Overview
An insider-threat investigation at a design agency. capture.pcap holds a full workday of traffic from a monitored workstation. Somewhere in there, something left the building that shouldn’t have.
Prerequisites / Setup
tshark(or Wireshark)stringscapture.pcapin the working directory
Recon
Step 1 — Initial analysis
Goal: Survey the capture’s protocol mix.
The capture contained a large amount of normal workstation traffic, including DNS, TCP, TLS, HTTP, and Microsoft/Google/Slack-related services.
Enumeration
Step 2 — Look for suspicious DNS activity
Goal: Rule out DNS tunneling or TXT-based exfiltration first.
I first looked for suspicious DNS activity, but the queries appeared mostly normal and did not show obvious DNS tunneling or TXT-based exfiltration.
Step 3 — Isolate TCP packets with raw application data
Goal: Find the anomalous high-entropy flows.
tshark -r capture.pcap \
-Y 'data.data' \
-T fields -E separator='|' \
-e frame.number \
-e tcp.stream \
-e frame.time_relative \
-e ip.src \
-e ip.dst \
-e tcp.srcport \
-e tcp.dstport \
-e tcp.seq \
-e tcp.len \
-e data.len \
-e data.data
This isolated 11 unusual packets.
They shared several characteristics:
- Internal
10.42.x.xhosts communicating with external IP addresses. - Destination port
443. - Payloads that were not normal TLS records.
- High-entropy binary-looking data.
- Payload sizes ranging from 41 to 291 bytes.
- Most appeared as single client-to-server data transmissions.
Example:
10[.]42[.]18[.]19:7937 -> 52[.]244[.]251[.]24:443
TCP payload: 198 bytes
and later:
10[.]42[.]18[.]19:26072 -> 52[.]71[.]136[.]222:443
TCP payload: 148 bytes
The repeated appearance of 10[.]42[.]18[.]19 made it particularly interesting, although it was not the only host involved.
Exploitation
Step 4 — Payload analysis
Goal: Confirm the payloads are not ordinary TLS application data.
The payloads did not begin with standard TLS record bytes such as:
16 03 03
Instead they looked like encrypted/random binary data, for example:
447b665b4080e9da5a28bd748c8e0246...
This suggested that the traffic was not ordinary HTTPS application data and was worth treating as a possible covert channel or exfiltration mechanism.
Step 5 — String search for recovery material
Goal: Hunt the capture for keys, tokens, or the flag itself.
strings -a capture.pcap | \
grep -Ei 'key|secret|password|token|nonce|aes|gcm|encrypt|decrypt|side|door|flag'
The capture contained:
recovery-token: r00t{REDACTED}
Do not distribute outside the design team.
The recovery token is the key finding from the capture and indicates that sensitive information was present in the traffic.
Flag / Secret
r00t{REDACTED}
Conclusion
The investigation narrowed the workday traffic down from thousands of packets to a small set of anomalous high-entropy TCP/443 transmissions. These packets were inconsistent with ordinary TLS traffic and represented the likely covert transfer associated with the insider threat.
A direct string search of the PCAP ultimately exposed the recovery token:
r00t{REDACTED}
Lessons / Takeaways
- Rank flows by rarity, not volume: 11 high-entropy TCP/443 packets stood out against thousands of normal packets.
- Non-TLS payloads on port 443 are worth inspecting closely.
- A direct string search of the capture can expose material that flow analysis alone misses.
Mitigations
- Inspect and log application-layer payloads on port 443, not just TLS metadata.
- Alert on high-entropy, non-TLS payloads to external IPs.
- Monitor data exfiltration volumes per internal host.
Techniques
References
- tshark / Wireshark: https://www.wireshark.org/

