Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Spectre and Meltdown are processor vulnerabilities disclosed in January 2018 that abuse speculative execution to read memory across security boundaries, leaving traces in the cache that an attacker can measure.

What Are Spectre and Meltdown?
On 3 January 2018, two research teams disclosed a class of vulnerability that was not in any piece of software. Spectre and Meltdown exploited speculative execution, a performance technique built into almost every high-performance processor made in the previous twenty years, to read memory the attacker had no right to see. The flaws were in the silicon, the affected population was effectively every modern computer, and the mitigations cost measurable performance.
Rather than waiting to discover whether a branch will be taken or whether a memory access is permitted, a modern processor guesses and carries on. If the guess was wrong, the architectural results are discarded and nothing in the programmer-visible state reflects what happened. The problem is that microarchitectural state is not discarded. Data loaded speculatively remains in the cache, and the difference between a cache hit and a cache miss is measurable in nanoseconds. Speculation therefore leaves a shadow, and the shadow can be read.
Meltdown, CVE-2017-5754, applied mainly to Intel processors. A user-space program reads a kernel address, which will fault, but speculative execution proceeds before the permission check completes and uses the fetched byte to index an array the attacker controls. The fault is then raised and the architectural read is discarded, yet one cache line of the attacker array has been loaded. Timing the array reveals which line, and therefore the value of the byte.
The result was that any unprivileged process could dump kernel memory, and in cloud environments that meant reading across tenancy boundaries. The mitigation, kernel page table isolation, removes kernel mappings from the user page tables so there is nothing to speculate against. It derived from the KAISER work at Graz and costs performance on syscall-heavy workloads, though later processors fixed the underlying behaviour in hardware.
Spectre is the harder family and has proved far more durable. Variant 1, CVE-2017-5753, is bounds check bypass: a mispredicted branch causes an array access beyond its bounds, and the speculatively loaded value is leaked through the cache. Variant 2, CVE-2017-5715, is branch target injection, where an attacker trains the indirect branch predictor so that a victim process speculatively executes a gadget of the attacker’s choosing.
Spectre leaks data within the victim’s own address space, so it breaks isolation between a browser and a page’s JavaScript, between a hypervisor and a guest, and between a sandbox and its host. Mitigation is piecemeal by nature: lfence barriers inserted at vulnerable patterns, retpolines to replace indirect branches with a construct the predictor cannot be trained against, and microcode features such as IBRS, IBPB, and STIBP. Browsers responded by reducing timer resolution, adding jitter, and introducing site isolation so that cross-origin data is not in the same process to begin with.
The disclosure opened a research field rather than closing an issue. Foreshadow attacked Intel SGX enclaves in 2018. ZombieLoad, RIDL, and Fallout in 2019 leaked data from internal processor buffers rather than the cache, a family grouped as microarchitectural data sampling. Retbleed in 2022 defeated retpoline on certain processors. Downfall, CVE-2022-40982, disclosed in August 2023, used gather instructions to leak data across boundaries on a wide range of Intel parts. Inception, CVE-2023-20569, did comparable damage on AMD Zen. Each required new microcode, new kernel work, and another small performance concession.
/sys/devices/system/cpu/vulnerabilities/; on Windows, Microsoft publishes a PowerShell module that reports which mitigations are active.These are side-channel attacks in the strict sense, and they illustrate why isolation claims need testing rather than trusting. Our Infrastructure and Architecture Review Services include verification that mitigation state matches policy across server estates, which is where it most often drifts.
Meltdown breaks the boundary between user space and the kernel by reading privileged memory before the permission check completes, and it was addressed effectively by kernel page table isolation and later by hardware changes. Spectre tricks a victim process into leaking its own data through mispredicted branches, which makes it a family of software-specific problems rather than one flaw with one fix.
Meltdown-style flaws were fixed in hardware from around 2019, but Spectre variants continue to be found, including Retbleed in 2022, Downfall in 2023, and Inception on AMD processors. Each generation adds hardware mitigations and each round of research finds a new path. Assume that speculative execution issues require ongoing microcode and kernel updates rather than a one-off remediation.
It depends entirely on the workload. Systems that make many system calls or context switches, such as databases and busy I/O servers, saw the largest impact from kernel page table isolation, sometimes in double-digit percentages on 2018-era hardware. Compute-bound workloads were barely affected. Newer processors with hardware fixes avoid most of the cost.
The original Spectre proof of concept ran in JavaScript, which is why browser vendors reduced timer precision, added jitter, and deployed site isolation so that content from different origins runs in separate processes. Those measures make practical exploitation much harder. Keeping browsers current and site isolation enabled is the effective control for this route.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.