INSIGHTS

Field notes from the ocean of data

Threat research, forensic tradecraft, and upcoming live sessions from Cyberharpoon operators.

Boyan Cluster streaming fraud case study cover
CASE STUDY

The Boyan Cluster: A 21-Artist Spotify Manipulation Network, Found From One URL

A routine BADS-Audio scan on one 119-monthly-listener Spotify artist flagged a 932× stream-to-listener anomaly. Thirty minutes and ~$0.60 in API costs later, Cyberharpoon had identified a suspected 21-artist manipulation network, a shared TuneCore Denmark distributor-pipeline fingerprint, two hub-node artists, and real-name attribution leads — using our new BADS-Audio Streaming Integrity methodology.

Published · Sep 18, 2026 Streaming Integrity · BADS-Audio
Read case study →
Counterfeit Consensus analysis cover
ANALYSIS

Counterfeit Consensus: Why the FTC's Fake-Account Op-Ed Is a Forensics Problem

A Washington Post op-ed by Sander Gerber and Eugene Kontorovich argues the FTC should regulate fake foreign social media accounts as a deceptive practice — the same legal bucket as fake reviews. The legal framing is catching up to the problem. Here's why forensic bot-network detection is the missing piece any enforcement effort will actually need.

Published · Oct 7, 2026 Disinformation Analysis · Bot Network Detection
Read analysis →
Flydubai FZ1073 methodology case study cover
METHODOLOGY CASE STUDY

Flydubai FZ1073: What Real-Time OSINT Would Have Seen Before the Cockpit Door Closed

When Flydubai Flight FZ1073 was nearly crashed by its own co-pilot on September 30, 2026, CNN, The Guardian, WSJ and others later surfaced years of extremist content on his archived social media — the exact category of signal continuous digital-footprint intelligence is built to flag. A walkthrough of how Cyberharpoon’s cross-platform OSINT methodology maps onto the publicly reported fact pattern, and why the two-year gap between a 2024 red flag and the 2026 incident is the gap this capability closes.

Published · Oct 4, 2026 Methodology · Aviation Vetting
Read case study →
Conversational Training-Data Poisoning — @Miriambendavid × Grok case study cover
CASE STUDY

Conversational Training-Data Poisoning — @Miriambendavid × Grok

Full link analysis of a sustained, multi-language conversational poisoning campaign against xAI's Grok LLM: 861 direct @grok interactions in 6 months, 47 “learned from Grok” attributions, 16 languages, and a repeatable 5-stage cycle designed to embed semantic anchors in future training data. Network graph, narrative clusters, evolution timeline, and evidence chain — all in one interactive report.

Published · Aug 27, 2026 Link Analysis
Read case study →