STEM Synapse

Subjects

Tech & Engineering

Today

  1. Paper of the DayPreprint, not yet peer reviewed

    Adversarial Reinforcement Learning for Adaptive Eavesdropping in BB84 Quantum Key Distribution

    Quantum key distribution lets two parties share a secret code with a guarantee from physics: anyone listening in disturbs the signal in a way that shows up as extra errors.

    Why it matters An eavesdropper that learns from feedback can slip past the standard error check in quantum key distribution far more often than fixed attack models assume.

    Security checks usually assume the eavesdropper intercepts at a fixed rate. Here the eavesdropper was instead a simple learning program that watched the error rate and decided, one quantum bit at a time, whether to intercept. With 1% channel noise, it cut its chance of being detected from 99.4% under the best fixed strategy to about 0.28%, while still collecting around 10 correct key bits per round. The agents also learned to burst their attacks at the very end, a trick that randomised check timing removed.

    Bagunu, arXiv cs.CR, 17 November 2026.

    Read the paper: Adversarial Reinforcement Learning for Adaptive Eavesdropping in BB84 Quantum Key Distribution

Tuesday 17 November

  1. Paper of the DayPreprint, not yet peer reviewed

    Design and Experimental Evaluation of a Low-Cost Immersion Cooling System

    Sinking computer hardware in oil is one way to cool crowded data centres, but rigs for studying it are usually costly.

    Why it matters Liquid cooling of data centre hardware can now be studied cheaply, with a rig anyone can rebuild to compare cooling oils and layouts.

    The researchers built an open design from parts available in India for about 380 US dollars: a heater standing in for a server sits in a glass tank of mineral oil, a pump pushes the oil through a fan-cooled radiator, and a small microcontroller logs the temperatures. At full load the oil stayed only 8 degrees Celsius above room temperature while the heater surface ran far hotter. Almost all the resistance to heat flow sat where the heater meets the slowly moving oil, so the cooling loop itself has capacity to spare. The plans, code and data are public.

    Mandal et al., arXiv eess.SY, 16 November 2026.

    Read the paper: Design and Experimental Evaluation of a Low-Cost Immersion Cooling System

Monday 16 November

  1. Paper of the DayPreprint, not yet peer reviewed

    Implementation of a 1D Gaussian-modulated CV-QKD under finite-size parameter estimation

    Quantum key distribution lets two parties share a secret encryption key using light, with security that rests on physics rather than on hard mathematics.

    Why it matters A simpler way to share encryption keys using quantum light delivered 0.364 megabits per second over 12 km of ordinary optical fibre.

    A leading version encodes the key in two properties of a light wave, which places heavy demands on the sending equipment. Encoding in only one property is simpler, but a complete working system had been hard to build. The researchers built one, running over 12 km of ordinary optical fibre with separate lasers at each end and the full chain of processing needed to distil a secure key. It produced 0.364 megabits of secret key per second under realistic assumptions about limited data. That figure shows what the simpler design costs in performance.

    Nascimento et al., arXiv quant-ph, 12 November 2026.

    Read the paper: Implementation of a 1D Gaussian-modulated CV-QKD under finite-size parameter estimation

Sunday 15 November

  1. Paper of the DayPreprint, not yet peer reviewed

    Reconfiguring System-Level Protection Through Bounded Delegation in RISC-V MCUs

    The small chips inside embedded devices wall off parts of their memory so that one piece of software cannot tamper with another.

    Why it matters Small embedded chips could change their memory protection rules while running, with limits on who may grant what, without slowing ordinary memory access.

    Increasingly those walls need to move while the device is running. Existing chips can limit who may change the rules, but not what rules they may then write, so a permitted program can overreach. The authors propose a scheme in which each piece of software owns a region and can only make or hand on rules inside it, with every change checked before it takes effect. Built into an open-source chip design, it did not slow ordinary memory access, and switching rules during an update was 89.5% faster than routing the change through a supervising software layer.

    Cunha et al., arXiv cs.AR, 9 November 2026.

    Read the paper: Reconfiguring System-Level Protection Through Bounded Delegation in RISC-V MCUs

Saturday 14 November

  1. Paper of the DayPreprint, not yet peer reviewed

    U.S. Data Center Growth Through 2030: A Critical Review of Grid Adequacy and Transmission Expansion Planning

    Can the US electricity system keep up with the data centers planned through 2030?

    Why it matters Whether the US grid can supply the data centers planned through 2030 cannot be read from national capacity totals; local limits and timing decide.

    This review of public forecasts and studies argues that the headline numbers cannot answer the question. Forecasts mix three different things: average demand over a year, peak demand, and the size of connection a site requests. And a national total of generating capacity does not show a dependable surplus, because what can be delivered depends on plant availability, regional supply, transmission bottlenecks and project timing. Data centers also behave unusually: a fault on the grid can make many switch to backup power at once and return later. The review finds that this behaviour is rarely built into the studies used to plan new power lines.

    Porkar et al., arXiv eess.SY, 13 November 2026.

    Read the paper: U.S. Data Center Growth Through 2030: A Critical Review of Grid Adequacy and Transmission Expansion Planning

Friday 13 November

  1. Paper of the DayPreprint, not yet peer reviewed

    Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents

    AI agents increasingly run for hours unattended with wide access to a computer, so a tampered model is a security risk.

    Why it matters Editing fewer than ten internal units of an AI model, with no retraining, can plant a hidden trigger for an attacker, though most such edits were detected.

    This preprint shows how little tampering it takes. The method edits fewer than ten of a model's unused internal units, with no retraining, so that when a chosen trigger text appears, the model outputs a command picked by the attacker. In one model, changing seven units out of more than 12,000 set off the command in at least 98.8% of test runs containing the trigger, and in none of the normal runs checked. Two other models were just as vulnerable. The edits leave traces: automated defences flagged 91.4% of tampered models, while wrongly flagging 9.7% of clean ones.

    Zeng et al., arXiv cs.CR, 11 November 2026.

    Read the paper: Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents

Thursday 12 November