def combine_analytic_raw_and_by_cmp_fields(
self, analytic_table, analytic_by_cmp_fields_table,
analytic_raw_and_by_cmp_fields_table):
if table_exists(self.conn, analytic_raw_and_by_cmp_fields_table):
return
self.conn.loadactionset("fedsql")
query = F"CREATE TABLE {analytic_raw_and_by_cmp_fields_table} AS "
tables = [
F"{analytic_table} as s", F"{analytic_by_cmp_fields_table} as c"
]
fields = [
's.* ', 'c._Min_ as "cmpMin"', 'c._Mean_ as "cmpMean"',
'c._Max_ as "cmpMax"', 'c._Nobs_ as "cmpCount"'
]
query += " SELECT " + ", ".join(fields)
query += " FROM " + ", ".join(tables)
where_clauses = list(
map(lambda x: "s." + x + " = " + "c." + x, self.comparison_fields))
if len(where_clauses) > 0:
query += " WHERE " + " AND ".join(where_clauses)
self.log.debug("Combine by_src and by_cmp tables:" +
" ".join(query.split()))
out = self.conn.fedsql.execdirect(query)
self.conn.promote(name=analytic_raw_and_by_cmp_fields_table)
def compute_evidence_detail(self, input_table, input_record_type,
all_event_table, ev_detail_table):
if table_exists(self.conn, ev_detail_table):
self.log.debug("Evidence details table already exists")
return
self.conn.loadactionset("fedsql")
query = F"CREATE TABLE {ev_detail_table} AS "
query += 'SELECT \'' + input_record_type + '\' AS "recordType", r.*, \
n.eventId as "eventId", n.anomalousFlag as "anomalousFlag" '
from_tables = [F"{all_event_table} AS n", F"{input_table} AS r"]
query += " FROM " + ", ".join(from_tables)
where_clauses = list(
map(lambda x: "n." + x + " = " + "r." + x,
self.src_id_fields + self.comparison_fields))
where_clauses.append("n.anomalousFlag = 1")
query += " WHERE " + " AND ".join(where_clauses)
self.log.debug("Evidence detail table query:" +
" ".join(query.split()))
self.conn.fedsql.execdirect(query)
self.conn.promote(name=ev_detail_table)
def run_operation(self, operation, input_field, groupby, input_table,
output_table):
if table_exists(self.conn, output_table):
return
if operation.upper() == 'DISTINCT':
out = self.conn.distinct(table={
"name": input_table,
"groupby": groupby
},
inputs=input_field,
casout={
"name": output_table,
"promote": True
})
elif operation.upper() == 'SUMMARY':
out = self.conn.summary(table={
"name": input_table,
"groupby": groupby
},
inputs=input_field,
casout={
"name": output_table,
"promote": True
})
else:
raise Exception("Unknown operation")
def compute_evidence(self,
input_table,
all_event_table,
evidence_table,
fields_of_interest=None):
if table_exists(self.conn, evidence_table):
self.log.debug("Evidence table already exists")
return
self.conn.loadactionset("fedsql")
ev_create_table = F"CREATE TABLE {evidence_table} AS "
ev_fields = ["n.eventTypeId", "n.eventId", 'n.anomalousFlag']
distinct_flag = (self.src_aggregation_action.upper() == 'DISTINCT')
if fields_of_interest is None:
fields_of_interest = []
if distinct_flag:
fields_of_interest += self.src_id_fields + self.comparison_fields + [
self.input_field
]
else:
fields_of_interest += ["starttime"] + self.src_id_fields + ["srcPort"] + self.dst_id_fields + \
["dstPort"] + self.comparison_fields + ["protocol", self.input_field]
fields_to_select = get_available_fields(self.conn, fields_of_interest,
input_table)
ev_fields += list(map(lambda x: "r." + x, fields_to_select))
ev_select = "SELECT "
if distinct_flag:
ev_select += " DISTINCT "
ev_select += ", ".join(ev_fields)
from_tables = [F"{input_table} AS r", F"{all_event_table} AS n"]
ev_from = " FROM " + ", ".join(from_tables)
fields_of_interest = self.src_id_fields + self.comparison_fields
fields_in_where = get_available_fields(self.conn, fields_of_interest,
all_event_table)
ev_where_clauses = list(
map(lambda x: "n." + x + " = " + "r." + x, fields_in_where))
ev_where = " WHERE " + " AND ".join(ev_where_clauses)
ev_query = ev_create_table + ev_select + ev_from + ev_where
self.log.debug("Evidence table query:" + " ".join(ev_query.split()))
self.conn.fedsql.execdirect(ev_query)
self.conn.promote(name=evidence_table)
def resolve_filter(self, filter):
self.log.debug(F"Resolving filter: {filter}")
output = self.filters[filter]['output'].format(intable=self.intable)
finput = self.filters[filter]['input'].format(intable=self.intable)
clause = self.filters[filter]['clause']
if not table_exists(self.conn, output):
if not table_exists(self.conn, finput):
if finput == self.intable:
table_loaded = load_table(
self.conn,
self.intable,
self.input_file_rel_path,
self.input_caslib,
import_options=self.input_import_options,
index_vars=self.index_vars)
if not table_loaded:
raise Exception(
F"Input table {self.input_file} not found.")
self.log.debug(F"Input table {self.intable} loaded.")
else:
ifilter = self.find_filter(finput)
self.resolve_filter(ifilter)
where = clause.format(table="", not_missing="Is Not Missing")
casout = {"name": output, "promote": True}
if self.index_vars is not None:
casout["indexvars"] = self.index_vars
self.conn.table.partition(table={
"name": finput,
"where": where
},
casout=casout)
self.log.debug(F"Completed resolving filter: {filter}")
def create_table_to_merge(self, table_name):
merge_table_name = F"{table_name}_merge"
if table_exists(self.conn, merge_table_name):
self.log.debug("Merge events table already exists")
return merge_table_name
analytic_field_name = self.event_tables_cols[table_name]
time_id_str = ''
dur = self.event_analytic_duration["value"]
if self.event_analytic_duration["unit"].upper() == 'HOUR':
time_id_str = F'dhms(datepart(eventStartTime_sas), ceil(hour(eventStartTime_sas)/{dur}) * {dur}, 0, 0)'
else:
time_id_str = F'dhms(datepart(eventStartTime_sas), hour(eventStartTime_sas), \
ceil(minute(eventStartTime_sas)/{dur}) * {dur}, 0)'
by_cols = self.src_id_fields + self.comparison_fields + [
analytic_field_name, 'score', 'anomalousFlag'
]
by_cols_str = " ".join(by_cols)
keep_cols = ["timeId"] + self.src_id_fields + self.comparison_fields
keep_cols += [
analytic_field_name, F"{analytic_field_name}_score",
F"{analytic_field_name}_anomalousFlag",
F"{analytic_field_name}_relEvents"
]
keep_cols_str = " ".join(keep_cols)
code = F'''
data {merge_table_name} (keep={keep_cols_str});
retain {keep_cols_str};
set {table_name};
by {by_cols_str};
length {analytic_field_name}_relEvents $ 100;
if first.srcIpAddress then {analytic_field_name}_relEvents = '';
{analytic_field_name}_relEvents = catx(', ', trim(eventId), {analytic_field_name}_relEvents);
eventStartTime_sas = eventStartTime/1000000 + 315619200; /* Convert from unix micros to sas */
timeId = {time_id_str}; /* Ceil to hour or min based on duration */
{analytic_field_name}_score = score;
{analytic_field_name}_anomalousFlag = anomalousFlag;
if last.srcIpAddress then output;
'''
self.log.debug("Merge events table DS code:" + " ".join(code.split()))
self.conn.datastep.runcode(code)
self.conn.promote(merge_table_name)
return merge_table_name
def standardize_analytic(self, input_table, analytic_field,
deviation_table):
if table_exists(self.conn, deviation_table):
return
copy_vars = self.src_id_fields + self.comparison_fields + [analytic_field] + \
['cmpMin', 'cmpMean', 'cmpMax', 'cmpCount']
groupby = self.comparison_fields
std_success = True
try:
self.conn.datapreprocess.transform(table={
"name": input_table,
"groupby": groupby
},
copyvars=copy_vars,
requestpackages=[{
"inputs": analytic_field,
"function": {
"method": "standardize",
"arguments": {
"location":
"median",
"scale": "iqr"
}
}
}],
casout={
"name": deviation_table,
"promote": True
})
self.log.debug("Standardization complete.")
except SWATCASActionError as e:
self.log.debug(F"Standardization failed: {e}")
# Create deviation table with all deviations set to 0
copy_vars_str = " ".join(copy_vars + groupby)
code = F'''
data {deviation_table};
retain copy_vars_str;
set {input_table};
_TR1_{analytic_field} = 0;
'''
self.conn.datastep.runcode(code)
self.conn.promote(deviation_table)
def summarize_analytic_by_cmp_fields(self, analytic_table, analytic_field,
analytic_by_cmp_fields_table):
if table_exists(self.conn, analytic_by_cmp_fields_table):
return
out = self.conn.summary(table={
"name": analytic_table,
"groupby": self.comparison_fields
},
inputs=analytic_field,
casout={
"name": analytic_by_cmp_fields_table,
"promote": True
})
def compute_all_events(self, input_table, cdf_table, cdf_field,
analytic_field, min_value, all_event_table):
if table_exists(self.conn, all_event_table):
self.log.debug("All events table already exists")
return
self.conn.loadactionset("fedsql")
ne_create_table = F"CREATE TABLE {all_event_table} AS "
event_type_id = self.event_type_id
ne_fields = [
F'{event_type_id} as "eventTypeId"',
'n.' + cdf_field + ' as "' + analytic_field + '"',
'n.cdf as "score"'
]
ne_fields += list(
map(lambda x: 'n.' + x,
['cmpMin', 'cmpMean', 'cmpMax', 'cmpCount']))
ne_fields += [
'min(r.starttime) as "eventStartTime"',
'max(r.starttime) as "eventEndTime"',
'max(r.starttime) - min(r.starttime) as "eventDuration"'
]
all_fields = ne_fields + list(
map(lambda x: "r." + x + ' as "' + x + '"',
self.src_id_fields + self.comparison_fields))
ne_select = "SELECT " + ", ".join(all_fields)
from_tables = [F"{input_table} AS r", F"{cdf_table} as n"]
ne_from = " FROM " + ", ".join(from_tables)
where_clauses = list(
map(lambda x: "n." + x + " = " + "r." + x, self.src_id_fields))
where_clauses += list(
map(lambda x: "n." + x + " = " + "r." + x, self.comparison_fields))
ne_where = " WHERE " + " AND ".join(where_clauses)
groupby_fields = ["n." + cdf_field, "n.cdf"]
groupby_fields += list(
map(lambda x: 'n.' + x,
['cmpMin', 'cmpMean', 'cmpMax', 'cmpCount']))
groupby_fields += list(map(lambda x: "r." + x, self.src_id_fields))
groupby_fields += list(map(lambda x: "r." + x, self.comparison_fields))
ne_group_by = " GROUP BY " + ", ".join(groupby_fields)
all_event_query = ne_create_table + ne_select + ne_from + ne_where + ne_group_by
self.log.debug("All event table query:" +
" ".join(all_event_query.split()))
self.conn.fedsql.execdirect(all_event_query)
ne_code = F"DATA {all_event_table};\
SET {all_event_table};\
FORMAT eventId $HEX32.; "
# all_fields = ["eventTypeId", analytic_field, "score",
# "eventStartTime", "eventEndTime", "eventDuration"]
#
# all_fields += self.src_id_fields + self.comparison_fields
all_fields = self.conn.columninfo(
all_event_table)["ColumnInfo"]["Column"].tolist()
all_fields_str = ", ".join(all_fields)
event_id = F" eventId = put(md5(cats({all_fields_str})), $hex32.);"
threshold = self.threshold
ne_flag = F" IF score > {threshold} AND {analytic_field} > {min_value} THEN anomalousFlag = 1; "
ne_flag += " ELSE anomalousFlag = 0;"
ne_code = ne_code + event_id + ne_flag
self.log.debug("All events DS code:" + " ".join(ne_code.split()))
self.conn.runcode(ne_code)
self.conn.promote(name=all_event_table)
def runAnalytic(self):
qual_all_events_table = self.merged_all_events_table.format(
intable=self.intable)
if table_exists(self.conn, qual_all_events_table):
self.log.debug("Merge events table already exists")
return
# Substitute intable in event_table_cols
event_tables_cols_mod = {}
for key, value in self.event_tables_cols.items():
event_tables_cols_mod[key.format(intable=self.intable)] = value
self.event_tables_cols = event_tables_cols_mod
# Get list of all event tables
tables = list(self.event_tables_cols.keys())
# Load each of the table if it is not already loaded
rel_paths = []
for table in tables:
event_type_id = table.split("_")[-3]
rel_path = self.output_file_rel_path(event_type_id)
rel_paths.append(rel_path)
loaded_tables = load_tables(self.conn, tables, rel_paths,
self.ae_caslib)
# Create tables with desired columns
tables_to_merge = list(map(self.create_table_to_merge, loaded_tables))
if len(tables_to_merge) == 0:
self.log.error("No tables to merge")
return
tables_to_merge_str = " ".join(tables_to_merge)
# Run datastep to merge the tables
groupby = "by " + " ".join(["timeId"] + self.src_id_fields +
self.comparison_fields) + "; "
code = "DATA " + qual_all_events_table + "; MERGE " + tables_to_merge_str + "; " + groupby + " run;"
self.log.debug(code)
self.conn.datastep.runcode(code)
# Impute missing values
code = "data {qual_all_events_table}; \
set {qual_all_events_table};\
array change _numeric_;\
do over change;\
if change=. then change=0;\
end;\
run ;"
code = code.format(qual_all_events_table=qual_all_events_table)
self.log.debug("Merge events DS code:" + " ".join(code.split()))
self.conn.datastep.runcode(code)
self.conn.promote(qual_all_events_table)
# Save the table
output_file_rel_path = self.output_file_rel_path()
save_table(self.conn,
qual_all_events_table,
output_file_rel_path,
self.ae_caslib,
replace=True)
self.log.debug("Saved merge events table")
