Mirai showed that a list of factory default passwords is enough to build an army of hundreds of thousands of cameras and routers, and its public source code has kept that army growing ever since. Created in 2016 and released on a criminal forum that autumn, Mirai and the families built from it still power large distributed denial of service (DDoS) botnets and, as a 2024 FBI and NSA advisory shows, state-linked operations. This guide explains where Mirai came from, who uses its descendants, how they map to ATT&CK, why device credentials are the core weakness, and how network owners and device makers can detect and prevent infection. It is for network engineers, IoT and OT security teams, SOC analysts and product security leads. It goes deeper on one family than the general Botnets Guide.
Key takeaways
- Default credentials are the root cause. Mirai spread by logging in to internet-facing IoT devices with factory usernames and passwords, not by exploiting sophisticated flaws.
- The code is public, so the family keeps mutating. The US Department of Justice notes that the source code was posted in 2016 and other criminals have since used Mirai variants.
- Scale is enormous. Academic analysis measured a peak of about 600,000 infections; a 2024 FBI-led advisory described a separate Mirai-family botnet of more than 260,000 devices.
- It is not only a criminal tool. The 2024 advisory tied a customised Mirai-family botnet to Integrity Technology Group and to activity consistent with the PRC-linked Flax Typhoon.
- Fixes are known and increasingly mandatory. Unique passwords, disabled remote management, segmentation, patching and replacing end-of-life devices stop most infections, and the UK has banned universal default passwords on consumer connectable products since April 2024.
Who they are (names and aliases)
Mirai is a malware family, not a single group. Several operators have run Mirai-based botnets under different names.
| Name or operation | Operator as stated by the source | What the source says |
|---|---|---|
| Mirai (original) | Paras Jha, Josiah White and Dalton Norman, who pleaded guilty in the US in December 2017 | Created in summer and autumn 2016; hundreds of thousands of compromised cameras, routers and DVRs at peak; used for DDoS |
| Clickfraud botnet | Jha and Norman | December 2016 to February 2017; more than 100,000 mainly US home routers; used for advertising fraud |
| Mirai variants | Many unrelated criminal actors | DOJ notes other actors have used Mirai variants since the source code release |
| Integrity Tech botnet (FBI, CNMF, NSA advisory, September 2024) | Integrity Technology Group, a PRC-based company; activity consistent with Flax Typhoon | Customised Mirai-family payload; active since mid-2021; more than 260,000 devices in June 2024 |
Other IoT DDoS families exist that are not Mirai-derived. The Flodrix botnet, described as a LeetHozer variant, built a DDoS botnet from exploited Langflow servers and was linked to hundreds of internet cameras. Track families by behaviour, not just by name.
Targeting and motivation
- Devices: DOJ names wireless cameras, routers and digital video recorders. The 2024 advisory adds SOHO routers, firewalls, network-attached storage, webcams and IP cameras, many running Linux kernels 2.6 to 5.4 and many still vendor-supported.
- Geography: infections follow device deployment. In the 2024 case, 126,000 of the devices (47.9 percent) were in the United States, and the management database held over 1.2 million records including more than 385,000 unique US devices.
- Criminal motivation: DDoS for hire, extortion and revenge, and advertising fraud, as in the Clickfraud case. DOJ also records that Jha attacked Rutgers University's network, taking its central authentication server offline for days at a time.
- State motivation: large pools of residential and small business devices give state operators deniable infrastructure for intrusions. The 2024 advisory reports Flax Typhoon-consistent use of the same infrastructure against US victims. See the Residential Proxy Networks and Abuse Guide for how compromised devices are used to relay traffic.
Tactics and techniques (MITRE ATT&CK)
ATT&CK v20 does not have a software entry for Mirai. The mapping below applies Enterprise techniques to the behaviour described by DOJ, the USENIX study and the 2024 advisory.
| Stage | Technique | Mirai-family behaviour |
|---|---|---|
| Reconnaissance | T1595.001 Scanning IP Blocks | Internet-wide scanning for devices with remote login services exposed |
| Initial access | T1078.001 Default Accounts | Login with factory default usernames and passwords |
| Credential access | T1110.001 Password Guessing | Cycling through a built-in list of common device credentials |
| Initial access (later variants) | T1190 Exploit Public-Facing Application | Exploiting web interface flaws in routers and cameras where login fails |
| Execution | T1059.004 Unix Shell | Commands run in the device's embedded Linux shell |
| Command and control | T1105 Ingress Tool Transfer | Downloading the architecture-specific bot binary |
| Resource development | T1584.005 Botnet; T1584.008 Network Devices | Infected devices form the attack and relay pool |
| Impact | T1498.001 Direct Network Flood; T1499 Endpoint Denial of Service | Volumetric and application-layer DDoS |
Note that the original Mirai code is widely reported to scan Telnet on TCP 23 and 2323; later variants added other services and exploits. Check the T1090 Proxy Defensive Guide for the relay side of device botnets.
How they abuse identities and credentials
Every IoT device has at least one machine credential: the administrative login that ships with it. Mirai treats those credentials as a shared secret known to the whole internet.
- Universal factory passwords: identical username and password pairs across a product line make one credential list work against millions of devices. The UK PSTI regime, in force since 29 April 2024, requires each product to have a unique password or one the user sets, and bans passwords based on sequential counters, public information or product identifiers such as serial numbers unless derived with accepted cryptographic methods. See the ETSI EN 303 645 Consumer IoT Guide.
- Hidden service accounts: debug and support accounts that users cannot see or change act like hardcoded credentials. The HPE Aruba Instant On hard-coded credentials case shows the same class of flaw in enterprise access points.
- Weak user-set passwords: where owners can change the default, short or common passwords remain guessable by the same dictionaries.
- Device identity beyond passwords: certificate-based device identity and secure onboarding remove shared secrets entirely. See the Device and IoT Identity Guide and the Identity for IoT Fleets Guide.
Notable incidents
- 2016, original Mirai: the USENIX Security 2017 paper "Understanding the Mirai Botnet" analysed seven months of data, found a peak of about 600,000 infections made up mainly of embedded and IoT devices, and documented large DDoS attacks against several high-profile targets in late 2016.
- December 2017, guilty pleas: DOJ announced guilty pleas by Mirai's three creators, and by two of them for the Clickfraud botnet.
- 2021 to 2024, Integrity Tech botnet: the FBI, Cyber National Mission Force and NSA, with allied partners, described a customised Mirai-family botnet active since mid-2021, managed through an application called Sparrow and controlled from China Unicom Beijing addresses.
- 2025, Flodrix: a non-Mirai DDoS family showing the same model applied to exploited servers and cameras (see the breach page linked above).
Detection and defence
What to monitor
| Signal | Where | Meaning |
|---|---|---|
| Inbound Telnet, SSH or web admin logins to IoT devices from the internet | Perimeter firewall, device logs | Exposure that internet-wide Mirai-family scanners are built to find |
| Outbound scanning from IoT devices to many external addresses | NetFlow, firewall logs | Infected device propagating |
| Sudden outbound traffic spikes from cameras, DVRs or NAS | NetFlow, IDS | Participation in a DDoS attack; the 2024 advisory recommends monitoring abnormal traffic volumes |
| IoT devices contacting unknown external hosts on unusual ports | DNS and firewall logs | Command and control or bot download |
| Devices that stop responding to management or reboot unexpectedly | Asset management | Possible compromise; the advisory notes reboots clear fileless malware but reinfection follows if the weakness remains |
Phased programme
- Inventory (weeks 1 to 2): find every camera, DVR, router, NAS and other embedded device, including those installed by facilities teams. Run an external scan of your own address space for exposed management services.
- Remove exposure: disable UPnP, remote management and file sharing that are not needed, as the 2024 advisory advises, and close inbound management ports at the firewall. Test with one site first and keep the old rule set for rollback.
- Change credentials: replace every default password with a unique, strong value stored in a vault; disable password hints where possible. Prioritise internet-facing devices.
- Segment: put IoT devices in their own VLANs with only the connectivity they need, and block outbound traffic except to known update and cloud endpoints.
- Patch and replace: apply firmware updates from trusted channels and replace end-of-life devices that no longer receive updates.
- Clean up infections: reboot to clear memory-resident bots, physically if remote reboot fails, then immediately change credentials and close exposure before reconnecting.
Decision guidance
- Home and small office: buy devices with unique default passwords and a published support period, turn off UPnP and remote admin, and replace routers when support ends. See the Securing Smart TVs and IoT at Home Guide.
- Enterprises: enforce network access control for unmanaged devices and alert on outbound scanning from IoT segments. See the IoT Security Guide.
- Device makers: unique per-device credentials, a vulnerability reporting route and a stated support period are now legal requirements in the UK; align to NIST IR 8259 as well. See the NIST IR 8259 IoT Device Cybersecurity Baseline Guide.
- Targets of DDoS: the defence is upstream mitigation capacity; see the DDoS Protection Guide.
Common failure patterns
- Cameras and DVRs installed by contractors with port forwards that IT never sees.
- Rebooting an infected device without changing its password, so it is quickly reinfected.
- Keeping end-of-life routers in service because they still work.
- Flat networks where an IoT device can reach servers and the internet freely.
Metrics
- Number of IoT devices with management services reachable from the internet (target: zero).
- Percentage of IoT devices with non-default, unique credentials.
- Percentage of IoT devices on vendor-supported firmware.
- Percentage of IoT devices in segmented networks with outbound allowlists.
- Mean time to isolate an IoT device that starts outbound scanning.
Framework mapping
| Control | NIST CSF 2.0 | NIST SP 800-53 Rev. 5 |
|---|---|---|
| Device inventory | ID.AM-01 | CM-8 |
| Unique credentials | PR.AA-01 | IA-5 |
| Disable unneeded services | PR.PS-01 | CM-7 |
| Segmentation | PR.IR-01 | SC-7 |
| Firmware updates and end of life | PR.PS-02 | SI-2, SA-22 |
| Traffic monitoring | DE.CM-01 | SI-4 |
Practitioner checklist
- Scan your public address space for exposed Telnet, SSH and web admin on embedded devices.
- Replace every factory default password with a unique value held in a vault.
- Disable UPnP, remote management and file sharing that are not required.
- Place IoT devices in segmented networks with outbound allowlists.
- Alert on outbound scanning and traffic spikes from IoT segments.
- Patch firmware from trusted sources and replace end-of-life devices.
- After any infection, reboot, change credentials and close exposure before reconnecting.
- Require unique passwords and a published support period in device procurement.
Standards and references
- FBI, CNMF, NSA: People's Republic of China-Linked Actors Compromise Routers and IoT Devices for Botnet Operations
- US Department of Justice: Charges and Guilty Pleas in Three Computer Crime Cases Involving Mirai and Clickfraud Botnets
- USENIX Security 2017: Understanding the Mirai Botnet
- UK Government: The UK Product Security and Telecommunications Infrastructure (Product Security) regime
- MITRE ATT&CK: Valid Accounts: Default Accounts (T1078.001)
- MITRE ATT&CK: Network Denial of Service (T1498)
- NIST: Cybersecurity Framework 2.0
Related NHI Mgmt Group resources: Botnets Guide · IoT Security Guide · Device and IoT Identity Guide · DDoS Protection Guide