agg[IPUA_unsup_columns].copy()4.4 Rescale FeaturesWe will use these new DataFrames for out unsupervised data exploration

4.5 Look at Covariance

4.6 Conclusions:The Inbox session data does look there is a fair amount of covariance, but without excessively dependent. PCA might make sense, but since we only have 7 features, we will skip this step.Between the 2 IP bases sessions, the correlation matrix does not look very different. Since historically we know the combination of IP and User Agent have be successful in filtering out BOT activity, we will choose to use it for out IP session.5 Explore Clustering

6 Clustering IP/UA Session Data6.1 Lets look at all the scatter plots to help us choose the best unsupervised clustering modelAs you can see, the data does not form any obvious globular clusters. Since this data is mostly aggregate counts, the distributions are more likely exponential decay then Gaussian.Because of this, K-Mean would probably not be a good fit. An approach using nearest neighbors like DBSCAN should be a good probably be a good starting point.

[15]: <seaborn.axisgrid.PairGrid at 0x18cdaf1f948>

7 DBSCAN IP/UA Session

7.1 Append the DBSCAN Lables to the 2 IP group by DataFrames[18]: df_IPUA_agg[IPUA_Labels] = labels #Origional Group By with additional agg,columns df_IPUA_agg_unsup[IPUA_Labels] = labels #Normilized Aggregations df_IPUA_agg_unsup_s[IPUA_Labels] = labels #Not=Normilized Aggregations7.2 Build visualizations to manually inspect effectiveness of clusters

[20]: np.seterr(divide=ignore, invalid=ignore) sns.pairplot(df_IPUA_agg_unsup_s, hue=IPUA_Labels, palette=palette)[20]: <seaborn.axisgrid.PairGrid at 0x18c9bc3cbc8>

8 Merge the DBSCAN labels for IP/US GroupBy with original df8.1 Remember, so far we have been looking at the group by DataFrameNow that we merged the DBSCAN Lables into the raw date, we will see different cardinality then our groupby dataframes. This is because many rows from the origional DataFrame can have the same IP and User Agent[21]: df = botDataSet.df#Note: This will be an inner joindf_IPUA_clusters = pd.merge(df, df_IPUA_agg[IPUA_Labels], on=IPUASessionID)[22]: p= pd.DataFrame(df_IPUA_clusters.groupby([IPUA_Labels])[InboxID].count()) p = p.unstack(0, ) p = p.reset_index(IPUA_Labels)p.columns = [IPUA_Labels, RequestCount]p.plot.bar(IPUA_Labels, RequestCount, figsize=(12,5))[22]: <matplotlib.axes._subplots.AxesSubplot at 0x18c9f540a88>

[23]: IPUA_Labels 0 1 2 3InboxID 10154 1409 2 10Company 1 1 1 1List 1 1 1 1Campaign 1 1 1 1Link 37 12 27 37IPAddress 10327 28 2 2UserAgent 1177 1 2 2AS Name 21 1 1 2AS Number 22 1 1 2CIDR Range 353 11 2 2EmailDomain 590 361 2 8EmailRootLevelDomain 29 9 1 28.2 Show distribution of these labels across the AS Names[24]: def groupByPivot(group_column_top, group_column_left, count_column): gb = df_IPUA_clusters.groupby([group_column_top,,group_column_left])[count_column].count() gb.fillna(0, inplace=True) gb = gb.unstack(0)gb.fillna(0, inplace=True)gb[All] = gb[0] + gb[1] + gb[2] + gb[3] #+ gb[4] gb.sort_values(All, ascending=False, inplace=True) return gb[25]: groupByPivot(IPUA_Labels, AS Name,InboxID)[25]: IPUA_Labels AS Name 0 1 2 Amazon.com, Inc. 1158.0 9608.0 40.0Unknown 10562.0 0.0 0.0AT&T Mobility LLC 1118.0 0.0 0.0MCI Communications Services, Inc. Verizon Business 294.0 0.0 0.0T-Mobile USA, Inc. 253.0 0.0 0.0Scalair SAS 217.0 0.0 0.0CenturyLink Communications, LLC 91.0 0.0 0.0Google LLC 86.0 0.0 0.0Comcast Cable Communications, LLC 81.0 0.0 0.0DigitalOcean, LLC 20.0 0.0 0.0Level 3 Parent, LLC 54.0 0.0 0.0Windstream Communications LLC 31.0 0.0 0.0TELUS Communications Inc. 27.0 0.0 0.0Shaw Communications Inc. 16.0 0.0 0.0Microsoft Corporation 12.0 0.0 0.0GENESCO INC 6.0 0.0 0.0Johns Hopkins University 4.0 0.0 0.0DoD Network Information Center 3.0 0.0 0.0Sprint 3.0 0.0 0.0Cellco Partnership DBA Verizon Wireless 3.0 0.0 0.0Cogent Communications 2.0 0.0 0.0IPUA_Labels AS Name 3 AllAmazon.com, Inc. 333.0 11139.0Unknown 0.0 10562.0AT&T Mobility LLC 0.0 1118.0MCI Communications Services, Inc. Verizon Business 0.0 294.0T-Mobile USA, Inc. 0.0 253.0Scalair SAS 0.0 217.0CenturyLink Communications, LLC 0.0 91.0Google LLC 0.0 86.0Comcast Cable Communications, LLC 0.0 81.0DigitalOcean, LLC 36.0 56.0Level 3 Parent, LLC 0.0 54.0Windstream Communications LLC 0.0 31.0TELUS Communications Inc. 0.0 27.0Shaw Communications Inc. 0.0 16.0Microsoft Corporation 0.0 12.0GENESCO INC 0.0 6.0Johns Hopkins University 0.0 4.0DoD Network Information Center 0.0 3.0Sprint 0.0 3.0Cellco Partnership DBA Verizon Wireless 0.0 3.0Cogent Communications[26]: groupByPivot(IPUA_Labels, AS Number,InboxID)[26]: IPUA_Labels 0 1 2 3 AllAS NumberUnknown 10562.0 0.0 0.0 0.0 10562.0 16509 261.0 9608.0 40.0 0.0 9909.014618 897.0 0.0 0.0 333.0 1230.0 20057 1118.0 0.0 0.0 0.0 1118.0701 294.0 0.0 0.0 0.0 294.021928 253.0 0.0 0.0 0.0 253.0206002 217.0 0.0 0.0 0.0 217.0209 91.0 0.0 0.0 0.0 91.015169 86.0 0.0 0.0 0.0 86.07922 81.0 0.0 0.0 0.0 81.014061 20.0 0.0 0.0 36.0 56.03356 54.0 0.0 0.0 0.0 54.07029 31.0 0.0 0.0 0.0 31.0852 27.0 0.0 0.0 0.0 27.0 0.0 2.0

6327 16.0 0.0 0.0 0.0 16.08075 12.0 0.0 0.0 0.0 12.0 20242 6.0 0.0 0.0 0.0 6.05723 4.0 0.0 0.0 0.0 4.06167 3.0 0.0 0.0 0.0 3.0721 3.0 0.0 0.0 0.0 3.01239 3.0 0.0 0.0 0.0 3.0174 2.0 0.0 0.0 0.0 2.0[27]: groupByPivot(IPUA_Labels, CIDR Range,InboxID).head(20)[27]: IPUA_Labels CIDR Range 0 1 2 3 AllUnknown 10562.0 0.0 0.0 0.0 10562.0{4ED21F0C-8573-052A-670B-E4D315133AF5} 75.0 3823.0 0.0 0.0 3898.0{7EF79BFD-BC3A-903D-C2BD-F5E295ACF4DE} 18.0 1731.0 0.0 0.0 1749.0{E3615A90-F5C2-C4A8-40A1-2CCC4102B390} 7.0 812.0 0.0 0.0 819.0{D2185CAC-6E67-82FB-603B-F5B33514B577} 9.0 781.0 0.0 0.0 790.0{F9164193-183A-7A69-1AE7-25BD6243ED72} 9.0 596.0 0.0 0.0 605.0{11E5AB29-4623-8156-AB63-E7A2FD82A79A} 6.0 584.0 0.0 0.0 590.0{60D94A4C-7F09-D0F9-2810-83B93ACBC03E} 12.0 438.0 0.0 0.0 450.0{7E737FAE-7868-5BE1-84B6-EA9011F3B32E} 33.0 0.0 0.0 333.0 366.0{6194D26B-CF4C-CA86-F018-CDE54FED8393} 291.0 0.0 0.0 0.0 291.0{855313BB-3482-1ADA-36A8-5D073B29C265} 2.0 235.0 20.0 0.0 257.0{6893DEE9-A105-29E9-F9D4-98B3411CF42B} 4.0 249.0 0.0 0.0 253.0{BF61FA52-8010-2731-2B4E-0C77A6C05E6A} 0.0 222.0 0.0 0.0 222.0{EEA776DA-8B86-986D-AAD8-0A4A169106A4} 217.0 0.0 0.0 0.0 217.0{9DA7FFE3-A644-988D-A6E7-0D6305395B19} 216.0 0.0 0.0 0.0 216.0{71EC7D71-D378-E92E-CE68-0212EA13E3FC} 192.0 0.0 0.0 0.0 192.0{98A8C5B5-B34B-963A-3240-EE64FE8394FD} 184.0 0.0 0.0 0.0 184.0{4D1C1E09-9D9F-8960-FEA5-FA478BC0CE02} 1.0 137.0 20.0 0.0 158.0{801AB1BE-FFF2-B6EF-D3FA-8768D3109CD4} 120.0 0.0 0.0 0.0 120.0{69CC6342-B73F-0C05-D298-B7EE23E2217E} 114.0 0.0 0.0 0.0 114.09 10 Looking at the raw data in SQL, we deteremined: Label 0 Organic click activity Label 1 BOT activity from AWS Oregon Label 2 BOT activity from AWS Oregon Label 3 BOT activity from AWS Virginia11 12 Extract Results12.0.1 Now that we have named the unsupervised labels, we can make our conclusions

14 IPAddress/UserAgent Reputation[28]: df_IPUA_rep = pd.DataFrame(df_IPUA_clusters.groupby([IPAddress, UserAgent,,IPUA_Labels])[InboxID].count()) df_IPUA_rep = df_IPUA_rep.unstack(IPUA_Labels).fillna(0) df_IPUA_rep.columns = df_IPUA_rep.columns.droplevel() df_IPUA_rep[TotalCount] = df_IPUA_rep[0] + df_IPUA_rep[1] + df_IPUA_rep[2] +,df_IPUA_rep[3] df_IPUA_rep[BotCount] = df_IPUA_rep[1] + df_IPUA_rep[2] + df_IPUA_rep[3] df_IPUA_rep[ReputationScore] = df_IPUA_rep[BotCount] /,df_IPUA_rep[TotalCount] df_IPUA_rep.sort_values(ReputationScore, ascending=False)[28]: IPUA_Labels0 IPAddress UserAgent{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}0.0 {55EC17BD-287C-2F25-173A-298E9A4FEE24} {F87DC183-2EAD-5AED-B536-4545C440750D}1.0{55EC7CAB-1C40-F94C-7A67-87C6859B96E6} {CBBC6612-3713-CBB1-C09F-0E774A647016}1.0{55F5ECBC-468A-ACE5-B1A6-B3A78D6C7B6C} {F87DC183-2EAD-5AED-B536-4545C440750D}1.0{55FAAD1E-483A-5850-8CE4-C33C6096E326} {F87DC183-2EAD-5AED-B536-4545C440750D}1.0{FFF90752-2F9B-BAF1-E41F-2C9A924C70D2} {F87DC183-2EAD-5AED-B536-4545C440750D}1.0IPUA_Labels1 IPAddress UserAgent{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}143.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}145.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}222.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}76.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}0.0 {55EC17BD-287C-2F25-173A-298E9A4FEE24} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55EC7CAB-1C40-F94C-7A67-87C6859B96E6} {CBBC6612-3713-CBB1-C09F-0E774A647016}0.0{55F5ECBC-468A-ACE5-B1A6-B3A78D6C7B6C} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55FAAD1E-483A-5850-8CE4-C33C6096E326} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{FFF90752-2F9B-BAF1-E41F-2C9A924C70D2} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0IPUA_Labels2 IPAddress UserAgent{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}20.0 {55EC17BD-287C-2F25-173A-298E9A4FEE24} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55EC7CAB-1C40-F94C-7A67-87C6859B96E6} {CBBC6612-3713-CBB1-C09F-0E774A647016} 0.0{55F5ECBC-468A-ACE5-B1A6-B3A78D6C7B6C} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55FAAD1E-483A-5850-8CE4-C33C6096E326} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{FFF90752-2F9B-BAF1-E41F-2C9A924C70D2} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0IPUA_Labels3 IPAddress UserAgent{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}0.0 {55EC17BD-287C-2F25-173A-298E9A4FEE24} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55EC7CAB-1C40-F94C-7A67-87C6859B96E6} {CBBC6612-3713-CBB1-C09F-0E774A647016}0.0{55F5ECBC-468A-ACE5-B1A6-B3A78D6C7B6C} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55FAAD1E-483A-5850-8CE4-C33C6096E326} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{FFF90752-2F9B-BAF1-E41F-2C9A924C70D2} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0IPUA_LabelsTotalCount IPAddress UserAgent{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}143.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}145.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}222.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}76.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}20.0{55EC17BD-287C-2F25-173A-298E9A4FEE24} {F87DC183-2EAD-5AED-B536-4545C440750D}1.0{55EC7CAB-1C40-F94C-7A67-87C6859B96E6} {CBBC6612-3713-CBB1-C09F-0E774A647016}1.0{55F5ECBC-468A-ACE5-B1A6-B3A78D6C7B6C} {F87DC183-2EAD-5AED-B536-4545C440750D}1.0{55FAAD1E-483A-5850-8CE4-C33C6096E326} {F87DC183-2EAD-5AED-B536-4545C440750D}1.0{FFF90752-2F9B-BAF1-E41F-2C9A924C70D2} {F87DC183-2EAD-5AED-B536-4545C440750D}1.0IPUA_LabelsBotCount IPAddress UserAgent{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}143.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}145.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}222.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}76.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}20.0 {55EC17BD-287C-2F25-173A-298E9A4FEE24} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55EC7CAB-1C40-F94C-7A67-87C6859B96E6} {CBBC6612-3713-CBB1-C09F-0E774A647016}0.0{55F5ECBC-468A-ACE5-B1A6-B3A78D6C7B6C} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55FAAD1E-483A-5850-8CE4-C33C6096E326} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{FFF90752-2F9B-BAF1-E41F-2C9A924C70D2} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0IPUA_LabelsReputationScoreIPAddress UserAgent{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}1.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}1.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9} 1.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}1.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}1.0 {55EC17BD-287C-2F25-173A-298E9A4FEE24} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55EC7CAB-1C40-F94C-7A67-87C6859B96E6} {CBBC6612-3713-CBB1-C09F-0E774A647016}0.0{55F5ECBC-468A-ACE5-B1A6-B3A78D6C7B6C} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{55FAAD1E-483A-5850-8CE4-C33C6096E326} {F87DC183-2EAD-5AED-B536-4545C440750D}0.0{FFF90752-2F9B-BAF1-E41F-2C9A924C70D2} {F87DC183-2EAD-5AED-B536-4545C440750D} 0.0[10688 rows x 7 columns]15 CIDR ReputationSince most BOT requests will use different IP Addresses for their HTTP Requests, we want to score the reputation of the CIDR Range.This might be too aggressive if the CIDR Range is used by multiple accounts, i.e. NAT belongs to AWS and is used by mutliple accounts.[29]: df_CIDR_rep = pd.DataFrame(df_IPUA_clusters.groupby([CIDR Range,,IPUA_Labels])[InboxID].count()) df_CIDR_rep = df_CIDR_rep.unstack(IPUA_Labels).fillna(0) df_CIDR_rep.columns = df_CIDR_rep.columns.droplevel() df_CIDR_rep[TotalCount] = df_CIDR_rep[0] + df_CIDR_rep[1] + df_CIDR_rep[2] +,df_CIDR_rep[3] df_CIDR_rep[BotCount] = df_CIDR_rep[1] + df_CIDR_rep[2] + df_CIDR_rep[3] df_CIDR_rep[ReputationScore] = df_CIDR_rep[BotCount] /,df_CIDR_rep[TotalCount] df_CIDR_rep.sort_values(ReputationScore, ascending=False)[29]: IPUA_Labels 0 1 2 3 TotalCount CIDR Range{BF61FA52-8010-2731-2B4E-0C77A6C05E6A} 0.0 222.0 0.0 0.0 222.0{4D1C1E09-9D9F-8960-FEA5-FA478BC0CE02} 1.0 137.0 20.0 0.0 158.0{855313BB-3482-1ADA-36A8-5D073B29C265} 2.0 235.0 20.0 0.0 257.0{E3615A90-F5C2-C4A8-40A1-2CCC4102B390} 7.0 812.0 0.0 0.0 819.0{11E5AB29-4623-8156-AB63-E7A2FD82A79A} 6.0 584.0 0.0 0.0 590.0{53C74F82-9196-8077-6415-809E96574400} 1.0 0.0 0.0 0.0 1.0{536F1AD9-7CB0-BD49-C927-CD9A7895DC2F} 1.0 0.0 0.0 0.0 1.0{5352877C-27D3-B09C-D4EA-55295EEC1358} 1.0 0.0 0.0 0.0 1.0{524E0F39-04BE-5781-E6FD-F38096077136} 1.0 0.0 0.0 0.0 1.0{FEF1369D-B9BF-3EB1-2C8C-30C548F1EC07} 2.0 0.0 0.0 0.0 2.0IPUA_Labels CIDR Range BotCount ReputationScore{BF61FA52-8010-2731-2B4E-0C77A6C05E6A} 222.0 1.000000{4D1C1E09-9D9F-8960-FEA5-FA478BC0CE02} 157.0 0.993671{855313BB-3482-1ADA-36A8-5D073B29C265} 255.0 0.992218{E3615A90-F5C2-C4A8-40A1-2CCC4102B390} 812.0 0.991453{11E5AB29-4623-8156-AB63-E7A2FD82A79A} 584.0 0.989831 {53C74F82-9196-8077-6415-809E96574400} 0.0 0.000000{536F1AD9-7CB0-BD49-C927-CD9A7895DC2F} 0.0 0.000000{5352877C-27D3-B09C-D4EA-55295EEC1358} 0.0 0.000000{524E0F39-04BE-5781-E6FD-F38096077136} 0.0 0.000000{FEF1369D-B9BF-3EB1-2C8C-30C548F1EC07} 0.0 0.000000[354 rows x 7 columns]15.0.1 Save lables to csv file for manual analysis with raw non-anonymized data in SQL.[30]: output = df_IPUA_rep[df_IPUA_rep[ReputationScore] > 0] output.to_csv(C:/data/IPUA_Reputation_Scores.csv) output[30]: IPUA_Labels0 IPAddress UserAgent{0793D471-B890-C0C1-2C0B-DF4882B1D58C} {30634283-D150-237A-87FD-DA097EBBE5B9}3.0{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{19547553-CF45-F7D9-4A02-394BEED09DC4} {30634283-D150-237A-87FD-DA097EBBE5B9}5.0{231BEBFF-35AE-DBE5-0B5B-FC948A6B647C} {30634283-D150-237A-87FD-DA097EBBE5B9}1.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}0.0{41D3D20E-D0C2-80D5-90F7-1DFE369F08E8} {30634283-D150-237A-87FD-DA097EBBE5B9}2.0{45EA4474-3101-DAE1-F2C7-AB5DCBDC5C9E} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{47C10F28-63E1-6830-A065-8378B6E54BB9} {30634283-D150-237A-87FD-DA097EBBE5B9} 3.0{49F7F9A3-19F9-6E8A-5896-BC10581E5868} {30634283-D150-237A-87FD-DA097EBBE5B9}1.0{4ADEB3FD-B967-E72B-B0C7-CDA39EA744F0} {30634283-D150-237A-87FD-DA097EBBE5B9}2.0{500B9043-A0ED-989D-B645-5B635BAE6841} {30634283-D150-237A-87FD-DA097EBBE5B9}6.0{543DBB24-4FA1-DA0B-E6FC-6E8BB392C8BF} {30634283-D150-237A-87FD-DA097EBBE5B9}1.0{663BEEA8-FAB0-8223-D501-75488AA44C8B} {30634283-D150-237A-87FD-DA097EBBE5B9}4.0{692FDBEC-0D93-2758-D331-CE66EB83FE72} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{8B00AE4B-9328-B1DC-57BA-17B7D11AFE9F} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{907BAFFD-A66B-2D34-89B1-768E3FCBD951} {30634283-D150-237A-87FD-DA097EBBE5B9}6.0{9CBE8C71-0131-336C-5211-6EE9060D8B10} {199313F3-4901-ACA6-FFBA-D8C868D9FA00}0.0{9EA619AE-3EC0-3405-1CC9-282C8FBFBDC2} {30634283-D150-237A-87FD-DA097EBBE5B9}2.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{BBB7787E-A8EA-C6AF-4BBD-DBF37A23230A} {30634283-D150-237A-87FD-DA097EBBE5B9}3.0{C39D9FAE-0933-7C52-B6C0-837E5BD72598} {30634283-D150-237A-87FD-DA097EBBE5B9}2.0{D418BA33-88B2-4F10-0B5B-878B938CABF5} {30634283-D150-237A-87FD-DA097EBBE5B9}4.0{DA8A5BA2-BCB6-0C4D-B449-DCD5EF31D732} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{DBE066F2-9F22-A9FC-9B3F-11CE933DB934} {75570E27-930A-E412-95FA-B6F033F3988E}0.0{E11DF7EC-7BAF-593B-C8F7-B84EC4255F66} {30634283-D150-237A-87FD-DA097EBBE5B9}4.0{E16B4100-C6C6-E5C7-4411-02AD0C4B2D88} {59E0B8D9-09F2-ADC8-9686-9015A09FB3B8}0.0{E8D9A46E-8D4D-FAE6-F845-2D16E4CD8720} {30634283-D150-237A-87FD-DA097EBBE5B9}4.0{F27685B7-6847-BAEE-2549-CF30545EF54E} {30634283-D150-237A-87FD-DA097EBBE5B9}3.0{F65478DB-6A17-9059-1B5A-E144A44D06F4} {30634283-D150-237A-87FD-DA097EBBE5B9}1.0{FC7FE18F-0F08-55F9-E6BC-67CB932298EE} {30634283-D150-237A-87FD-DA097EBBE5B9}1.0IPUA_Labels1 IPAddress UserAgent{0793D471-B890-C0C1-2C0B-DF4882B1D58C} {30634283-D150-237A-87FD-DA097EBBE5B9}512.0{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}143.0{19547553-CF45-F7D9-4A02-394BEED09DC4} {30634283-D150-237A-87FD-DA097EBBE5B9}604.0{231BEBFF-35AE-DBE5-0B5B-FC948A6B647C} {30634283-D150-237A-87FD-DA097EBBE5B9}192.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}0.0{41D3D20E-D0C2-80D5-90F7-1DFE369F08E8} {30634283-D150-237A-87FD-DA097EBBE5B9}636.0{45EA4474-3101-DAE1-F2C7-AB5DCBDC5C9E} {30634283-D150-237A-87FD-DA097EBBE5B9}67.0{47C10F28-63E1-6830-A065-8378B6E54BB9} {30634283-D150-237A-87FD-DA097EBBE5B9}290.0{49F7F9A3-19F9-6E8A-5896-BC10581E5868} {30634283-D150-237A-87FD-DA097EBBE5B9}313.0{4ADEB3FD-B967-E72B-B0C7-CDA39EA744F0} {30634283-D150-237A-87FD-DA097EBBE5B9}577.0{500B9043-A0ED-989D-B645-5B635BAE6841} {30634283-D150-237A-87FD-DA097EBBE5B9}764.0{543DBB24-4FA1-DA0B-E6FC-6E8BB392C8BF} {30634283-D150-237A-87FD-DA097EBBE5B9}235.0{663BEEA8-FAB0-8223-D501-75488AA44C8B} {30634283-D150-237A-87FD-DA097EBBE5B9}305.0{692FDBEC-0D93-2758-D331-CE66EB83FE72} {30634283-D150-237A-87FD-DA097EBBE5B9}91.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}222.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}145.0{8B00AE4B-9328-B1DC-57BA-17B7D11AFE9F} {30634283-D150-237A-87FD-DA097EBBE5B9}123.0{907BAFFD-A66B-2D34-89B1-768E3FCBD951} {30634283-D150-237A-87FD-DA097EBBE5B9}770.0{9CBE8C71-0131-336C-5211-6EE9060D8B10} {199313F3-4901-ACA6-FFBA-D8C868D9FA00}0.0{9EA619AE-3EC0-3405-1CC9-282C8FBFBDC2} {30634283-D150-237A-87FD-DA097EBBE5B9}513.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9} 76.0{BBB7787E-A8EA-C6AF-4BBD-DBF37A23230A} {30634283-D150-237A-87FD-DA097EBBE5B9}604.0{C39D9FAE-0933-7C52-B6C0-837E5BD72598} {30634283-D150-237A-87FD-DA097EBBE5B9}215.0{D418BA33-88B2-4F10-0B5B-878B938CABF5} {30634283-D150-237A-87FD-DA097EBBE5B9}249.0{DA8A5BA2-BCB6-0C4D-B449-DCD5EF31D732} {30634283-D150-237A-87FD-DA097EBBE5B9}133.0{DBE066F2-9F22-A9FC-9B3F-11CE933DB934} {75570E27-930A-E412-95FA-B6F033F3988E}0.0{E11DF7EC-7BAF-593B-C8F7-B84EC4255F66} {30634283-D150-237A-87FD-DA097EBBE5B9}438.0{E16B4100-C6C6-E5C7-4411-02AD0C4B2D88} {59E0B8D9-09F2-ADC8-9686-9015A09FB3B8}0.0{E8D9A46E-8D4D-FAE6-F845-2D16E4CD8720} {30634283-D150-237A-87FD-DA097EBBE5B9}522.0{F27685B7-6847-BAEE-2549-CF30545EF54E} {30634283-D150-237A-87FD-DA097EBBE5B9}584.0{F65478DB-6A17-9059-1B5A-E144A44D06F4} {30634283-D150-237A-87FD-DA097EBBE5B9}215.0{FC7FE18F-0F08-55F9-E6BC-67CB932298EE} {30634283-D150-237A-87FD-DA097EBBE5B9}70.0IPUA_Labels2 IPAddress UserAgent{0793D471-B890-C0C1-2C0B-DF4882B1D58C} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{19547553-CF45-F7D9-4A02-394BEED09DC4} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{231BEBFF-35AE-DBE5-0B5B-FC948A6B647C} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}20.0{41D3D20E-D0C2-80D5-90F7-1DFE369F08E8} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{45EA4474-3101-DAE1-F2C7-AB5DCBDC5C9E} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{47C10F28-63E1-6830-A065-8378B6E54BB9} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{49F7F9A3-19F9-6E8A-5896-BC10581E5868} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{4ADEB3FD-B967-E72B-B0C7-CDA39EA744F0} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{500B9043-A0ED-989D-B645-5B635BAE6841} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{543DBB24-4FA1-DA0B-E6FC-6E8BB392C8BF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{663BEEA8-FAB0-8223-D501-75488AA44C8B} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{692FDBEC-0D93-2758-D331-CE66EB83FE72} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{8B00AE4B-9328-B1DC-57BA-17B7D11AFE9F} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{907BAFFD-A66B-2D34-89B1-768E3FCBD951} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{9CBE8C71-0131-336C-5211-6EE9060D8B10} {199313F3-4901-ACA6-FFBA-D8C868D9FA00}0.0{9EA619AE-3EC0-3405-1CC9-282C8FBFBDC2} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{BBB7787E-A8EA-C6AF-4BBD-DBF37A23230A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{C39D9FAE-0933-7C52-B6C0-837E5BD72598} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{D418BA33-88B2-4F10-0B5B-878B938CABF5} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{DA8A5BA2-BCB6-0C4D-B449-DCD5EF31D732} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{DBE066F2-9F22-A9FC-9B3F-11CE933DB934} {75570E27-930A-E412-95FA-B6F033F3988E}20.0{E11DF7EC-7BAF-593B-C8F7-B84EC4255F66} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{E16B4100-C6C6-E5C7-4411-02AD0C4B2D88} {59E0B8D9-09F2-ADC8-9686-9015A09FB3B8}0.0{E8D9A46E-8D4D-FAE6-F845-2D16E4CD8720} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{F27685B7-6847-BAEE-2549-CF30545EF54E} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{F65478DB-6A17-9059-1B5A-E144A44D06F4} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{FC7FE18F-0F08-55F9-E6BC-67CB932298EE} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0IPUA_Labels3 IPAddress UserAgent {0793D471-B890-C0C1-2C0B-DF4882B1D58C} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{19547553-CF45-F7D9-4A02-394BEED09DC4} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{231BEBFF-35AE-DBE5-0B5B-FC948A6B647C} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}0.0{41D3D20E-D0C2-80D5-90F7-1DFE369F08E8} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{45EA4474-3101-DAE1-F2C7-AB5DCBDC5C9E} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{47C10F28-63E1-6830-A065-8378B6E54BB9} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{49F7F9A3-19F9-6E8A-5896-BC10581E5868} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{4ADEB3FD-B967-E72B-B0C7-CDA39EA744F0} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{500B9043-A0ED-989D-B645-5B635BAE6841} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{543DBB24-4FA1-DA0B-E6FC-6E8BB392C8BF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{663BEEA8-FAB0-8223-D501-75488AA44C8B} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{692FDBEC-0D93-2758-D331-CE66EB83FE72} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{8B00AE4B-9328-B1DC-57BA-17B7D11AFE9F} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{907BAFFD-A66B-2D34-89B1-768E3FCBD951} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{9CBE8C71-0131-336C-5211-6EE9060D8B10} {199313F3-4901-ACA6-FFBA-D8C868D9FA00}36.0{9EA619AE-3EC0-3405-1CC9-282C8FBFBDC2} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{BBB7787E-A8EA-C6AF-4BBD-DBF37A23230A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{C39D9FAE-0933-7C52-B6C0-837E5BD72598} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{D418BA33-88B2-4F10-0B5B-878B938CABF5} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{DA8A5BA2-BCB6-0C4D-B449-DCD5EF31D732} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{DBE066F2-9F22-A9FC-9B3F-11CE933DB934} {75570E27-930A-E412-95FA-B6F033F3988E}0.0{E11DF7EC-7BAF-593B-C8F7-B84EC4255F66} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{E16B4100-C6C6-E5C7-4411-02AD0C4B2D88} {59E0B8D9-09F2-ADC8-9686-9015A09FB3B8}333.0{E8D9A46E-8D4D-FAE6-F845-2D16E4CD8720} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{F27685B7-6847-BAEE-2549-CF30545EF54E} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{F65478DB-6A17-9059-1B5A-E144A44D06F4} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0{FC7FE18F-0F08-55F9-E6BC-67CB932298EE} {30634283-D150-237A-87FD-DA097EBBE5B9}0.0IPUA_LabelsTotalCount IPAddress UserAgent{0793D471-B890-C0C1-2C0B-DF4882B1D58C} {30634283-D150-237A-87FD-DA097EBBE5B9}515.0{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}143.0{19547553-CF45-F7D9-4A02-394BEED09DC4} {30634283-D150-237A-87FD-DA097EBBE5B9}609.0{231BEBFF-35AE-DBE5-0B5B-FC948A6B647C} {30634283-D150-237A-87FD-DA097EBBE5B9}193.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}20.0{41D3D20E-D0C2-80D5-90F7-1DFE369F08E8} {30634283-D150-237A-87FD-DA097EBBE5B9}638.0{45EA4474-3101-DAE1-F2C7-AB5DCBDC5C9E} {30634283-D150-237A-87FD-DA097EBBE5B9}67.0{47C10F28-63E1-6830-A065-8378B6E54BB9} {30634283-D150-237A-87FD-DA097EBBE5B9}293.0{49F7F9A3-19F9-6E8A-5896-BC10581E5868} {30634283-D150-237A-87FD-DA097EBBE5B9}314.0{4ADEB3FD-B967-E72B-B0C7-CDA39EA744F0} {30634283-D150-237A-87FD-DA097EBBE5B9}579.0{500B9043-A0ED-989D-B645-5B635BAE6841} {30634283-D150-237A-87FD-DA097EBBE5B9}770.0{543DBB24-4FA1-DA0B-E6FC-6E8BB392C8BF} {30634283-D150-237A-87FD-DA097EBBE5B9}236.0{663BEEA8-FAB0-8223-D501-75488AA44C8B} {30634283-D150-237A-87FD-DA097EBBE5B9} 309.0{692FDBEC-0D93-2758-D331-CE66EB83FE72} {30634283-D150-237A-87FD-DA097EBBE5B9}91.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}222.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}145.0{8B00AE4B-9328-B1DC-57BA-17B7D11AFE9F} {30634283-D150-237A-87FD-DA097EBBE5B9}123.0{907BAFFD-A66B-2D34-89B1-768E3FCBD951} {30634283-D150-237A-87FD-DA097EBBE5B9}776.0{9CBE8C71-0131-336C-5211-6EE9060D8B10} {199313F3-4901-ACA6-FFBA-D8C868D9FA00}36.0{9EA619AE-3EC0-3405-1CC9-282C8FBFBDC2} {30634283-D150-237A-87FD-DA097EBBE5B9}515.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}76.0{BBB7787E-A8EA-C6AF-4BBD-DBF37A23230A} {30634283-D150-237A-87FD-DA097EBBE5B9}607.0{C39D9FAE-0933-7C52-B6C0-837E5BD72598} {30634283-D150-237A-87FD-DA097EBBE5B9}217.0{D418BA33-88B2-4F10-0B5B-878B938CABF5} {30634283-D150-237A-87FD-DA097EBBE5B9}253.0{DA8A5BA2-BCB6-0C4D-B449-DCD5EF31D732} {30634283-D150-237A-87FD-DA097EBBE5B9}133.0{DBE066F2-9F22-A9FC-9B3F-11CE933DB934} {75570E27-930A-E412-95FA-B6F033F3988E}20.0{E11DF7EC-7BAF-593B-C8F7-B84EC4255F66} {30634283-D150-237A-87FD-DA097EBBE5B9}442.0{E16B4100-C6C6-E5C7-4411-02AD0C4B2D88} {59E0B8D9-09F2-ADC8-9686-9015A09FB3B8}333.0{E8D9A46E-8D4D-FAE6-F845-2D16E4CD8720} {30634283-D150-237A-87FD-DA097EBBE5B9}526.0{F27685B7-6847-BAEE-2549-CF30545EF54E} {30634283-D150-237A-87FD-DA097EBBE5B9}587.0{F65478DB-6A17-9059-1B5A-E144A44D06F4} {30634283-D150-237A-87FD-DA097EBBE5B9}216.0{FC7FE18F-0F08-55F9-E6BC-67CB932298EE} {30634283-D150-237A-87FD-DA097EBBE5B9}71.0IPUA_LabelsBotCount IPAddress UserAgent{0793D471-B890-C0C1-2C0B-DF4882B1D58C} {30634283-D150-237A-87FD-DA097EBBE5B9}512.0{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}143.0{19547553-CF45-F7D9-4A02-394BEED09DC4} {30634283-D150-237A-87FD-DA097EBBE5B9} 604.0{231BEBFF-35AE-DBE5-0B5B-FC948A6B647C} {30634283-D150-237A-87FD-DA097EBBE5B9}192.0{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}20.0{41D3D20E-D0C2-80D5-90F7-1DFE369F08E8} {30634283-D150-237A-87FD-DA097EBBE5B9}636.0{45EA4474-3101-DAE1-F2C7-AB5DCBDC5C9E} {30634283-D150-237A-87FD-DA097EBBE5B9}67.0{47C10F28-63E1-6830-A065-8378B6E54BB9} {30634283-D150-237A-87FD-DA097EBBE5B9}290.0{49F7F9A3-19F9-6E8A-5896-BC10581E5868} {30634283-D150-237A-87FD-DA097EBBE5B9}313.0{4ADEB3FD-B967-E72B-B0C7-CDA39EA744F0} {30634283-D150-237A-87FD-DA097EBBE5B9}577.0{500B9043-A0ED-989D-B645-5B635BAE6841} {30634283-D150-237A-87FD-DA097EBBE5B9}764.0{543DBB24-4FA1-DA0B-E6FC-6E8BB392C8BF} {30634283-D150-237A-87FD-DA097EBBE5B9}235.0{663BEEA8-FAB0-8223-D501-75488AA44C8B} {30634283-D150-237A-87FD-DA097EBBE5B9}305.0{692FDBEC-0D93-2758-D331-CE66EB83FE72} {30634283-D150-237A-87FD-DA097EBBE5B9}91.0{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}222.0{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}145.0{8B00AE4B-9328-B1DC-57BA-17B7D11AFE9F} {30634283-D150-237A-87FD-DA097EBBE5B9}123.0{907BAFFD-A66B-2D34-89B1-768E3FCBD951} {30634283-D150-237A-87FD-DA097EBBE5B9}770.0{9CBE8C71-0131-336C-5211-6EE9060D8B10} {199313F3-4901-ACA6-FFBA-D8C868D9FA00}36.0{9EA619AE-3EC0-3405-1CC9-282C8FBFBDC2} {30634283-D150-237A-87FD-DA097EBBE5B9}513.0{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}76.0{BBB7787E-A8EA-C6AF-4BBD-DBF37A23230A} {30634283-D150-237A-87FD-DA097EBBE5B9}604.0{C39D9FAE-0933-7C52-B6C0-837E5BD72598} {30634283-D150-237A-87FD-DA097EBBE5B9}215.0{D418BA33-88B2-4F10-0B5B-878B938CABF5} {30634283-D150-237A-87FD-DA097EBBE5B9}249.0{DA8A5BA2-BCB6-0C4D-B449-DCD5EF31D732} {30634283-D150-237A-87FD-DA097EBBE5B9}133.0{DBE066F2-9F22-A9FC-9B3F-11CE933DB934} {75570E27-930A-E412-95FA-B6F033F3988E} 20.0{E11DF7EC-7BAF-593B-C8F7-B84EC4255F66} {30634283-D150-237A-87FD-DA097EBBE5B9}438.0{E16B4100-C6C6-E5C7-4411-02AD0C4B2D88} {59E0B8D9-09F2-ADC8-9686-9015A09FB3B8}333.0{E8D9A46E-8D4D-FAE6-F845-2D16E4CD8720} {30634283-D150-237A-87FD-DA097EBBE5B9}522.0{F27685B7-6847-BAEE-2549-CF30545EF54E} {30634283-D150-237A-87FD-DA097EBBE5B9}584.0{F65478DB-6A17-9059-1B5A-E144A44D06F4} {30634283-D150-237A-87FD-DA097EBBE5B9}215.0{FC7FE18F-0F08-55F9-E6BC-67CB932298EE} {30634283-D150-237A-87FD-DA097EBBE5B9}70.0IPUA_LabelsReputationScoreIPAddress UserAgent{0793D471-B890-C0C1-2C0B-DF4882B1D58C} {30634283-D150-237A-87FD-DA097EBBE5B9}0.994175{1555FBA3-2FF0-D329-AC22-154C821082CF} {30634283-D150-237A-87FD-DA097EBBE5B9}1.000000{19547553-CF45-F7D9-4A02-394BEED09DC4} {30634283-D150-237A-87FD-DA097EBBE5B9}0.991790{231BEBFF-35AE-DBE5-0B5B-FC948A6B647C} {30634283-D150-237A-87FD-DA097EBBE5B9}0.994819{2EDE7EB8-4F33-998A-C2B0-464296C61AC0} {5437D43E-47AD-69BD-A1D7-D310B07D7D5A}1.000000{41D3D20E-D0C2-80D5-90F7-1DFE369F08E8} {30634283-D150-237A-87FD-DA097EBBE5B9}0.996865{45EA4474-3101-DAE1-F2C7-AB5DCBDC5C9E} {30634283-D150-237A-87FD-DA097EBBE5B9}1.000000{47C10F28-63E1-6830-A065-8378B6E54BB9} {30634283-D150-237A-87FD-DA097EBBE5B9}0.989761{49F7F9A3-19F9-6E8A-5896-BC10581E5868} {30634283-D150-237A-87FD-DA097EBBE5B9}0.996815{4ADEB3FD-B967-E72B-B0C7-CDA39EA744F0} {30634283-D150-237A-87FD-DA097EBBE5B9}0.996546{500B9043-A0ED-989D-B645-5B635BAE6841} {30634283-D150-237A-87FD-DA097EBBE5B9}0.992208{543DBB24-4FA1-DA0B-E6FC-6E8BB392C8BF} {30634283-D150-237A-87FD-DA097EBBE5B9}0.995763{663BEEA8-FAB0-8223-D501-75488AA44C8B} {30634283-D150-237A-87FD-DA097EBBE5B9}0.987055{692FDBEC-0D93-2758-D331-CE66EB83FE72} {30634283-D150-237A-87FD-DA097EBBE5B9}1.000000{6AC33884-ADE3-2F3E-6DD7-1731EC5D2945} {30634283-D150-237A-87FD-DA097EBBE5B9}1.000000{810FA2C8-982B-6FEE-C299-5CA6CDE57F3D} {30634283-D150-237A-87FD-DA097EBBE5B9}1.000000{8B00AE4B-9328-B1DC-57BA-17B7D11AFE9F} {30634283-D150-237A-87FD-DA097EBBE5B9}1.000000{907BAFFD-A66B-2D34-89B1-768E3FCBD951} {30634283-D150-237A-87FD-DA097EBBE5B9}0.992268{9CBE8C71-0131-336C-5211-6EE9060D8B10} {199313F3-4901-ACA6-FFBA-D8C868D9FA00}1.000000{9EA619AE-3EC0-3405-1CC9-282C8FBFBDC2} {30634283-D150-237A-87FD-DA097EBBE5B9}0.996117{BA3EBEE5-218C-5865-B9D7-E6C93F3C852A} {30634283-D150-237A-87FD-DA097EBBE5B9}1.000000{BBB7787E-A8EA-C6AF-4BBD-DBF37A23230A} {30634283-D150-237A-87FD-DA097EBBE5B9}0.995058{C39D9FAE-0933-7C52-B6C0-837E5BD72598} {30634283-D150-237A-87FD-DA097EBBE5B9}0.990783{D418BA33-88B2-4F10-0B5B-878B938CABF5} {30634283-D150-237A-87FD-DA097EBBE5B9}0.984190{DA8A5BA2-BCB6-0C4D-B449-DCD5EF31D732} {30634283-D150-237A-87FD-DA097EBBE5B9}1.000000{DBE066F2-9F22-A9FC-9B3F-11CE933DB934} {75570E27-930A-E412-95FA-B6F033F3988E}1.000000{E11DF7EC-7BAF-593B-C8F7-B84EC4255F66} {30634283-D150-237A-87FD-DA097EBBE5B9}0.990950{E16B4100-C6C6-E5C7-4411-02AD0C4B2D88} {59E0B8D9-09F2-ADC8-9686-9015A09FB3B8}1.000000{E8D9A46E-8D4D-FAE6-F845-2D16E4CD8720} {30634283-D150-237A-87FD-DA097EBBE5B9}0.992395{F27685B7-6847-BAEE-2549-CF30545EF54E} {30634283-D150-237A-87FD-DA097EBBE5B9}0.994889{F65478DB-6A17-9059-1B5A-E144A44D06F4} {30634283-D150-237A-87FD-DA097EBBE5B9}0.995370{FC7FE18F-0F08-55F9-E6BC-67CB932298EE} {30634283-D150-237A-87FD-DA097EBBE5B9}0.985915[31]: output = df_CIDR_rep[df_CIDR_rep[ReputationScore] > 0] output.to_csv(C:/data/CIDR_Reputation_Scores.csv) output[31]: IPUA_Labels 0 1 2 3 TotalCount CIDR Range{11E5AB29-4623-8156-AB63-E7A2FD82A79A} 6.0 584.0 0.0 0.0 590.0{4D1C1E09-9D9F-8960-FEA5-FA478BC0CE02} 1.0 137.0 20.0 0.0 158.0{4ED21F0C-8573-052A-670B-E4D315133AF5} 75.0 3823.0 0.0 0.0 3898.0{60D94A4C-7F09-D0F9-2810-83B93ACBC03E} 12.0 438.0 0.0 0.0 450.0{6893DEE9-A105-29E9-F9D4-98B3411CF42B} 4.0 249.0 0.0 0.0 253.0{7E737FAE-7868-5BE1-84B6-EA9011F3B32E} 33.0 0.0 0.0 333.0 366.0{7EF79BFD-BC3A-903D-C2BD-F5E295ACF4DE} 18.0 1731.0 0.0 0.0 1749.0{855313BB-3482-1ADA-36A8-5D073B29C265} 2.0 235.0 20.0 0.0 257.0{BF61FA52-8010-2731-2B4E-0C77A6C05E6A} 0.0 222.0 0.0 0.0 222.0{D2185CAC-6E67-82FB-603B-F5B33514B577} 9.0 781.0 0.0 0.0 790.0{E3615A90-F5C2-C4A8-40A1-2CCC4102B390} 7.0 812.0 0.0 0.0 819.0{F9164193-183A-7A69-1AE7-25BD6243ED72} 9.0 596.0 0.0 0.0 605.0{FBDFF399-BA15-4F23-EC40-C8E961EF58A7} 1.0 0.0 0.0 36.0 37.0IPUA_Labels CIDR Range BotCount ReputationScore{11E5AB29-4623-8156-AB63-E7A2FD82A79A} 584.0 0.989831{4D1C1E09-9D9F-8960-FEA5-FA478BC0CE02} 157.0 0.993671{4ED21F0C-8573-052A-670B-E4D315133AF5} 3823.0 0.980759{60D94A4C-7F09-D0F9-2810-83B93ACBC03E} 438.0 0.973333{6893DEE9-A105-29E9-F9D4-98B3411CF42B} 249.0 0.984190{7E737FAE-7868-5BE1-84B6-EA9011F3B32E} 333.0 0.909836{7EF79BFD-BC3A-903D-C2BD-F5E295ACF4DE} 1731.0 0.989708{855313BB-3482-1ADA-36A8-5D073B29C265} 255.0 0.992218{BF61FA52-8010-2731-2B4E-0C77A6C05E6A} 222.0 1.000000{D2185CAC-6E67-82FB-603B-F5B33514B577} 781.0 0.988608{E3615A90-F5C2-C4A8-40A1-2CCC4102B390} 812.0 0.991453{F9164193-183A-7A69-1AE7-25BD6243ED72} 596.0 0.985124{FBDFF399-BA15-4F23-EC40-C8E961EF58A7} 36.0 0.97297316 Findings16.1 When the CIDR Reputation Scores were matched against this dataset, 100% of the suspected BOTs were identified!But since the BOT determination is subjective, this finding must be taken with a grain of salt. The primary way of identifying BOTs in this dataset is by finding any activity the IP Address traces back to AWS. At this current point, it is very unlikely any organic activity would come from AWS, but this is not a good long-term assumption.But since unsupervised learning in this notebook only looked at IP Address and User Agent, this bias would not be included in this approach. Random sampling of about 50 rows with high BOT Reputation Scores was done and all pointed to the same known BOT.There were 36 requests from DigitalOcean, LLC. Digital Ocean is another cloud-based platform. Manually research into these 36 requests revealed that there were from a BOT that is used to harvest coupons from eCommerces email campaigns. So, the DBSCAN clustering was able to detect this same type of pattern outside of AWS.17 Future WorkThere were 1158 requests with false negatives in that some requests from AWS did not have a Reputation score. When tracking these down, it looked like the request counts were low for these IP Addresses.Since AWS has thousands and thousands of IP Address and this dataset was only from about 2 days and from a single campaign, chances are these IPs were only used once or twice in this dataset.Another know BOT was also not detected here, mainly due to the fact the email list did not include this subset of email domains known to be associated with this dataset.17.1 A larger dataset should be run through this same processA large dataset (i.e. a full week) might be needed to identify all the CIDRs bound to BOT activity. But this would be 10s of millions of requests. This large a dataset is needed for two primary reasons: Heterogenous Data Most companies send only once a day, some less. So, a subset of data will miss data from email addresses, campaigns and accounts. Number of IP Address There are thousands of IP Address for some of these cloud platforms. A subset of small timeframe of data might not show activity on all the IP Addresses.It would be a simple task to just extend the dataset to a much broader set of requests, but trials on datasets about 4 times larger make DBSCAN not complete due to the time complexity math on the algorithm. So, another clustering model or a new scaling approach would be needed. Some of the possible ways of scaling might be dimension reduction (i.e. PCA) or running the analysis in blocks of time (i.e. 4 hours of requests at a time).