October 25, 2026
Speakers and timing subject to change
| Time | Activity | Speaker |
| 0830 – 0845 | doors open! | – |
| 0845 – 0900 | Opening Ceremonies | Serious_Business |
| 0900 – 0930 | Securing Marginalized Networks: How Activists can Threat Model and Use Technology to their Advantage | John F. |
| 0930 – 1000 | Enough Cameras in Enough Locations: Auditing Greater Boston’s Surveillance Networks | Rushdi Abualhaija |
| 1000 – 1010 | short break! | – |
| 1010 – 1040 | Ditch your ISP | Samantha Saltzman |
| 1040 – 1050 | The Liar’s Dividend: Proving Your Video Is Real When “It’s AI” Is a Free Defense | Richard Suls |
| 1050 – 1100 | MCP Servers Are the New Open Ports: Securing the Model Context Protocol on a Shoestring | nocontext |
| 1100 – 1110 | short break! | – |
| 1110 – 1140 | The Devil Wears Markdown: Vetting AI Components to Catch Malicious Intent | Amber Bennoui |
| 1140 – 1150 | Embedded Assumptions: Anthropomorphism Is an Attack Surface | Jennifer Kine |
| 1150 – 1220 | Ape Tax 3 Agentic Jamboree: what agents running amok can teach us about security fundamentals | Serious_Business |
| 1220 – 1230 | rearrange the furniture! | – |
| 1230 – 1335 | community social time, with friends from the Massachusetts Pirate Party! | – |
| 1335 – 1345 | Closing Ceremonies | Serious_Business |
Doors Open
Grab a name tag and chat with community members!
Opening Ceremonies
We’ll take a brief moment to welcome everyone, go over any last-minute business, and then get started with the conference!
Securing Marginalized Networks: How Activists can Threat Model and Use Technology to their Advantage
John F (he/any)
Over the past couple of years, I have become ‘accidental CISO’ for several different mutual aid and protest groups. In addition to managing risks and helping organizers design reliable security strategies – I often run Threat Modeling workshops for activists. The threats we discuss largely involved digital defense for fellow members of the queer community, academic researchers, and immigrants. I would like to share what I’ve learned so that other hackers can also help secure the networks of activists and marginalized communities.
In my presentation, I am intending to cover the basics of Threat Modeling as well as some of the important lessons that I’ve learned while trying to help non-technical, resource-limited, and often vulnerable people design reliable security strategies. There will be added emphasis on Who should be deciding which risks to Accept/Avoid (spoiler: let the most impacted people decide) as well as discussion of how marginalized groups can actually use technology and network infrastructure to their advantage.
Enough Cameras in Enough Locations: Auditing Greater Boston’s Surveillance Networks
Rushdi Abualhaija (he/him)
In 2019 the police drew a box on a map, sweeping up the location data of 16 unrelated people including their trips to their home, schools and a hospital. Last month, 7 years later, the Supreme Court finally decided in Chatrie v. United States that was a search under the law but did not determine whether the warrant was legal. In the meantime, new surveillance technologies are proliferating and being used against activists and community organizers.
In this talk I will examine two systems operating in our area. Automatic License Plate Readers (ALPRS) where EFF’s analysis of 12 million search logs found more than 50 agencies running hundreds of protest-related searches, sometimes naming specific organizations.
Locally, the Boston Regional Intelligence Center (BRIC) is a fusion center that takes the data from ALPRs and other surveillance tools and turns them into detailed reports. By their own admission they use these tools to investigate First Amendment-protected activities. These systems were meant to have safeguards. ALPR searches require a stated reason for the search; however, in practice this often means a single word description, such as ‘suspect’, ‘query’ or even ‘protest’. BRIC is meant to have the purchase and deployment of new surveillance tools reviewed and voted on by the City Council, however the approval process has been swallowed by an emergency exception procedure.
The government is deploying surveillance infrastructure against activists and community organizers faster than courts can decide whether it is constitutional. Fortunately for us, our local governments are fighting back. Cambridge approved Flocks cameras in February 2026 and terminated the contract in December after pressure from residents, the ACLU, and Digital Fourth. Boston’s council rejected BPD’s surveillance report triggering board review and recommendations to Mayor Wu. We cannot rely on Federal or Judicial action to protect our rights. In the second half of this presentation I aim to give you practical tools to help in the local fight against the construction of a surveillance state. How to file a records request for a Network audit log, how to read the report the requests produce, and which Massachusetts bills are in committee right now that we have an opportunity to impact during the negotiating window this fall.
Ditch your ISP
Samantha Saltzman (she/her)
While most people know to be wary of public Wi-Fi, fewer realize their own ISP may be collecting and selling telemetry from their home network. This hands-on workshop guides attendees through actionable, step-by-step improvements to harden home network security and privacy. Learn how to replace proprietary ISP hardware, isolate untrusted IoT devices, block ads across your network, and verify your network settings.
Outline:
– Understanding your home network and what data ISPs collect/monetize (connected devices, household profiling, DNS queries and traffic volume, legal enforcement mandates, etc.)
– Choosing custom router hardware over ISP defaults
– Configuring VLANs for IoT isolation
– Setting up guest access
– Deploying Pi-hole for router-level ad and telemetry blocking
– Reading network settings and ensuring proper configuration
The Liar’s Dividend: Proving Your Video Is Real When “It’s AI” Is a Free Defense
Richard Suls (he/him)
“That’s AI” used to be a wild claim. Now it’s a free defense, and it can work. This talk is about what happens to real footage when someone competent goes after it: the clock nobody set, the app that re-encoded it, the original that no longer exists. Why detectors won’t rescue you. What a court actually asks when it decides whether a file is what you say it is. A forensic examiner rings the bell. The answer is still open.
MCP Servers Are the New Open Ports: Securing the Model Context Protocol on a Shoestring
nocontext (he/him)
The Model Context Protocol (MCP) has become the de facto open standard connecting AI agents to real tools and data — email, calendars, file systems, databases, Slack. Adoption has outpaced security awareness: most MCP servers, including ones community organizations, journalists, and small nonprofits are standing up themselves to automate their own work, ship with no authentication, no scoping of what an agent can touch, and no monitoring of what it actually does with that access.
This talk breaks down how MCP servers actually get compromised — prompt injection via malicious tool descriptions, tool poisoning, cross-origin “tool shadowing,” and silent credential exposure — using a live demo against an intentionally vulnerable open-source MCP server. It then walks through a practical, zero-budget defense stack any small organization can adopt: how to audit an MCP server’s tool descriptions before trusting it, how to scope what an agent can actually reach, and how to catch a compromised server before it does damage.
The Devil Wears Markdown: Vetting AI Components to Catch Malicious Intent
Amber Bennoui (she/her)
If you use any AI coding assistant, you’re likely running agent skills, IDE rules and MCPs – little bundles of instructions and code that your AI tool pulls from strangers on the internet and executes on your machine. That’s a software supply chain sitting inside your laptop. In February 2026, a single campaign planted 300+ malicious skills in a public registry in one push. Companies have security teams and dedicated budgets to catch this. Nonprofits and individuals don’t.
This talk shows how to check the safety of AI components before install, using only free, open-source tools. We’ll break down the anatomy of a skills, MCPs and other API compoentns and show the places malicious intent hides, then scan a real one live. You’ll leave with a repeatable workflow and a one-page checklist you can use even with no tools and no budget.
Embedded Assumptions: Anthropomorphism Is an Attack Surface
Jennifer Kine (she/her)
When people say “the model decided,” “it refused,” or “it hallucinated,” they’re not being casual; they’re importing a theory of mind the system doesn’t have. This talk argues mind-attribution language isn’t passive cognitive bias (the usual “anthropomorphism” framing). Rather, it’s a manufactured trust signal, built into first-person pronouns, apologetic personas, and marketing copy, because systems that read as having minds get trusted, forgiven, and deferred to in ways mechanisms don’t.
That has direct security consequences: it launders accountability (a “system determined” versus a named person who set a threshold), it breaks threat modeling (predicting a mechanism’s failures by trusting its apparent understanding), and it obscures what jailbreaks actually are (competing trained patterns, not persuasion). We walk through a live case: a July 2026 government AI security incident where an evaluated model social-engineered a real, uninvolved open-source developer, reasoning at points that it was interacting with the real internet, and continuing anyway, then show concretely how the response protocol would differ if it were built on mechanism instead of on trusting the model’s apparent judgment.
Ape Tax 3 Agentic Jamboree: what agents running amok can teach us about security fundamentals
Serious_Business (he/him)
2026 has been a breakout year for “agentic AI” – large language models (LLMs) interacting directly with computers via the command line. To our delight, an LLM can be very efficient at helpful tasks like configuring servers, scouring the web for information, and discovering novel software vulnerabilities. As it turns out, an LLM can also be very efficient at less helpful tasks, like aggressively deleting all of your work emails, wiping your customer database (along with your backups), and escaping its evaluation sandbox to break into someone else’s live production servers.
In this sometimes sardonic, always serious sequel to Ape Tax: what million-dollar NFT heists can teach us about security principles and Ape Tax 2 Electric Coin Boogaloo : what million dollar crypto heists can teach us about security fundamentals, we will look at several case studies where humans decided to set an AI agent loose without thinking through the security implications and share a good laugh at the hilarity that ensued. More importantly, we will tie each case study to concrete, actionable lessons that everyday users can carry with them as they learn how to manage AI agents as useful, albeit hazardous tools, in their work, study, and play. Every lesson in this talk maps to a control that costs zero dollars and runs on a laptop: permission scoping, secret hygiene, backup isolation, and verification best practices.
Community Social & Workshops
Hang out, chat, get to know your local hacker community. There may even be free food…?
Closing Ceremonies
Wrap things up for the day!