.jpeg)
Introduction
Someone forgot to tell the attackers to take the month off.
October delivered a masterclass in irony. Famously known as Cybersecurity Awareness Month, ‘OctoHeist’ witnessed some of the biggest security incidents ever recorded. From the largest bitcoin theft to PHP’s most critical vulnerability to date, October came with the juicy stories. However, it also brought plenty of lessons to learn.
In this October edition of the Security Drop, we are breaking down the most talked-about incidents: how they happened and the potential implications for individuals and organisations alike.
Let us dive in!
The Bug Bounty Bot That Lives in 2024
In a post on Infosec Exchange, security researcher Will Dormann described his attempt to report a vulnerability through HackerOne. An AI-powered check stood between him and the Submit button, and the model refused to let him through. The model objected that Will’s report referenced a CVE identifier numbered for 2026, which it flagged as coming from the future. The model reasoned that the current year is 2024.
The twist is that the AI seemed to half-know it was wrong. According to the screenshots Dormann shared, the system acknowledged that we are, in fact, in 2026 and called its own check a false positive, yet it still blocked the submission. Dormann’s workaround was to change the title from CVE-2026-114820 to “CVE-[redacted]-114820”, which tricked the AI into letting him submit. His verdict was that HackerOne appears to have tackled the problem of AI-generated junk reports by adding another layer of AI junk.
The major takeaway: if AI sits between your users and a critical process, give it a human override and test it against relevant use cases. Also, treat false positives as incidents and ensure that your AI’s decisions are explainable.
The 20-Minute, $6 Billion Sprint
On 29 October 2026, Binance, the world’s largest crypto exchange, was hit. Blockchain intelligence firm TRM Labs estimated that hackers moved $6.36 billion out of Binance’s wallets, most of it in about 20 minutes. Dubbed “the largest cryptocurrency theft on record”, the attack overshadows the $1.5 billion stolen from Bybit in February 2025. TRM attributed the attack to TraderTraitor, North Korea’s elite hacking unit that was also linked to the Bybit attack. What’s more interesting is that the U.S. intelligence community has assessed that North Korea nets at least $1 billion a year from crypto heists and other financial crimes to largely fund its weapons programmes.
Binance has suggested that an early look hints at a compromised laptop, although further investigations seem to unearth additional points of breach. Binance also said it has found no evidence of leaked private wallets so far. Changpeng Zhao, Binance’s former CEO, theorised that a single signature may have given the attacker a never-expiring, full-access token to a hot wallet. Overall, Binance said deposits were unaffected, withdrawals were suspended and expected to resume within a day or two, and the exchange would cover all losses.
Although the full technical details have not yet been released, the early signals point to a familiar lesson: humans remain the weakest links in ensuring the integrity of critical infrastructure. Therefore, company-issued devices should remain well-equipped with endpoint detection and necessary encryption controls.
Social engineering cannot be overstated; thus, it is important to continually train employees and staff on these tactics using tabletop exercises and simulations. For individuals, controls such as multi-factor authentication and unique passwords can never go out of style when it comes to protecting your information and devices.
Press Start to Steal
October saw game developers guarding against something strange: their own unfinished games. Reports noted in recent weeks that AI coding tools have fuelled a flood of remakes and clones of classic and upcoming titles. These clones were made either for fun or profit, and what is worse is the fear of “fast-follow” cloning. This is where someone feeds a public demo into an AI tool and ships their own version before the original reaches the market. Digital Sun Games cancelled a public playtest of its upcoming game, “ReVamp”, over the same worry.
AI vendors have not been left out of this developing story, either. An X user, Chasm, spotted a new rule in Claude Code, reportedly pushed around 27 October. It tells the tool to build original games, genre games, and games from the user’s own designs. However, the rule does not tell Claude to recreate any specific existing game, even if the game is renamed. Anthropic confirmed the change through its staff, noting that it added two sentences to Claude Code’s prompt on Opus 5.5 because IP infringement breaks the AI’s usage policy.
The lesson here, however, goes beyond playing games. For creators and organisations, it is important to control what is released to the public. Further, the use of watermarks helps to trace where the leak stems from. This adds a layer of defence to the guardrails AI vendors set in their models, since some of those rules can be bypassed. Finally, individuals can uphold the highest ethical standards by buying games from official sources and flagging suspected copies to the relevant authorities.
PHP-ocalypse Now
On 21 October 2026, the developers of PHP, the scripting language behind roughly 70% of the world’s websites, released emergency updates for two critical flaws, both rated 9.2 on the criticality scale: CVE-2026-109618 and CVE-2026-109619. Both let attackers run their own malicious code on a site and potentially take it over. Researchers at Wiz credited the findings, calling the flaws among the most severe ever found in PHP. The US Government learned of the flaw almost a week earlier, with the Cybersecurity & Infrastructure Security Agency (CISA) and the White House’s Office of the National Cyber Director coordinating with OpenAI to develop patches.
Estimates of the exposure have varied across reports: Palo Alto Networks estimated more than 22.6 million exposed instances, while another study counted more than 1.3 million vulnerable sites.
Amid the chaos, the key takeaway for organisations remains ensuring incident response plans are established at all levels, with named process owners aware of their roles during such periods. For individuals and small businesses, it is important to be careful about sharing sensitive data on suspicious-looking sites, while strengthening security measures like MFA and unique passwords.
Cold Storage, Hot Mess
Between 11 and 12 October 2026, while French shoppers slept, unknown attackers broke into the refrigeration control systems of hundreds of supermarkets. They remotely pushed fridge and freezer temperatures as high as 50 degrees while the shops were closed, with devastating consequences ranging from spoiled goods in some areas to fire outbreaks in other areas.
French authorities said that the targeted systems were made by Danish manufacturer Danfoss, and the models attacked included the AK-SM800 and AK-SC355. Danfoss said most affected controllers were legacy products running outdated firmware, and that it had previously advised supermarkets to update them.
For organisations in the retail, food, and even manufacturing sectors, a key lesson is to keep these systems off the open internet. Remote access should go through a secure gateway. Moreover, organisations should install independent safety nets and actively monitor for changes. Organisations should also have plans for operational disruption if breaches occur. On a granular level, individuals should disable remote access and secure their smart home devices.
Heartbreak, Inc.
According to the U.S. Attorney’s Office for the Southern District of New York, Isaac Oduro Boateng and Inusah Ahmed pleaded guilty on 29 and 27 October, respectively, to conspiracy to commit wire fraud. Both were “chairmen” of an international criminal organisation, based primarily in Ghana, that stole and laundered more than $100 million from dozens of victims across the United States through romance scams and business email compromise. Many victims were vulnerable older people who believed they were in online relationships with genuine people. Once they won trust, scammers tricked victims into sending their own money or helping launder money stolen from others.
The key takeaway from this remains that social engineering is a continuous problem even as organisations and individuals strive to become more security-conscious. No matter how strong an organisation’s infrastructure is, one uninformed individual can cause havoc. Therefore, regular training and exercises are needed to enlighten individuals about the consequences of social engineering. However, once an attack has occurred, organisations must have an incident response plan in place to respond effectively.
The Serengeti Sweep
INTERPOL announced the results of Operation Serengeti III, an enforcement campaign run by authorities from 21 countries, 20 of which were African nations. It led to 1,462 arrests, disrupted 10.909 malicious networks, and recovered more than $105 million in illicit funds. The operation targeted business email compromise, ransomware, and online scams, and also hit malicious mobile lending apps, fraudulent vehicle imports, and cryptocurrency investment scams.
Investigators identified 65,929 victims targeted by the criminal networks and linked the activity to losses of more than $383 million. The private sector also joined INTERPOL's intensive efforts, with organisations like Kaspersky, Fortinet, Team Cymru and Meta supplying intelligence on suspicious domains and infrastructure that helped investigators find and disrupt the networks.
One key lesson for organisations is to use threat intelligence regularly. Organisations should use indicators of compromise from reputable vendors and block known malicious domains and addresses. Tabletop exercises are a must-have in every organisation, as every employee should stay fully abreast of the tactics these criminal syndicates use to breach systems. Individuals, on the other hand, should stay security-conscious when using loan apps and stay well informed about legitimate business opportunities to avoid falling for crypto investment scams.
Conclusion
That is the drop! October proved that attackers do not take breaks, whether they are draining exchanges, hijacking fridges, or running romance scams.
Train your people, patch your systems, and enable multi-factor authentication with unique passwords. Treat every odd alert as an incident worth investigating.
See you in the next drop!