#badactors — Public Fediverse posts

Refactored Analysis of TikTok's Content Moderation

The #TikTok CEO's comments seem to confirm my suspicion: the #platform is designed to deliver #content without adequately filtering for #extremistviewpoints during its rapid-fire serving process. While a faster #review is technically possible when a #video is uploaded, the current #moderationprocess is clearly insufficient before #content is served to the public.

This #system works exactly as #badactors need it to. They abuse the platform to spread #radicalizedcontent, and #TikTok appears to do little about it. This is less a failure of #proactive effort and more a consequence of the platform's core #design.

The Algorithmic "#RabbitHole"

The system's priorities appear to be purely #political and #engagement-driven. If a user engages with one piece of #extremistcontent—even through an unrelated #hashtag—the #algorithm starts pushing a flood of similar extremist material to them. This creates a confusing array of user profiles that engage with conflicting #ideologies (e.g., both #extremistright and #extremistleft content), making it difficult to pin down a person's exact #politicalalignment.

I believe the #AI is designed to profile every possible vulnerability—#loneliness, #desperation, #politicalviews—to keep users #engaged for as long as possible. The system operates without any discernible #ethicallogic. If ethical guardrails exist, the #AI certainly doesn't appear to be learning in a safe or positive way.
https://www.ted.com/talks/shou_chew_tiktok_s_ceo_on_its_future_and_what_makes_its_algorithm_different?fbclid=IwY2xjawNMJHxleHRuA2FlbQIxMQABHpe_32vhUBup2q9JgJ5EXh7mN4NDXxRQQ_MC5m9g_RF2IcYGIwnXB6Wc09gI_aem_WbzHkJqzEFEySlER6tmtLg

Refactored Analysis of TikTok's Content Moderation

The #TikTok CEO's comments seem to confirm my suspicion: the #platform is designed to deliver #content without adequately filtering for #extremistviewpoints during its rapid-fire serving process. While a faster #review is technically possible when a #video is uploaded, the current #moderationprocess is clearly insufficient before #content is served to the public.

This #system works exactly as #badactors need it to. They abuse the platform to spread #radicalizedcontent, and #TikTok appears to do little about it. This is less a failure of #proactive effort and more a consequence of the platform's core #design.

The Algorithmic "#RabbitHole"

The system's priorities appear to be purely #political and #engagement-driven. If a user engages with one piece of #extremistcontent—even through an unrelated #hashtag—the #algorithm starts pushing a flood of similar extremist material to them. This creates a confusing array of user profiles that engage with conflicting #ideologies (e.g., both #extremistright and #extremistleft content), making it difficult to pin down a person's exact #politicalalignment.

I believe the #AI is designed to profile every possible vulnerability—#loneliness, #desperation, #politicalviews—to keep users #engaged for as long as possible. The system operates without any discernible #ethicallogic. If ethical guardrails exist, the #AI certainly doesn't appear to be learning in a safe or positive way.
https://www.ted.com/talks/shou_chew_tiktok_s_ceo_on_its_future_and_what_makes_its_algorithm_different?fbclid=IwY2xjawNMJHxleHRuA2FlbQIxMQABHpe_32vhUBup2q9JgJ5EXh7mN4NDXxRQQ_MC5m9g_RF2IcYGIwnXB6Wc09gI_aem_WbzHkJqzEFEySlER6tmtLg

Refactored Analysis of TikTok's Content Moderation

The #TikTok CEO's comments seem to confirm my suspicion: the #platform is designed to deliver #content without adequately filtering for #extremistviewpoints during its rapid-fire serving process. While a faster #review is technically possible when a #video is uploaded, the current #moderationprocess is clearly insufficient before #content is served to the public.

This #system works exactly as #badactors need it to. They abuse the platform to spread #radicalizedcontent, and #TikTok appears to do little about it. This is less a failure of #proactive effort and more a consequence of the platform's core #design.

The Algorithmic "#RabbitHole"

The system's priorities appear to be purely #political and #engagement-driven. If a user engages with one piece of #extremistcontent—even through an unrelated #hashtag—the #algorithm starts pushing a flood of similar extremist material to them. This creates a confusing array of user profiles that engage with conflicting #ideologies (e.g., both #extremistright and #extremistleft content), making it difficult to pin down a person's exact #politicalalignment.

I believe the #AI is designed to profile every possible vulnerability—#loneliness, #desperation, #politicalviews—to keep users #engaged for as long as possible. The system operates without any discernible #ethicallogic. If ethical guardrails exist, the #AI certainly doesn't appear to be learning in a safe or positive way.
https://www.ted.com/talks/shou_chew_tiktok_s_ceo_on_its_future_and_what_makes_its_algorithm_different?fbclid=IwY2xjawNMJHxleHRuA2FlbQIxMQABHpe_32vhUBup2q9JgJ5EXh7mN4NDXxRQQ_MC5m9g_RF2IcYGIwnXB6Wc09gI_aem_WbzHkJqzEFEySlER6tmtLg

Refactored Analysis of TikTok's Content Moderation

The #TikTok CEO's comments seem to confirm my suspicion: the #platform is designed to deliver #content without adequately filtering for #extremistviewpoints during its rapid-fire serving process. While a faster #review is technically possible when a #video is uploaded, the current #moderationprocess is clearly insufficient before #content is served to the public.

This #system works exactly as #badactors need it to. They abuse the platform to spread #radicalizedcontent, and #TikTok appears to do little about it. This is less a failure of #proactive effort and more a consequence of the platform's core #design.

The Algorithmic "#RabbitHole"

The system's priorities appear to be purely #political and #engagement-driven. If a user engages with one piece of #extremistcontent—even through an unrelated #hashtag—the #algorithm starts pushing a flood of similar extremist material to them. This creates a confusing array of user profiles that engage with conflicting #ideologies (e.g., both #extremistright and #extremistleft content), making it difficult to pin down a person's exact #politicalalignment.

I believe the #AI is designed to profile every possible vulnerability—#loneliness, #desperation, #politicalviews—to keep users #engaged for as long as possible. The system operates without any discernible #ethicallogic. If ethical guardrails exist, the #AI certainly doesn't appear to be learning in a safe or positive way.
https://www.ted.com/talks/shou_chew_tiktok_s_ceo_on_its_future_and_what_makes_its_algorithm_different?fbclid=IwY2xjawNMJHxleHRuA2FlbQIxMQABHpe_32vhUBup2q9JgJ5EXh7mN4NDXxRQQ_MC5m9g_RF2IcYGIwnXB6Wc09gI_aem_WbzHkJqzEFEySlER6tmtLg

Refactored Analysis of TikTok's Content Moderation

The #TikTok CEO's comments seem to confirm my suspicion: the #platform is designed to deliver #content without adequately filtering for #extremistviewpoints during its rapid-fire serving process. While a faster #review is technically possible when a #video is uploaded, the current #moderationprocess is clearly insufficient before #content is served to the public.

This #system works exactly as #badactors need it to. They abuse the platform to spread #radicalizedcontent, and #TikTok appears to do little about it. This is less a failure of #proactive effort and more a consequence of the platform's core #design.

The Algorithmic "#RabbitHole"

The system's priorities appear to be purely #political and #engagement-driven. If a user engages with one piece of #extremistcontent—even through an unrelated #hashtag—the #algorithm starts pushing a flood of similar extremist material to them. This creates a confusing array of user profiles that engage with conflicting #ideologies (e.g., both #extremistright and #extremistleft content), making it difficult to pin down a person's exact #politicalalignment.

I believe the #AI is designed to profile every possible vulnerability—#loneliness, #desperation, #politicalviews—to keep users #engaged for as long as possible. The system operates without any discernible #ethicallogic. If ethical guardrails exist, the #AI certainly doesn't appear to be learning in a safe or positive way.
https://www.ted.com/talks/shou_chew_tiktok_s_ceo_on_its_future_and_what_makes_its_algorithm_different?fbclid=IwY2xjawNMJHxleHRuA2FlbQIxMQABHpe_32vhUBup2q9JgJ5EXh7mN4NDXxRQQ_MC5m9g_RF2IcYGIwnXB6Wc09gI_aem_WbzHkJqzEFEySlER6tmtLg

Bad Actors Are Grooming LLMs to Produce Falsehoods

https://americansunlight.substack.com/cp/168074209

#HackerNews #BadActors #LLMs #Falsehoods #Misinformation #AIethics

I came across the #Fail2Ban #docker image from #linuxserverio and thought to myself that it's finally time to set up Fail2Ban. I admit i never used it before and it was a bit difficult to add it to my #playbook as all of my #servers have different services and therefore different #logfile paths, but that's nothing #jinja #templating can't fix.

Now that i've got #Discord notifications for banned #IPs, it's time to work on actual #IPblocking. I also want to use the #IPComplaint and #AbuseIPDB actions as i really like the idea of reporting abuse (even though i have no idea how effective that may be).

I may also want to replace the discord #webhook with #email notifications later as that's mostly the reason i've set up a #mailserver ( #stalwart ) in the first place.

I mean, most of my services are only accessible from #tailscale or my #homenetwork, but since #Ansible makes it so much easier to apply higher standards, i just can't resist. My #homelab is changing every day and i think setting up additional #security, even though i don't need it yet, is never a bad idea.

#networking #badactors #firewall #automation #linux #selfhosting #homeserver

I came across the #Fail2Ban #docker image from #linuxserverio and thought to myself that it's finally time to set up Fail2Ban. I admit i never used it before and it was a bit difficult to add it to my #playbook as all of my #servers have different services and therefore different #logfile paths, but that's nothing #jinja #templating can't fix.

Now that i've got #Discord notifications for banned #IPs, it's time to work on actual #IPblocking. I also want to use the #IPComplaint and #AbuseIPDB actions as i really like the idea of reporting abuse (even though i have no idea how effective that may be).

I may also want to replace the discord #webhook with #email notifications later as that's mostly the reason i've set up a #mailserver ( #stalwart ) in the first place.

I mean, most of my services are only accessible from #tailscale or my #homenetwork, but since #Ansible makes it so much easier to apply higher standards, i just can't resist. My #homelab is changing every day and i think setting up additional #security, even though i don't need it yet, is never a bad idea.

#networking #badactors #firewall #automation #linux #selfhosting #homeserver

I came across the #Fail2Ban #docker image from #linuxserverio and thought to myself that it's finally time to set up Fail2Ban. I admit i never used it before and it was a bit difficult to add it to my #playbook as all of my #servers have different services and therefore different #logfile paths, but that's nothing #jinja #templating can't fix.

Now that i've got #Discord notifications for banned #IPs, it's time to work on actual #IPblocking. I also want to use the #IPComplaint and #AbuseIPDB actions as i really like the idea of reporting abuse (even though i have no idea how effective that may be).

I may also want to replace the discord #webhook with #email notifications later as that's mostly the reason i've set up a #mailserver ( #stalwart ) in the first place.

I mean, most of my services are only accessible from #tailscale or my #homenetwork, but since #Ansible makes it so much easier to apply higher standards, i just can't resist. My #homelab is changing every day and i think setting up additional #security, even though i don't need it yet, is never a bad idea.

#networking #badactors #firewall #automation #linux #selfhosting #homeserver

I came across the #Fail2Ban #docker image from #linuxserverio and thought to myself that it's finally time to set up Fail2Ban. I admit i never used it before and it was a bit difficult to add it to my #playbook as all of my #servers have different services and therefore different #logfile paths, but that's nothing #jinja #templating can't fix.

Now that i've got #Discord notifications for banned #IPs, it's time to work on actual #IPblocking. I also want to use the #IPComplaint and #AbuseIPDB actions as i really like the idea of reporting abuse (even though i have no idea how effective that may be).

I may also want to replace the discord #webhook with #email notifications later as that's mostly the reason i've set up a #mailserver ( #stalwart ) in the first place.

I mean, most of my services are only accessible from #tailscale or my #homenetwork, but since #Ansible makes it so much easier to apply higher standards, i just can't resist. My #homelab is changing every day and i think setting up additional #security, even though i don't need it yet, is never a bad idea.

#networking #badactors #firewall #automation #linux #selfhosting #homeserver

I came across the #Fail2Ban #docker image from #linuxserverio and thought to myself that it's finally time to set up Fail2Ban. I admit i never used it before and it was a bit difficult to add it to my #playbook as all of my #servers have different services and therefore different #logfile paths, but that's nothing #jinja #templating can't fix.

Now that i've got #Discord notifications for banned #IPs, it's time to work on actual #IPblocking. I also want to use the #IPComplaint and #AbuseIPDB actions as i really like the idea of reporting abuse (even though i have no idea how effective that may be).

I may also want to replace the discord #webhook with #email notifications later as that's mostly the reason i've set up a #mailserver ( #stalwart ) in the first place.

I mean, most of my services are only accessible from #tailscale or my #homenetwork, but since #Ansible makes it so much easier to apply higher standards, i just can't resist. My #homelab is changing every day and i think setting up additional #security, even though i don't need it yet, is never a bad idea.

#networking #badactors #firewall #automation #linux #selfhosting #homeserver

1. If you're going to #report somebody you need to give #examples of what #activity you're reporting. Just #reporting a #profile and telling an #instance #admin that an #account is being "#spammy" doesn't cut it. Not when we can clearly see the user is not.

2. Again, not a very helpful report. You report a user and say they're a #SCAM account but offer up absolutely no #evidence to support your #claim.

3. Every account was new once, including the person that reported this one. As long as #NSFW #content is properly #flagged, there's no issue. Suspecting an account to be a "scam account or worse" is not a valid #reason for reporting an account.

4. Another report without any basis or #justification.

5. Again, another baseless report about somebody with absolutely ZERO evidence to back up their claim.

So here's what's going to happen....

The instances that forwarded these #bogus #reports to me, and wasted my time, each have a strike against them.

Strike three and we tell #Sharkey to #ignore any and all reports coming from the #instance in the future.

No #admin wants #BadActors on their instances, and no admin wants a #BadActor to make someone else feel unwelcome.

But you can't just flag and report an account for any bullshit reason either.

You have to give us proof and evidence.

This is not high school and this is not middle school. If somebody is being disruptive or causing problems, let us know.

But if you just suspect someone might cause problems or they might be disruptive... wait until you have proof and evidence and then let us know!

#FediAdmin #AdminStuff #administrivia #UserReports #FediReports #InstanceAdmins #MastoAdmin

This is an ad served by Google Ads. It appears superimposed on a (legitimate) web page, with the rest of the page blurred out. I didn't check to see where clicking the Open button would take me, but I doubt it's anywhere good.

Google does allow users to give feedback on this ad, but the only option is "Seen too many times”. There is no option to flag it as the blatant cross-site scripting/probable #phishing attack that it is.

#security #Google #PhishingAsAService #BadActors

A #Conservative #Parent Got 444 #Books #Removed from a #Wisconsin #SchoolDistrict

This #story illustrates how #badactors #manipulate #bad #policy to create #chaos, while #wasting significant amounts of #time, #money, and #energy for #publicinstitutions and #taxpayers

#Women #Transgender #LGBTQ #LGBTQIA #Conservatives #Extremism #Fascism #Censorship #RepublicanParty #Hate #Bigotry #Violence #Genocide #Discrimination #Racism #Homophobia #Transphobia #ThePartyOfHate

https://bookriot.com/one-parent-got-444-books-removed-from-a-wisconsin-school-district/

The level of disappointment I feel in @[email protected] right now is off the charts. And the blindness of so many to obvious attempts to destroy leftist unity. I've had to constantly fight against this in real world organizing, and I started quietly discussing some obvious bad actors in this space with some people here but here we are.

To be clear, over the course of a month an effective leftist instance was expanded, infiltrated, it's most powerful voice evicted and the instance admin outmaneuvered. The instance may not exist soon, and it provided tooling and lists used to keep much of leftist ActivityPub/Mastodon safe.

People need to take a big step back and look at the big picture of what is going on here. This was effective organizing until others organized to undermine and destroy it.

#oliphant #badactors #anarchistnina

#MhairiBlack: #Trans people should be ‘left the hell alone’.

The #SNP’s #deputy #Westminster #leader said “#badactors” are using the #debate around #trans #people as a “#wedgeissue to cause #chaos

#Women #Transgender #LGBTQ #LGBTQIA #UK #SNP #Allies #Tories #Labour #TERFS #Hate #Bigotry #Homophobia #Transphobia #TERFIsland #MoralPanic #HateSpeech #StochasticTerrorism #Hatecrimes https://www.newschainonline.com/news/mhairi-black-trans-people-should-be-left-the-hell-alone-352062

Of course #aws features are going to be coming fast and furious in the run up to #ReInvent but another neat #CloudSecurity feature for #CloudFront is
hashing of the TLS Client Hello packet to create the JA3 #fingerprint.

Allows you to create a #BlockList (or #AllowList if you're lucky enough) of #BadActors in your #WAF (or just to log for anomalies)

https://aws.amazon.com/about-aws/whats-new/2022/11/amazon-cloudfront-supports-ja3-fingerprint-headers/

#mastodon #security I'm concerned about #vulnerabilities in the #federation of servers. It seems to me that #badActors will be able to use throw away #instances as an #attackVector

#mastodon #security I'm concerned about #vulnerabilities in the #federation of servers. It seems to me that #badActors will be able to use throw away #instances as an #attackVector

#mastodon #security I'm concerned about #vulnerabilities in the #federation of servers. It seems to me that #badActors will be able to use throw away #instances as an #attackVector