Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Tracking State-Sponsored Threat Actors with OS...

Sponsored · Ship Features Fearlessly Turn features on and off without deploys. Used by thousands of Ruby developers. →

Tracking State-Sponsored Threat Actors with OSINT: A DPRK Case Study

This talk was presented at ISCR 2026.

JeongGak Lyu

August 28, 2026

Transcript

15 slides
  1. Cyber Threat Actor Categories Financial Gain Political & Social Causes

    Malice, Bribery, Negligence Espionage & Political Influence 2
  2. Attribution of Cyber Threat Actors • The Diamond Model as

    a practical framework https://www.activeresponse.org/wp-content/uploads/2013/07/diamond.pdf https://blog.talosintelligence.com/compartmentalized-threat-modeling/ • Collaboration, Initial Access Brokers (IABs), Operational Relay Boxes (ORBs), false flags, and AI increase attribution difficulty 3
  3. DPRK Operations Across All Categories Ransomware, Crypto Heists DDoS, Disruptive

    Attacks IT Worker Operations Espionage, Data Breaches, Disruptive Attacks 5
  4. CTI Lifecycle with LLM Agents • Running Hermes Agent on

    GPT-5.5 Publish Reports and Share Findings Identify Relationships and Draft Reports https://hermes-agent.nousresearch.com/ Monitor RSS Feeds, Mailing Lists, and Other Public Sources Summarize Reports, Extract and Enrich IoCs, Suggest Tags 6
  5. Results of OSINT Collection Annual Reporting Volume 792 800 627

    600 657 520 400 296 172 200 183 204 218 98 0 7 0 16 0 31 18 11 31 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026.8 7
  6. From 246 Names to 8 Clusters • Linking each new

    code name to the most closely related existing code name • Candidate clusters emerge through network analysis • 246 code names resolved into eight clusters 8
  7. Measuring the Cost of Naming Chaos • Coverage Curve: the

    share of reports retrievable when an analyst knows k code names - Single-name coverage ranges from 16% to 77% across clusters Coverage Curve by Cluster Cluster Kimsuky Single Names to Names to Name Reach 90% Reach 100% 77% 5 42 Konni 71% 2 12 Lazarus 65% 11 57 Andariel 61% 5 17 Bluenoroff 29% 13 34 Famous Chollima 27% 9 24 Scarcruft 16% 9 26 100 75 50 25 Kimsuky Bluenoroff 0 1 4 Lazarus Konni Scarcruft Andariel Famous Chollima 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 9
  8. The Shift to Financial Gain • Observed shift toward cryptocurrency

    theft and financial gain since 2015 Annual Incident Counts by Attack Motivation Targeted Sectors Overall 40 Other 8% Government 4% 30 Defense Healthcare 7% 4% 20 Technology 8% 10 0 2009 Espionage 2011 2013 Data Breach 2015 2017 Financial Gain 2019 Destruction 2021 2023 Supply Chain 2025 Cryptocurrency 49% Finance 20% Watering Hole 10
  9. Case Study: Linking a New Code Name • Expel publishes

    a report on Hexagonal Rodent - Report references Lazarus, Famous Chollima, and Contagious Interview D • Expel's Report A links to Report B through shared tags and IoCs • Reports C, D and E connect to B through IoCs and Contagious Interview C B A E ‣ Links to Contagious Interview https://expel.com/blog/inside-lazarus-how-north-korea-uses-ai-to-industrialize-attacks-on-developers/ https://lazarus.day/reports/inside-lazarus-how-north-korea-uses-ai-to-industrialize-attacks-on-developers-R7BvS/ 11
  10. Case Study: Attributing a New Incident • Supply chain attack

    on the JavaScript library Axios • Multiple vendors publish reports using different code names Vendor CrowdStrike Elastic Google(Mandiant) Microsoft Proofpoint Attribution Stardust Chollima REF9135 UNC1069 Sapphire Sleet TA444 ‣ Attributed to Bluenoroff https://lazarus.day/incidents/axios/ 12
  11. Threat Actors for Law Enforcement Primarily subject to investigation and

    arrest Often requiring diplomatic and policy responses 13
  12. Attribution for Law Enforcement • A recent joint advisory was

    unusually detailed, with technical data and IoCs published alongside CTI vendors - Attribution, however, was omitted • Value of in-house attribution capability and publishing attributed findings - Provides an industry verification baseline and a foundation for follow-up research and defensive measures https://police.go.kr/user/bbs/BD_selectBbs.do?q_bbsCode=1001&q_bbscttSn=20260730110134617 https://thehackernews.com/2026/07/hackers-exploit-anysign4pc-via-hacked.html https://www.yna.co.kr/view/AKR20260730121200017 14
  13. Closing • With a different dataset, the same method applies

    to other state sponsors • All data presented here is available at https://lazarus.day 15

More Decks by JeongGak Lyu

Other Decks in Technology