Awesome
Weird Anti-Cheat Ideas
Objective:
- Minimize the impact of cheaters in FPS games by 2025
Key Ideas:
- Some anti-cheat systems are more resilient to the arms race
- ML detection frameworks can robustly model cheats
- Occlusion culling systems prevent hard-to-detect wallhacks and are guaranteed to be secure
- Automatic server side detections must be highly explainable anbd/or integrate with orthogonal methods
- Proctoring tournaments with a camera for mouse and monitor
- Tie efforts with anti-harassment work
Detection:
- Quantifying impact of cheaters
- Reports / game? Stratify data by player skill, trust, spending?
- Player retention
- Quantifying perception across many games
- Quantifying distribution of types of cheaters and tools to prioritize targets
- Proctoring major online tournaments
- "Help, he's definitely spinbotting" button
- Implement efficient replay system
- Tweak systems to catch cheaters red-handed
- Improved labels provided to deep network
- Labels in time
- Labels with type of hack and degree of certainty/severity
- Record abusive text and voice (legality?)
- Kaggle competition for classifiers. Why crowdsource just the labels?
- Manually generated aimbot data, or clean labels extracted from client-side anti-cheat
- Cool features:
- Trust factor
- Lack of non-aim skills (movement, decision making, pre aim, discipline)
- Drastic changes in skill
- Network access patterns
- Knowledge of enemy positions, never looking in a direction without enemies
- Burst DPS
- Inspiring architectures:
- Generating artificial aimbot data. Will basic smoothing aimbots be classified as abnormal?
- Generating aimbot data from constrained self-play
Prevention:
- Occlusion culling/fog of war
- Increasing cheater cost of reentry--2FA, trust factor, time requirements
- Protecting high-profile streamers from playing against suspicious accounts
- Securing the client
- Quantify effectivness to determine development focus
- What can you do against adversaries with physical access?
- Sandbox highly skilled new accounts, accounts with recently changed passwords or IP
- Stricter identification requirements for top ~1% ranks
- Messing with pixel bots by tweaking in-game colors. Noise or replacing red outlines with a lack of blue outlines.
- Feasibility of hardware security, root of trust, enclaves
Deterence:
- False-positive quantification, defining and constraining error budget
- Tiered sanctions, tets, and punishments--increasing with severity and probability of hacking
- Integrate with VAC
- Ban alternatives
- Shadowban, use bots to supplement population
- Make shadow pool engaging to cheaters with leaderboards and features. Unlikely to stop malicious types.
- Spyro DRM. Pester cheaters in subtle ways that make them blame the game or the hack.
Behavioral:
- Compensate players for running into cheaters, without making the system gameable
- Rollback competitive game changes if a cheater is detected
- Freeze the ranking updates of a suspicious game for manual approval
- Thank them for reporting cheaters
- Incentivize players to help fight cheaters
- Again, incentivize cheaters to stay in isolated lobbies
Anti-Harassment
- Harassment likely causes more damage than cheaters
- Really easy to get organizational support
- Best damage control and rehabilitation policies?
- Harassment and hatred aren't specific to gaming, but it's enhanced by it.
- How to cultivate a community that brings out the goodness in everyone?
Industry Goals:
- Encourage collaboration, publications, and conferences
- Individual developers often like to share ideas
- Large-scale efforts hampered by organizational pursuit of competitive advantage
- Simple, open-source tools to help small developers protect themselves.
- Incorporating ideas from many genres? FPS, MMO, RTS, MOBA, table-top.
- Consulting with former and current cheat creators
- Specific companies to reach out to (in loose order):
- Valve
- Riot
- Tencent
- EA / EAP
- Denuvo, EAC, BattleEye
- Epic Games
- PUBGC
- Blizzard / Activision
- Jagex
- Chess.com and poker sites
- Bungie
- Ubisoft
- Bohemia Interactive
- New World Interactive
- Other indies
- ESA, Fair Play Alliance
Economic:
- Measure economic impact of cheaters on current and past games
- Consider studies on economics of drug industry
- Measuring elasticity of cheat demand
- Possible to undercut cheat providers? Or does supply just create more demand?
Developer Practices:
- EAC dev recommends to not announce ban waves (small number looks lazy, big number normalizes cheating)
- Don't get involved with trolls, only interact with polite users
- Be transparent with community, hold both parties accountable
- Act like a duck--always calm above water
- Stay safe
- Don't be a target. Would someone be incentivized to harm you or your friends?
- Use good personal security practices. Passwords, 2FA.
- Separate hardware network and VM for dangerous work.
Community Interaction
- Success is primarily about player experience and perception
- Survey community for perceptions and suggestions
- Controlled in-game form, Google form, forum analytics?
- Gather additional data on player trust, skill, playtime, etc.
Project Management:
- How to prioritize anti-cheat strategies?
- Impact (short/long term)
- Cost (to develop and distribute)
- Strategies of competitors
- Community feedback
- Ideal anti-cheat team composition. Size, skills?
- Collaborating with infosec/security team?
- Getting organizational resources, using analytics to determine areas of focus
- How to integrate and delegate tasks to third parties like EAC?
- Testing framework. Playtest with pros? Roll out in one map/mode? Random rollout with RCT?
Career:
- How to ramp up quickly and sustainably
- How to decide between offers?
- Potential to learn and network
- Potential to make an impact (freedom, development resources, open-sourcing, company vision)
- Turn down a good offer for a chance of a better offer? Uncertainty of appraisal, expected utility
- Conisder contracting:
Thoughts on Hack vs. Hack
- Could be a mode with competitive integrity, save for problems with balance.
- Hard to get into because it requires investment in VPN and VM/separate machine.
- Rewards decision making over mechanical aim in fights.
- Rewards strategy and teamplay over positional game sense and information control.
- Rewards usage, tuning, and creation of tools.
- Hopefully reduce casual cheating a little bit. Ineffective at deterring people who hack to win.
Misc:
- Examine MOBA market's response to Riot's anti-cheat
- Trust factor, possibly integrate with Valve's.
- Does trust factor stratification increase cheat contagion?
- Custom game modes with wallhacks and easy aim, maybe speed. Modes with asymmetric player power.
- Legally taking down cheat providers. Studies on effectiveness? Economics.
- Laying traps for cheat developers, bait, or directed evolution.
- Outsourcing not only detection, but grunt development work to the community.
News
Favorite talks
