Wireshark says No interfaces found, or it lists an interface and then captures zero packets while the traffic is clearly flowing. Both failures have the same short list of causes and none of them is "reinstall Wireshark".
The order below is the order to check them. On Linux, the answer is usually the second one; on Windows, the first.
The one-minute triage
# What does the OS think the interfaces are?
ip -brief link # Linux
ifconfig -a # macOS, BSD
ipconfig /all # Windows, in cmd
# Can the user actually read raw packets?
getcap /usr/bin/dumpcap # Linux: should print cap_net_raw,cap_net_admin
If ip -brief link shows an interface with state UP and Wireshark shows nothing, the problem is permissions or the capture driver, not the network. If the interface is DOWN or missing, fix that first — no capture tool can fix a downed link.
1. Windows: Npcap is missing or was installed with the wrong box ticked
Wireshark on Windows cannot capture without Npcap, and the installer asks two questions that decide whether capturing works at all.
- "Restrict Npcap driver's access to Administrators only" — leave this unticked unless you always run Wireshark elevated. Ticked, and a normal user sees interfaces but captures nothing.
- "Support raw 802.11 traffic" — needed only for monitor-mode Wi-Fi capture; harmless to enable.
Reinstalling is the fix, not repairing Wireshark. In the installer, choose the option to install Npcap and keep the "restrict to administrators" box clear. After the reboot, the interface list populates.
2. Linux: dumpcap has no capability, so the user cannot capture
This is the classic Ubuntu and Debian failure. Installing Wireshark packages the capture helper dumpcap, and the post-install question — "Should non-superusers be able to capture packets?" — is answered No by default when the install is unattended. The result: interfaces appear, capture buttons are greyed out, or you get Couldn't run /usr/bin/dumpcap in child process: Permission denied.
# Re-answer the packaging question, choose Yes
sudo dpkg-reconfigure wireshark-common
# Then put yourself in the group the package manages
sudo usermod -aG wireshark "$USER"
# Verify the binary actually carries the capability
getcap /usr/bin/dumpcap
# expected: /usr/bin/dumpcap cap_net_admin,cap_net_raw=eip
Log out and back in. Group membership is not re-read by a running session, and this is why half the answers online "do not work" — the person ran the commands in one terminal and tested in a session that still had the old groups. newgrp wireshark in the current shell is a shortcut for testing.
Running sudo wireshark also works and is the wrong answer: Wireshark's own documentation warns against it, because dissectors are a large attack surface and running them as root turns a malformed packet into a root compromise. If you must run elevated once to prove the diagnosis, do that and then fix the capability.
3. The capture filter is eating everything
A capture filter is applied by libpcap before packets reach Wireshark. If it is wrong, the capture is genuinely empty — not filtered in the display, absent. This is also why host 10.0.0.5 as a capture filter shows nothing when the host is 10.0.0.55 or when traffic is IPv6.
# Prove it: no capture filter at all
tshark -i any -c 20 # 20 packets, whatever they are
dumpcap -i any -c 5 -w /tmp/t.pcap
# Then narrow, one clause at a time
tcpdump -i any -n 'port 443' -c 5
Remember the split that trips people: capture filters use BPF syntax (tcp port 443 and host 10.0.0.5) and display filters use Wireshark syntax (tcp.port == 443 && ip.addr == 10.0.0.5). Typing tcp.port == 443 into the capture filter box is silently invalid in older versions. Drop the capture filter entirely while troubleshooting and use a display filter afterwards; the display filters cheat sheet has the syntax that does work.
4. You are capturing on the wrong interface
Traffic that reaches the machine over a VPN tunnel, a Docker bridge or WSL appears on tun0, docker0 or vEthernet (WSL) — not on the Wi-Fi interface you selected. Loopback traffic never leaves lo at all, so a request to a local service is invisible on eth0.
# Which interface is actually carrying the conversation?
ip route get 1.1.1.1 # shows the egress interface for a destination
ss -tnp | head # which sockets are talking, and over what
# Capture everything at once when you are unsure
tshark -i any -f 'port 443' -c 20
On Windows, Hyper-V virtual adapters are listed alongside physical ones and have near-identical names. The one with the real DHCP address in ipconfig is the physical one.
5. Promiscuous mode is not doing what people think
Promiscuous mode lets the NIC accept frames not addressed to it — on a hub, or on a mirrored port. On a modern switched network or Wi-Fi, enabling it in Wireshark changes nothing: the switch only sends you your own traffic, and a Wi-Fi client cannot see other stations' frames without monitor mode.
If the goal is "see someone else's traffic", the answer is a SPAN/mirror port on the switch or airmon-ng monitor mode on the wireless card, not a checkbox. Expect this to be the whole conversation on a managed network.
6. Offload, so what you capture is not what was sent
Checksum errors on every packet, or giant frames that Wireshark calls malformed, are usually not a capture problem at all: TCP segmentation offload hands the NIC a large buffer and lets the hardware slice it. The capture then records packets larger than the MTU, with a checksum that has not been calculated yet.
# Disable offload on the capture interface (Linux)
sudo ethtool -K eth0 gro off gso off tso off lro off
# Re-enable when finished; the setting does not survive a reboot
sudo ethtool -K eth0 gro on gso on tso on lro on
Fix this on the machine you are capturing on, not on the remote one. It is the single most common reason a legitimate capture "looks broken".
7. It is a driver problem, and it is specific
Two cases worth knowing because no amount of Wireshark reconfiguration helps:
- Windows on Wi-Fi with Npcap in "no monitor mode" builds cannot see 802.11 management frames. Capture on the physical adapter and you get converted Ethernet frames only, which is usually what you want anyway.
- Containers and WSL2 do not share the host's capture points. Capture inside the container namespace (
nsenter -t <pid> -n tcpdump ...) or install tshark in the container; the host's Wireshark cannot see itseth0. This is the case the Wireshark vs tcpdump comparison exists to solve: tcpdump inside the namespace, pcap copied out for Wireshark to dissect.
8. Antivirus and endpoint agents block the capture driver
Corporate EDR products routinely block raw socket access or unload Npcap's driver at boot. The symptom is an interface list that stays empty even after a clean Npcap install, or a driver that works for one session and dies on reboot.
Checking is quick and worth doing before blaming the network:
# Windows: is the driver loaded and started?
sc query npcap
sc query npf
Get-Service npcap, npf | Select Name, Status
If a policy is blocking it, you will not fix this locally. Capture with tcpdump on a Linux host on the same segment instead, and bring the pcap back for analysis.
What "empty capture" actually looks like
| Symptom | Almost always |
|---|---|
| No interfaces listed at all | Npcap (Windows) or dumpcap capability (Linux) |
| Interface listed, capture button greyed out | Same — permissions |
| Capture runs, zero packets | Capture filter, or wrong interface |
| Only broadcast and ARP visible | Switched network — you need a mirror port |
| Packets marked bad checksum | Offload — disable on the capturing host |
| Everything works, then stops after reboot | Driver not started, or EDR unloaded it |
Frequently asked
Why does Wireshark show no interfaces on Ubuntu after installing it?
dumpcap lacks the cap_net_raw capability, usually because the packaging question was answered No. Re-run sudo dpkg-reconfigure wireshark-common, add yourself to the wireshark group, and log out and back in.
Do I need to run Wireshark as root?
No, and its documentation says not to. Give dumpcap the capability instead, or use tcpdump for capture and open the file in Wireshark.
I can see interfaces but capture still gives me nothing. What now?
Remove the capture filter and try tshark -i any -c 20. If that returns packets, the filter was wrong. If it returns nothing, generate traffic deliberately — curl https://example.com — while the capture runs.
Why can't I see another computer's traffic on my Wi-Fi?
Wi-Fi clients only receive frames addressed to them. You need monitor mode on the wireless adapter, or a mirrored port. Promiscuous mode does not help on a switched or wireless segment.
Is -i any the same as capturing on every interface?
On Linux, any is a pseudo-interface that merges traffic from all interfaces and cannot enter promiscuous mode; it is fine for "what is this host talking to" and unsuitable for link-layer analysis. Prefer the real interface when you know it.
Why do the packets look malformed or have bad checksums?
TCP segmentation and checksum offload. Disable them with ethtool -K on the machine doing the capture and the frames will match what left the stack.
Where this fits
The fix is usually one command, and the analysis after the fix is the actual work: display filters to cut a capture down, a tshark or tcpdump run when there is no GUI, and the tool guide for the rest of the workflow. If you are capturing on a client engagement, read the scope first — packet capture on a network you do not own is a different conversation.
See also: Wireshark "No Interfaces Found" — when the interface list itself is empty.