Deno presents its permission system as “secure by default”: programs must be granted access to files, environment variables, or subprocesses, and can ask users for permission at runtime. The critique is that these controls govern Deno’s own standard-library calls, not the operating-system process as a whole. Even the permission prompt can be manipulated because it shares a terminal with untrusted code: a program can alter terminal output to obscure the prompt or exploit terminal responses to make an approval appear to come from the user. The deeper problem is trusting input on a channel the code being authorized can already control.
The “run one command” permission is also unreliable. Deno resolves command names through PATH, which a program can change to substitute a malicious executable; even if command resolution is fixed, environment variables such as LD_PRELOAD can make the dynamic loader execute attacker-controlled code. Blocking particular variables or configuration tricks only patches individual escape routes, since many programs - including editors - can be configured to run other commands. The author argues that Deno’s permissions create a misleading impression of isolation while leaving code in a fully privileged process, with further risks involving symlinks and resource exhaustion. Untrusted code should instead run inside an operating-system sandbox such as bubblewrap. Deno’s restrictions may help against unsophisticated malicious code, but the author considers them too easy to misinterpret as a real security boundary.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
A covert Russian network used fake media posts to exploit French riots, spreading false claims about migrants and portraying the government as weak. French and EU assessments also found Israeli-linked and MAGA accounts pushing hostile narratives about the unrest. (reuters.com)
A Microsoft-signed boot shim can be reused to undermine Secure Boot protections on corporate computers. The research explains how Secure Boot validates trusted boot components and previews a practical technique for exploiting that trust. (leosmith.wtf)
ChatGPT Desktop forwards your local SSH agent to remote machines. Remote agents can use your SSH credentials to access private repositories and push code. (news.ycombinator.com)
Chinese surveillance technology is helping repressive African governments monitor citizens and suppress dissent. The systems give authorities tools to track critics and control public activity. (economist.com)
U.S. officials say Alice Huang, a Chinese businesswoman close to Nvidia CEO Jensen Huang, was recruited by China’s Ministry of State Security to gather intelligence on him and seek access to advanced AI chips. House committee leaders have requested a closed-door briefing from Jensen Huang about his interactions with her. (nytimes.com)
Security researchers found vulnerabilities in a major automaker’s connected-car systems while examining the infrastructure behind its remote-unlock app. An exposed admin tool and weaknesses in the platform enabled access to cars across much of the world. (shubs.io)
Today's best Hacker News stories, summarized and screenshotted, one email a day.