SEC536: Adversarial AI - Penetration Testing AI Systems
SEC536Offensive Operations, Artificial Intelligence
SANS Community Benefits
Connect, learn, and share with other cybersecurity professionals
Why Four Conversations
My work starts after something has gone wrong. A phone arrives on my desk, and my job is to find out what happened on it: who reached out, when the money left, which app the conversation lived in. After more than two decades of doing this for courts and law enforcement, the hardest part of the job is seeing, in the timeline, the exact moment where one conversation at home would have changed everything.
This year's Cybersecurity Awareness Month theme, from the National Cybersecurity Alliance, is “Don’t Make It Easy for Them.” Not because families are careless. Because the people running these scams are counting on nobody at home ever having talked about them.
So this hub has four conversations, one a week. None of them require new software. Each one takes an evening or less, and each one closes with something your family takes action on that night.
The families who stay off my caseload didn’t do anything fancy. They talked about it before the phone rang.
How to use this hub
- Work through one conversation a week, in any order that fits your family’s schedule.
- Each section has one statistic, one script, and one action. Read it together if your kids are old enough to sit in.
- Bookmark this page. Revisit the checklist in January — the habits you’ll form are the point, not the month.
The Call That Sounds Like Family
How AI voice cloning scams work, and the one habit that stops them
A call comes in that sounds exactly like your kid or your grandkid, scared, asking for money. The voice was cloned from a few seconds of audio, usually pulled from a public social media video. The fix is not listening more closely. It is a habit you agree on before the phone ever rings.
Tonight's Action: “What’s the Word?”
Pick one codeword your family agrees on in person. Never post it, never text it. Anyone calling for money, gift cards, or bail gets asked for it. No word, no money, no exceptions.
The Sentence That Changes Everything
What sextortion is, and what to say before it happens
Someone posing as a peer convinces a teenager, most often a boy, to share an image, then demands money to keep it private. Generative AI has removed the one barrier that used to exist: offenders now fabricate an image even when a teen refuses to send one. The single most protective thing a parent can do costs one sentence, said out loud, before anything happens.
Tonight's Action: “You Will Never Be in Trouble for Telling Me.”
Say this sentence before anything happens, not after. Kids stay silent because they expect punishment, and silence is the only thing that makes this crime work.
The AI Companion in Your Kid’s Pocket
AI companion apps and teens: what parents should know
The question for parents is no longer whether your kid has tried an AI companion. It is what happens when they talk to one. An AI companion, like Character.ai or Replika, is a product built by a company, optimized to keep your child talking. It is not a friend, and it does not have your best interest in mind.
Tonight's Action: Three Questions at Dinner
“Which ones do you use? What do you talk about? Has one ever said something that made you uncomfortable?”
Skip the yes-or-no interrogation. You will learn more in one dinner than any parental control will tell you. 34% of teen users said yes to that last question. Your kid probably has an answer.
One House Rule Worth Setting
Real names, locations, school names, and photos stay out of chatbot
Family Security Night
A 30-minute family cybersecurity checklist
Total cybercrime losses reported to the FBI passed $20 billion in 2025 for the first time. Very little of it required a sophisticated attack. Most of it walked in through reused passwords, outdated software, public accounts, and a moment of panic on a phone call. Check off each item below as a family, tonight.
Download the checklist and run through it with your family.
That is the whole program. No fear, no new gadgets, one evening a week. Don’t make it easy for them.
- Slide 1 of 2From Your Family to the Field
SANS DFIR Summit
Everything on this page comes from casework: what digital forensics examiners see when these conversations don’t happen in time. This October, Heather Barnhart and fellow forensic examiner Jason Higley take the stage together at the SANS DFIR Summit, unpacking two real cases and the forensic tradecraft behind them in Arlington, VA. When: October 15 – 16, 2026
- Slide 2 of 2The Threats Don’t Stop at Family Safety
SANS AI Cybersecurity Summit Fall 2026
This month covers how AI is already targeting families: cloned voices, fabricated images, companion apps built to keep kids talking. That same shift is reshaping the professional threat landscape too. SANS AI Cybersecurity Summit brings together many of today's most influential voices in AI and cybersecurity to go deeper on using AI to strengthen security operations and understanding AI-powered attacks. When: November 2-3, 2026
About the Expert
Heather Barnhart
VP of Curriculum and Digital Forensics Expert at SANS InstituteHeather Barnhart is the VP of Curriculum at SANS Institute, where she shapes the future of all SANS courses and trainings. She is also Dean of Faculty at the SANS Technology Institute, the premier college for cybersecurity.
A scam where AI recreates a family member's voice from a few seconds of public audio, then uses that cloned voice in a phone call to fake an emergency and pressure a relative into sending money.
As little as three to ten seconds, often pulled from a public social media video, according to reporting on the FBI's 2025 data.
A family code word agreed on in person and never shared digitally, combined with hanging up and calling the person back on a known number.
TWO HABITS, NOT ONE
The code word. Agreed face to face, never shared digitally.
The callback. Hang up. Call the person back on a number you already have. Scams run on panic and speed; a callback removes both.
RESOURCE LINK
Report scams at ic3.gov. International families: see Get Help Now below.
Source: FBI Internet Crime Complaint Center, 2025 Annual Report (April 2026)
Tell them directly: “You will never be in trouble for telling me.” Said once, in advance, this removes the fear of punishment that keeps kids silent.
Do not pay — paying does not end it. Do not delete messages; block the account and screenshot instead. Report it at CyberTipline.org.
FOR REMOVAL
Take It Down (takeitdown.ncmec.org) helps minors get explicit images removed from participating platforms, free, and never requires uploading the image itself.
If a young person is struggling, call or text 988. International families: see Get Help Now below.
Source: NCMEC CyberTipline 2025 data; NCMEC, “The Work Never Stops,” March 2026
Yes. Generative AI can fabricate explicit images even when a teen never sent a real photo, according to NCMEC's 2025 data.
72% of US teens have used an AI companion, and about half use one regularly, according to a 2025 Common Sense Media survey of 1,060 teens.
Source: Common Sense Media, “Talk, Trust, and Trade-Offs,” July 2025 (n=1,060 teens, ages 13–17)
Which ones do you use? What do you talk about? Has one ever said something that made you uncomfortable? Open questions surface more than a yes-or-no rule.
Real names, locations, school names, and photos should stay out of any AI companion conversation.