Iron March - Fascist Forum Network Analysis (incomplete)

Introduction

In November 2019, the SQL database for defunct fascist forum Iron March was leaked to the public. Iron March, closed in 2017 for reasons that remain unclear, served as a breeding ground for right-wing political extremism, often edging into terrorism. Analysis of users and patterns found on this forum provides interesting insight into the trends present in the modern far-right through 2017 and may expose vectors for further research going forward.

Data

For this analysis, I’ve pulled all the necessary data directly from the MySQL dump, converting each user’s IP address to a set of estimated coordinates through the IPStack API. From here it’s been exported to CSV and is ready for use. We’ll start by converting our CSV into a Pandas DataFrame.

members = pd.read_csv('data/core_members_coords_extra.csv', index_col=0)

Let’s get a quick idea of what our data looks like.

members.sort_values(by="Member ID").head()

	Member ID	Name	Email	IP	Latitude	Longitude
0	1	Александр Славрос	slavros_a@mail.ru	178.140.119.217	55.764408	37.636600
5	2	PhalNat	illuminatienlightened@hotmail.com	68.37.21.125	42.250519	-83.172760
9	3	Blood and Iron	renegader23@aim.com	68.10.255.89	36.764111	-76.341103
10	4	Mierce	hominemcura@gmail.com	82.29.169.221	52.471390	-1.735000
11	5	Will to Power	tashkentfox@hotmail.com	90.214.150.70	53.810280	-1.544440

Great! This is well-formatted, easy-to-read data. However, it doesn’t tell us much by itself. We want to perform analysis on this data, and I’m particularly interested in two things: generating a clear vizualization of the forum and identifying key figures in the community. Let’s start with a visualization.

Visualization

There are a number of options at our disposal for visualizing our data (in fact, my earliest approach used QGIS and can be found here). For the purposes of keeping this analysis in the Python family, let’s use the geopandas library. Here, we’ll read in a few geographic datasets for visualization.

world = gpd.read_file(gpd.datasets.get_path('naturalearth_lowres'))
states = gpd.read_file('data/gis/cb_2018_us_state_500k').to_crs(epsg=4326)
members_points = gpd.GeoDataFrame(members, geometry = gpd.points_from_xy(members.Longitude, members.Latitude), crs="epsg:4326")

Let’s make sure each of our datasets shares the same coordinate reference system (CRS). We’ll use

all(world.crs == x.crs for x in [cities, states, members_points])

because

1. It makes it clear that we want world to set our baseline CRS.
2. It’s one operation performed against a list of objects, and I prefer to keep that obvious and explicit. Daisy-chaining comparisons obscures the purpose of the operation.

all(world.crs == x.crs for x in [states, members_points])

True

Now that we’re sure everything matches up, let’s take a look at what’s changed in our data.

members_points.sort_values(by="Member ID").head()

	Member ID	Name	Email	IP	Latitude	Longitude	geometry
0	1	Александр Славрос	slavros_a@mail.ru	178.140.119.217	55.764408	37.636600	POINT (37.63660 55.76441)
5	2	PhalNat	illuminatienlightened@hotmail.com	68.37.21.125	42.250519	-83.172760	POINT (-83.17276 42.25052)
9	3	Blood and Iron	renegader23@aim.com	68.10.255.89	36.764111	-76.341103	POINT (-76.34110 36.76411)
10	4	Mierce	hominemcura@gmail.com	82.29.169.221	52.471390	-1.735000	POINT (-1.73500 52.47139)
11	5	Will to Power	tashkentfox@hotmail.com	90.214.150.70	53.810280	-1.544440	POINT (-1.54444 53.81028)

You’ll notice we now have a geometry column filled with POINTs! members_points is what’s called a GeoDataFrame, which is just an extension of the standard Pandas DataFrame that has a higher understanding of geographical information.

Plotting our Data

To plot our data, we’ll set our axis to be world and plot each additional set of information (individual states and the location of each member) on top of it.

ax = world.plot(figsize=(30,30), color='none', edgecolor='grey')
states.plot(ax=ax, color='none', edgecolor='grey')
members_points.plot(ax=ax, markersize=5)

<AxesSubplot:>

def get_member_state(member_id):
    return members_usa.loc[members_usa["Member ID"] == member_id]["State"]

def get_member_location(member_id):
    try:
        x = members_points.loc[members_points["Member ID"] == member_id].geometry.x.values[0]
        y = members_points.loc[members_points["Member ID"] == member_id].geometry.y.values[0]
        return (x, y)
    except:
        return (0, 0)

to = gpd.GeoSeries(LineString([get_member_location(1), get_member_location(7)]))

edges = pd.read_csv("edges.csv")

def dict_head(d, n):
    for item in list(d.items())[0:n]:
        print(item)

def get_member_name(id):
    try:
        return members.loc[members["Member ID"] == id].Name.item()
    except ValueError:
        return ''

messages = pd.read_csv("all_messages_headers.csv", index_col="mt_id")
connections = pd.read_csv("distinct_connections.csv")

d = defaultdict(list)
connection_tuples = zip(connections["msg_author_id"], connections["mt_to_member_id"])

for sender, recipient in connection_tuples:
    d[sender].append(recipient)

g = nx.Graph()

for k, v in d.items():
    g.add_edges_from([(k, t) for t in v if k != t])

with open("edges.csv", 'w') as f:
    f.write("Sender,Recipient\n")
    for edge in g.edges():
        f.write(f"{edge[0]},{edge[1]}")
        f.write('\n')

def convert_connection_to_location(tuple_of_ids):
    return (get_member_location(tuple_of_ids[0]), get_member_location(tuple_of_ids[1]))

for edge in list(g.edges)[0:5]:
    print(edge, convert_connection_to_location(edge))

(1, 3) ((37.63660049438477, 55.76440811157226), (-76.34110260009766, 36.76411056518555))
(1, 11) ((37.63660049438477, 55.76440811157226), (-6.267000198364258, 53.29199981689453))
(1, 20) ((37.63660049438477, 55.76440811157226), (0, 0))
(1, 23) ((37.63660049438477, 55.76440811157226), (115.8829574584961, -31.987459182739254))
(1, 25) ((37.63660049438477, 55.76440811157226), (25.65834045410156, 60.97164916992188))

lineseries = []
for edge in list(g.edges):
    lineseries.append(gpd.GeoSeries(LineString(convert_connection_to_location(edge))))

lineseries_pd = pd.DataFrame(lineseries)
lineseries_pd.rename(columns={0: "geometry"}, inplace=True)

lineseries_gpd = gpd.GeoDataFrame(lineseries_pd)

lineseries_gpd.head()

	geometry
0	LINESTRING (37.63660 55.76441, -76.34110 36.76...
1	LINESTRING (37.63660 55.76441, -6.26700 53.29200)
2	LINESTRING (37.63660 55.76441, 0.00000 0.00000)
3	LINESTRING (37.63660 55.76441, 115.88296 -31.9...
4	LINESTRING (37.63660 55.76441, 25.65834 60.97165)

ax = world.plot(figsize=(30,30), color='none', edgecolor='grey')
states.plot(ax=ax, color='none', edgecolor='grey')
members_points.plot(ax=ax, markersize=5)
lineseries_gpd.plot(ax=ax, color='black', alpha=0.05)

<AxesSubplot:>

bt = nx.betweenness_centrality(g)
centrality = [(get_member_name(k), k, v) for k, v in sorted(bt.items(), key=lambda x: x[1], reverse=True)]

for line in centrality[:50]:
    print(f"#{centrality.index(line) + 1}\t{line[1]}\t{line[0]+'':<20}\t{line[2]}")

#1	1	Александр Славрос   	0.24749951007026946
#2	7	Daddy Terror        	0.11138219902228376
#3	7600	Odin                	0.08294539794003127
#4	9174	Atlas               	0.062343767883346626
#5	353	SpookHunter         	0.05620958276449609
#6	2170	Zeiger              	0.0542133015520747
#7	130	Pro Patria Mori     	0.05101861417992674
#8	2306	Sammy               	0.04833233980581861
#9	6113	Aquila              	0.04319930048452017
#10	6168	Raycis              	0.04010608213921796
#11	35	American_Blackshirt 	0.03932823869527507
#12	132	Myrrysmies          	0.037157120344800613
#13	3491	New Canadian Empire 	0.035986004270294886
#14	7424	Bear                	0.03426406418693662
#15	49	Владимир_Борисов    	0.033637839606292116
#16	960	                    	0.03328299868224244
#17	7816	kllш                	0.03159984495073807
#18	9393	Blackshirt 13       	0.02838381655749529
#19	4873	Rintrah             	0.026853258702452924
#20	2075	EvilCatholicNaziGoy 	0.02620845655983659
#21	2	PhalNat             	0.024780602906112358
#22	168	Growth of the Soil  	0.023543498873975305
#23	1558	Spöket              	0.02315096126098888
#24	6322	Асенов              	0.022951479245250105
#25	9503	HermannTheGerman    	0.022390651394737424
#26	9304	TheWeissewolfe      	0.021976726108282035
#27	6249	The Yank            	0.021918791051853674
#28	72	T-34-85Forthewin    	0.020986748048481427
#29	288	Clive Bissel        	0.02020353366276106
#30	4	Mierce              	0.019875739014479567
#31	16	Talleyrand          	0.019532703339924507
#32	9144	Neizbezhnost        	0.018914314909373346
#33	7481	RadioFreeGB         	0.01825898645824197
#34	3721	suspiciouscelerystalk	0.018018732299192576
#35	150	Insurrectionist     	0.01767605239362129
#36	3	Blood and Iron      	0.016697829349899367
#37	315	Бронеизтребител     	0.01624798989444871
#38	9927	Victor Breivik      	0.01622369540683362
#39	67	État de Stase       	0.015685250972063098
#40	293	☧☧☧                 	0.015275113481666165
#41	84	FascistCapitalist   	0.014628389068116072
#42	161	Loved I Not Honour More	0.014489454003828783
#43	2392	gingertoast         	0.012930395203228334
#44	9475	mengligiraykhan     	0.01289428196441316
#45	11	Four Suited Jack    	0.012484008161833971
#46	7757	Carcamano           	0.012451261917570088
#47	9288	Змајевит            	0.011751561429145494
#48	2220	Kulturkampf         	0.010976563605598308
#49	6155	AlbaNuadh           	0.010884779568055595
#50	158	TotalitarianSocialist	0.010717973897778022

get_member_name(1)

'Александр Славрос'